Systems and methods for peak searching and artificial tissue response, including for processing functional response waveforms.

Abstract

A system and method for processing functional response waveforms obtained from artificial tissue are described herein. In one exemplary embodiment, the system and method includes obtaining a first waveform, which includes the contraction and relaxation responses of the artificial tissue during a contraction-relaxation cycle. A model that independently parameterizes the growth of the contraction and relaxation responses is then fitted to the first waveform. A second waveform is then generated from the model fitted to the first waveform, such that the second waveform includes a noise-filtered representation of the first waveform. Feature values(s) can then be extracted from the second waveform, which can be used to develop artificial intelligence (AI) or, otherwise, for downstream tasks such as drug discovery and development, by performing noise filtering and enabling high-fidelity and large synthetic training datasets.

Description

[Technical Field] 【0001】 This application claims the benefits of U.S. Provisional Application No. 63 / 457,574 (filed April 6, 2023), U.S. Provisional Application No. 63 / 466,105 (filed May 12, 2023), U.S. Provisional Application No. 63 / 505,257 (filed May 31, 2023), and U.S. Provisional Application No. 63 / 526,566 (filed July 13, 2023). Each of the aforementioned provisional applications in whole is incorporated herein by reference. [Background technology] 【0002】 Tissue behavior can be measured and modeled as a waveform of tissue response over time (e.g., functional response or contractile force). The resulting functional response waveform is a fundamental and valuable unit for representing and encoding tissue data. Representing and encoding tissue behavior as a waveform provides a rich foundation for downstream analysis. However, existing methods for characterizing such waveforms require extracting specific features from peaks within the waveform. Such feature extraction is often hindered by factors such as noise, and as a result, the effectiveness of these features used in downstream analysis is often limited. Furthermore, building models to accurately represent the underlying peaks within the waveform is often limited by the amount of in vitro data that can be used to train such models. 【0003】 Therefore, new methods are needed to process functional response data that enable accurate and efficient feature extraction. 【0004】 Furthermore, existing methods typically rely on tissues that exhibit regular or periodic contraction responses (e.g., due to external stimuli applied to the tissue). When tissues exhibit spontaneous or irregular contraction behavior, existing methods may fail to adequately identify individual contraction responses and extract relevant features. 【0005】 Therefore, there is a need for a new approach to detect and process spontaneous behavior within functional response data. 【0006】 Furthermore, many functional response waveforms may also include combinations of single contractions and double contractions (e.g., ectopic heartbeats). The ability to process such waveforms often depends on the ability to identify the contraction types present. 【0007】 Therefore, there is a need for a new approach to process functional response data that enables accurate and efficient feature extraction and effective classification of contraction types. 【0008】 Moreover, in some systems, tissue can be attached to a tissue scaffold within a device such that an image capturing the deflection of the tissue scaffold can be used to obtain a measurement of the tissue response at the time the image is captured. However, the conversion of the position of the tissue scaffold within an image frame to a measure of the contraction force can be inhibited by tissue scaffold misalignment or inaccurate tracking, or an inaccurate conversion of scaffold deflection to contraction force. 【0009】 Therefore, there is a need for a new approach to track the deflection of a tissue scaffold and then model the force exerted on the tissue scaffold. SUMMARY OF THE INVENTION 【0010】 A system and method for processing functional response waveforms. According to one aspect of the present disclosure, a method for processing a functional response waveform is provided. The method includes obtaining a first waveform including a contraction response and a relaxation response of an artificial tissue during a single contraction-relaxation cycle, fitting a model to the first waveform, the model independently parameterizing the growth of the contraction response and the relaxation response, and generating a second waveform from the model fitted to the first waveform such that the second waveform includes a noise-filtered representation of the first waveform. 【0011】 According to a further aspect of the present disclosure, a method for training a model using synthetic training data is provided. The method includes obtaining a plurality of waveforms including functional responses of one or more artificial tissues during a single contraction-relaxation cycle, and extracting a plurality of parameter sets from the plurality of waveforms, wherein the parameter sets among the plurality of parameter sets characterize corresponding waveforms of the plurality of waveforms. The method further includes determining a parameter set distribution from the plurality of parameter sets. The method further includes generating a synthetic training data set, wherein each element of the synthetic training data set includes a synthetic waveform and a corresponding parameter set used to generate the synthetic waveform, and the corresponding parameter set is obtained from the parameter set distribution. The method further includes training a prediction model using the synthetic training data set, wherein the prediction model is trained to estimate an output parameter set from an input waveform. 【0012】 According to an additional aspect of the present disclosure, a method for extracting a contraction-relaxation cycle waveform is provided. The method includes obtaining a first waveform including a plurality of functional responses of an artificial tissue stimulated at a first frequency, and convolving the first waveform with a pulse train by one or more processors to generate a convolved waveform, wherein the pulse train is generated at the first frequency. The method further includes identifying a first position associated with a maximum value of the convolved waveform, wherein the first position corresponds to an expected position of a first contraction-relaxation cycle, and extracting a second waveform including the first contraction-relaxation cycle from the first position of the first waveform, wherein the second waveform has a first duration proportional to the first frequency. 【0013】 A further aspect of the present disclosure provides a method for predicting therapeutic effects. The method comprises acquiring a plurality of signals, including a baseline signal and a perturbation signal, wherein the baseline signal comprises a first plurality of functional responses of the manipulated tissue under reference conditions, and the perturbation signal comprises a second plurality of functional responses of the manipulated tissue under perturbation conditions with a first perturbation. The method further comprises splitting the plurality of signals into a first plurality of waveforms, each waveform of the first plurality of waveforms comprising a contraction and relaxation response of the manipulated tissue during a single contraction-relaxation cycle. The method further comprises fitting a model to each waveform of the first plurality of waveforms, wherein the model independently parameterizes the growth of the contraction and relaxation response of the manipulated tissue during a single contraction-relaxation cycle for each waveform, and generating a second plurality of waveforms from the models fitted to each waveform of the first plurality of waveforms, wherein the second plurality of waveforms comprises a plurality of filtered baseline waveforms associated with the baseline signal and a plurality of filtered perturbation waveforms associated with the perturbation signal. The method further includes extracting a first feature value of a first feature from multiple filtered baseline waveforms, extracting a second feature value of the first feature from multiple filtered perturbation waveforms, and determining the effect associated with the first perturbation based on a comparison between the first and second feature values. 【0014】 Further aspects of this disclosure provide a method for processing functional response waveforms. The method comprises using one or more processors to acquire a first waveform including the contraction and relaxation responses of an artificial tissue during a single contraction-relaxation cycle, the first waveform being acquired from a bioreactor containing the artificial tissue. The method further comprises using one or more processors to fit a model to the first waveform, the model independently parameterizing the growth of the contraction and relaxation responses. The method further comprises using one or more processors to generate a second waveform from the model fitted to the first waveform, such that the second waveform includes a noise-filtered representation of the first waveform. The method further comprises using one or more processors to extract one or more feature values ​​from the second waveform. The method further comprises training a machine learning model on one or more feature values ​​from the second waveform. The method further comprises inputting a dataset of human tissue into the machine learning model to generate predictions defining one or more properties of human tissue, the predictions corresponding to at least one of the one or more feature values ​​of the second waveform. 【0015】 In accordance with the foregoing and herein disclosures, this disclosure includes applying certain features or embodiments using or by using certain machines, such as bioreactors. In various embodiments, a bioreactor may include a device configured to grow or manipulate human tissue (e.g., human tissue such as muscle tissue, cardiac tissue, and / or skeletal muscle tissue). Additionally or alternatively, a bioreactor may include a sensor assembly configured to detect one or more functional responses of tissue within the device. 【0016】 Furthermore, the disclosure includes converting or reducing a particular article to a different state or substance, for example, converting or reducing a functional response of, for example, human tissue in a bioreactor detected as a waveform by a sensor assembly to a different state or substance, for example, generating, creating, or more accurately fitting a model based on an improved waveform that has been filtered to remove errors (e.g., noise) in the original received or raw waveform signal detected (e.g., by one or more sensors) from an artificial tissue. 【0017】 Furthermore, this disclosure includes improvements in computer functions or other technologies, at least because the disclosure herein discloses systems and methods for reducing errors in an underlying computing device by, for example, producing an enhanced (e.g., a second) waveform filtered from data noise caused by measurement hardware (e.g., a sensor assembly that extracts data from artificial tissue). Furthermore, once the enhanced (e.g., a second) waveform is generated, fitted, or otherwise acquired, a high-fidelity training dataset can be trained or generated therefrom. Such a high-fidelity, large-volume synthetic training dataset can then be used to train or update models, such as new or updated machine learning models, to produce more accurate and error-free outputs, and as a result, to produce high-quality pharmaceuticals, such as therapeutic drugs. Furthermore, once the predictive model is deployed on the underlying system, the systems and methods of this disclosure enable the execution of a predictive model with fewer iterations and the use of fewer computing resources than the relevant systems and methods of the prior art. In other words, this disclosure describes improvements to the capabilities of the computer itself or “any other art or art field” because the increased predictive improvements provided by the predictive model enable the underlying computer system to utilize fewer processing and memory resources compared to prior art systems and methods. This is at least because a predictive model trained on an enhanced or second waveform with reduced error (less noise) can generate or determine with greater accuracy the probability that a given organization will exhibit a given behavior without requiring various tests and / or empirical computer simulations over a wide range of tests using multiple computation cycles and data. Thus, using a predictive model results in fewer computation cycles or other iterations with less impact on the underlying computing device compared to previous prior art systems and methods.In other words, the systems and methods of this disclosure are an improvement over the prior art, at least because the systems and methods of the prior art require empirical or trial-and-error methods, which may involve real-world testing on human tissues, resulting in and potentially requiring large database and memory utilization as well as processor usage to arrive at similar real-world or simulated results having the same or similar outcomes. On the other hand, the disclosed systems and methods describe the generation and / or use of bioreactors for tissue growth and testing to define a limited set of data specific to the tissue (e.g., human engineered tissue), which requires less memory usage and / or processing utilization compared to conventional methods that use or require large sets of unknown, potentially irrelevant data. Furthermore, the disclosures herein enable the identification and use of high-fidelity datasets, further reducing the need for additional computational cycles. 【0018】 In addition, this disclosure relates to improvements in other arts or technologies to provide robust, efficient, and comparable encoding of tissue behavior, which can be used to improve the efficiency and performance of several downstream drug discovery and development tasks. This can be carried out, for example, by predictive models trained or otherwise generated using spectral representations as training data that define the functional responses of tissues (e.g., artificial human tissues). The predictive models can be deployed on an underlying computing device or system, thereby improving accuracy and predictability in carrying out the drug discovery and development tasks described herein. 【0019】 Furthermore, this disclosure includes certain features that are well understood in the art, different from routine and customary activities, and / or otherwise, adds non-customary steps that limit this disclosure to specific useful applications, such as systems and methods for processing functional response waveforms of artificial tissues for the purpose of generating high-fidelity feature data that can be used, for example, to train more accurate models and / or to improve downstream tasks such as drug discovery and development. 【0020】 A system and method for detecting spontaneous tissue contraction. According to one aspect of the present disclosure, a method is provided for determining the spontaneous behavior of an engineered tissue. The method includes: obtaining a waveform containing the functional response of the engineered tissue over a period of time; applying a frequency-based global classifier to the waveform to generate a first classification score, the first classification score indicating whether the waveform contains periodic contractions of the engineered tissue over the period; if the first classification score indicates the absence of periodic contractions in the waveform, the method further includes: applying a local classifier to the waveform to generate a second classification score, the second classification score indicating whether the waveform contains spontaneous contractions of the engineered tissue over the period; and generating a behavioral profile of the engineered tissue over the period based on the second classification score. 【0021】 A further aspect of the present disclosure provides a system for determining the spontaneous behavior of manipulated tissue. The system comprises a bioreactor having a device configured to grow tissue, and a sensor assembly configured to detect one or more functional responses of the manipulated tissue within the device. The system further comprises a processing unit communicatively coupled to the bioreactor, the processing unit comprising one or more processors configured to: acquire a waveform containing the functional responses of the manipulated tissue within the device over a period of time; and generate a periodicity classifier applied to the waveform, thereby generating a first classification score, the first classification score indicating whether the waveform contains periodic contractions of the manipulated tissue over the period. If the first classification score indicates that there are no periodic contractions in the waveform, the one or more processors are configured to: apply a spontaneous contraction classifier to the waveform, thereby generating a second classification score, the second classification score indicating whether the waveform contains spontaneous contractions of the manipulated tissue over the period; and generate a behavioral profile of the manipulated tissue over the period based on the second classification score. 【0022】 In addition to the present disclosure, a non-temporary computer-readable medium is provided for storing instructions for determining the spontaneous behavior of an engineered tissue. When executed by one or more processors, the instructions cause one or more processors to: acquire a waveform containing the functional response of the engineered tissue over a period of time; apply a first classifier to the waveform to generate a first classification score, the first classification score indicating whether the waveform contains periodic contractions of the engineered tissue over the period. If the first classification score indicates that there are no periodic contractions in the waveform, the one or more processors are configured to further apply a second classifier to the waveform to generate a second classification score, the second classification score indicating whether the waveform contains spontaneous contractions of the engineered tissue over the period; and generate a behavioral profile of the engineered tissue over the period based on the second classification score. 【0023】 In accordance with the foregoing and herein disclosures, this disclosure includes applying certain features or embodiments using or by using certain machines, such as bioreactors. In various embodiments, a bioreactor may include a device configured to grow or manipulate human tissue (e.g., human tissue such as muscle tissue, cardiac tissue, and / or skeletal muscle tissue). Additionally or alternatively, a bioreactor may include a sensor assembly configured to detect one or more functional responses of tissue within the device. 【0024】 Furthermore, the disclosure includes converting or reducing a particular article to a different state or substance, for example, converting or reducing a functional response of, for example, human tissue in a bioreactor detected as a waveform by a sensor assembly to a different state or substance, for example, generating, creating, or otherwise developing a behavioral profile of an engineered tissue based on an originally received or raw waveform signal detected (for example, by one or more sensors) from an artificial tissue. 【0025】 Furthermore, this disclosure includes improvements in computer functions or other technologies, at least because the disclosure herein discloses a system and method for reducing errors in an underlying computing device by, for example, implementing a hierarchical system of multiple classifiers, wherein one classifier (e.g., a periodic classifier) ​​can filter waveforms known to exhibit periodic or regular behavior. This reduces the amount of data used by the system, and only waveforms that are not expected to have periodic or regular behavior are fed to subsequent local classifiers (e.g., spontaneous classifiers). This not only reduces the amount of data analyzed by the underlying computing system but also streamlines the system's ingestion of waveforms (or waveform data) that may exhibit spontaneous behavior, thereby providing improved identification of spontaneous behavior (and the location of spontaneous contractions), because the system will have fewer false positives, as it will analyze only a narrower or otherwise reduced set of questionable data. In other words, once waveforms (or waveform data) associated with the spontaneous behavior of an manipulated tissue are identified by the classifiers, such data can be selected (filtered) for analysis. Such waveform data produces more accurate and error-free outputs, and as a result, includes high-fidelity data for generating behavioral profiles of manipulated tissues in order to produce high-quality pharmaceuticals such as therapeutic drugs. 【0026】 Furthermore, when the classifier hierarchy is deployed on the underlying system, it enables the systems and methods of this disclosure to be executed with fewer iterations and use fewer computing resources than the relevant systems and methods of the prior art. In other words, this disclosure describes an improvement in the capabilities of the computer itself or “any other art or technical field” because the increased predictive improvement provided by the classifier hierarchy allows the underlying computer system to utilize fewer processing and memory resources compared to the systems and methods of the prior art. This is at least because the classifier hierarchy, designed to filter periodic waveform data to provide spontaneous waveform data with reduced errors (less noise), can generate or determine with greater accuracy the probability that a given organization exhibits a given behavior without requiring various tests and / or empirical computer simulations across a wide range of tests using multiple computation cycles and data. Thus, using the classifier hierarchy results in fewer computation cycles, or in other cases fewer iterations, compared to the aforementioned systems and methods of the prior art, which has less impact on the underlying computing device. In other words, the systems and methods of the present disclosure are an improvement over the prior art, at least because the systems and methods of the prior art require empirical or trial-and-error methods, which may involve real-world testing on human tissues, resulting in and potentially requiring large database and memory utilization as well as processor usage to achieve similar real-world or simulated results having the same or similar outcomes. On the other hand, the disclosed systems and methods describe the generation and / or use of a bioreactor for tissue growth and testing to define a limited set of data specific to the tissue (e.g., human engineered tissue), which requires less memory usage and / or processing utilization compared to conventional methods that use or require large sets of unknown, potentially irrelevant data.Furthermore, the disclosures herein enable the identification and use of high-fidelity datasets, further reducing the need for additional computational cycles. 【0027】 In addition, this disclosure relates to improvements in other arts or technologies to provide a robust, efficient, and comparable encoding of tissue behavior, which can be used to improve the efficiency and performance of several downstream drug discovery and development tasks. This can be carried out, for example, by a hierarchy of classifiers used to filter data of specific waveform types (e.g., spontaneous waveforms and / or related data) through classification, which can then be used to identify or define the functional response of a tissue (e.g., manipulated human tissue). The hierarchy of classifiers can be deployed on an underlying computing device or system, thereby improving accuracy, classification, and / or prediction when carrying out the drug discovery and development tasks described herein. 【0028】 Furthermore, this disclosure includes certain characteristics that are well understood in the art, and that differ from routine and customary activities, and / or otherwise, it adds non-customary steps that limit this disclosure to specific useful applications, such as systems and methods for determining the spontaneous behavior of an engineered tissue, which can be used, for example, to determine a more accurate behavioral profile(s) based on the engineered tissue, which can improve downstream tasks such as drug discovery and development. 【0029】 A system and method for classifying the type of shrinkage of manipulated tissue. According to one aspect of the present disclosure, a method for processing functional response waveforms is provided. The method comprises obtaining a first waveform including at least one contraction response and at least one relaxation response of an engineered tissue, the first waveform having a predetermined length corresponding to the expected length of a contraction-relaxation cycle of the engineered tissue. The method further comprises determining a predicted contraction type among a plurality of contraction types for the first waveform, the plurality of contraction types including a single contraction type and a dual contraction type. The method further comprises fitting a model to the first waveform based on the predicted contraction type, the model parameterizing the growth of at least one contraction response of the engineered tissue independently of the growth of at least one relaxation response of the engineered tissue. The method further comprises generating a second waveform from the model fitted to the first waveform such that the second waveform includes a noise-filtered representation of the first waveform. 【0030】 A further aspect of the present disclosure provides a method for training a classifier to predict tissue contraction types from functional response waveforms. The method comprises obtaining a plurality of waveforms, each of which has a predetermined length corresponding to the expected length of a contraction-relaxation cycle of the manipulated tissue. The method further comprises, by one or more processors, extracting a first plurality of parameter sets from a first subset of the plurality of waveforms associated with a single contraction type, wherein the first parameter set of the first plurality of parameter sets characterizes a first waveform of the first subset of waveforms. The method further comprises extracting a second plurality of parameter sets from a second subset of the plurality of waveforms associated with a dual contraction type, wherein the second parameter set of the second plurality of parameter sets characterizes a second waveform of the second subset of waveforms. The method further comprises determining a plurality of parameter set distributions, wherein the plurality of parameter set distributions include a first parameter set distribution determined from the first plurality of parameter sets and a second parameter set distribution determined from the second plurality of parameter sets. The method further comprises generating a synthetic training dataset, where each element of the synthetic training dataset includes a synthetic waveform and a corresponding tissue contraction type associated with the synthetic waveform, and the synthetic waveform is generated using a parameter set distribution from a plurality of parameter set distributions associated with the corresponding tissue contraction type. The method further comprises training a classifier using the synthetic training dataset, where the classifier trained using the synthetic training dataset determines the predicted tissue contraction type of the input waveform. 【0031】 An additional aspect of the present disclosure provides a method for processing functional response waveforms. The method includes acquiring a first waveform comprising multiple functional responses of an artificial tissue stimulated at a first frequency, and convolving the first waveform with a pulse train to generate a convolutional waveform, wherein the pulse train is generated at the first frequency. The method further includes identifying a first position associated with a first maximum value of the convolutional waveform, the first position corresponding to an expected position of a first contraction-relaxation cycle. The method further includes extracting a second waveform comprising a first contraction-relaxation cycle from the first position of the first waveform, the second waveform having a first duration proportional to the first frequency. 【0032】 A further aspect of the present disclosure provides a method for predicting perturbation effects. The method comprises acquiring a plurality of signals, including a baseline signal and a perturbation signal, wherein the baseline signal comprises a first plurality of functional responses of the manipulated tissue under control conditions, and the perturbation signal comprises a second plurality of functional responses of the manipulated tissue under a first set of perturbation conditions. The method further comprises splitting the plurality of signals into a first plurality of waveforms, each waveform of the first plurality of waveforms having a predetermined length and comprising at least one contraction response and at least one relaxation response of the manipulated tissue, the predetermined length corresponding to the expected length of a contraction-relaxation cycle of the manipulated tissue. For each of the first plurality of waveforms, the method further comprises determining a predicted contraction type from a plurality of contraction types, wherein the plurality of contraction types include a single contraction type and a dual contraction type. The method further comprises fitting a model to each waveform of the first plurality of waveforms based on the corresponding predicted contraction type, the model parameterizing the growth of at least one contraction response independently of the growth of at least one relaxation response of the manipulated tissue. The method further comprises generating a second set of waveforms from a model fitted to each of a first set of waveforms, the second set of waveforms comprising a set of filtered baseline waveforms associated with a baseline signal and a set of filtered perturbation waveforms associated with a perturbation signal. The method further comprises extracting a first feature value of the first feature from the set of filtered baseline waveforms, extracting a second feature value of the first feature from the set of filtered treatment waveforms, and determining the effect associated with a first set of perturbation conditions based on a comparison of the first and second feature values. 【0033】 According to additional aspects of the present disclosure, a method is provided for training a model using synthetic training data. The method includes: obtaining a plurality of waveforms, each containing a functional response of one or more prosthetic tissues during a single contraction-relaxation cycle; and extracting a plurality of parameter sets from the plurality of waveforms, wherein the parameter sets among the plurality of parameter sets characterize the corresponding waveforms of the plurality of waveforms. The method further includes: determining a parameter set distribution from the plurality of parameter sets; and generating a synthetic training dataset, wherein each element of the synthetic training dataset includes a synthetic waveform and a corresponding parameter set used to generate the synthetic waveform, the corresponding parameter set being obtained from the parameter set distribution. The method further includes: training a parameter estimation model using the synthetic training dataset, wherein the parameter estimation model is trained to estimate an output parameter set from an input waveform. 【0034】 In accordance with the foregoing and herein disclosures, this disclosure includes applying certain features or embodiments using or by using certain machines, such as bioreactors. In various embodiments, a bioreactor may include a device configured to grow or manipulate human tissue (e.g., human tissue such as muscle tissue, cardiac tissue, and / or skeletal muscle tissue). Additionally or alternatively, a bioreactor may include a sensor assembly configured to detect one or more functional responses of tissue within the device. 【0035】 Furthermore, the disclosure includes converting or reducing a particular article to a different state or substance, for example, converting or reducing a functional response of human tissue in a bioreactor, detected as a waveform by a sensor assembly, to a different state or substance, for example, generating, creating, or otherwise developing a noise-filtered second waveform based on an originally received or raw waveform signal detected (e.g., by one or more sensors) from artificial tissue using a fitting model. 【0036】 Furthermore, this disclosure includes improvements in computer functions or other technologies, at least because the disclosure herein discloses a system and method for reducing errors in an underlying computing device by implementing a contraction-relaxation cycle model, for example, in which a first waveform of an operated tissue is improved by noise filtering using model fitting to eliminate or reduce data usage, and also enables the generation of high-fidelity, large-scale synthetic training datasets. The contraction-relaxation cycle model is also efficiently extensible to model different contraction types (e.g., functional response waveforms including single or double contraction responses). Thereafter, the amount of data used by the system is reduced, and the noise-reduced (and therefore data-reduced) waveform can be used as the output of a final process, such as the output of an improved downstream task, such as drug discovery and development. This not only reduces the amount of data analyzed by the underlying computing system, but also streamlines the system's input to waveforms (or waveform data) that have both contraction and relaxation responses, because the system will have fewer false positives since it analyzes only filtered or otherwise reduced sets of data that rely on synthetic (controlled) data as determined by the model. Such data is also extensible, as it allows the data to predict, classify, or otherwise output different contraction types (e.g., functional response waveforms including single or dual contraction responses). In other words, once waveforms (or waveform data) related to the relaxation and contraction responses of manipulated tissue are input, such data can be input into a model (e.g., a classifier model) to determine or otherwise classify multiple contraction types (e.g., single contraction type and / or dual contraction type) identified by the model, and such data can then be selected (filtered) for analysis.Such waveform data includes high-fidelity data, and the input waveform (e.g., the first waveform) is fitted to a model to parameterize the growth of the contraction response of the manipulated tissue, which is independent of the growth of the relaxation response of the manipulated tissue, thereby generating a second noise-filtered version of the input (first) waveform. This generates a more accurate and error-free output, which in turn makes it possible to produce high-quality pharmaceuticals such as therapeutic drugs. 【0037】 Furthermore, when the contraction-relaxation cycle model is deployed on an underlying system, it enables the systems and methods of this disclosure to be executed with fewer iterations and use fewer computing resources than the relevant systems and methods of the prior art. In other words, this disclosure describes an improvement in the capabilities of the computer itself or "any other art or technical field" because the increased predictive improvement provided by the contraction-relaxation cycle model allows the underlying computer system to utilize fewer processing and memory resources compared to systems and methods of the prior art. This is at least because the contraction-relaxation cycle model, designed to filter periodic waveform data to provide waveform data with reduced errors (less noise), can generate or determine with greater accuracy the probability that a given organization exhibits a given behavior without requiring various tests and / or empirical computer simulations across a wide range of tests using multiple computational cycles and data. Thus, using a classifier hierarchy results in fewer computational cycles, or in other cases fewer iterations, compared to the aforementioned systems and methods of the prior art, which has less impact on the underlying computing device. In other words, the systems and methods of the present disclosure are an improvement over the prior art, at least because the systems and methods of the prior art require empirical or trial-and-error methods, which may involve real-world testing on human tissues, resulting in and potentially requiring large database and memory utilization as well as processor usage to arrive at similar real-world or simulated results having the same or similar outcomes.On the other hand, the disclosed systems and methods describe the generation and / or use of bioreactors for tissue growth and testing to define a limited set of data specific to an organization (e.g., human engineered tissue), which requires less memory usage and / or processing utilization compared to conventional methods that use or require large sets of unknown, potentially irrelevant data. Furthermore, the disclosure herein enables the identification and use of high-fidelity datasets, further reducing the need for additional computational cycles. 【0038】 In addition, this disclosure relates to improvements in other arts or technologies to provide a robust, efficient, and comparable encoding of tissue behavior, which can be used to improve the efficiency and performance of several downstream drug discovery and development tasks. This can be carried out, for example, by a contraction-relaxation cycle model (e.g., a type of classifier) ​​used to filter data of a particular waveform type (e.g., waveform and / or related data) through classification, which can then be used to identify, classify, or define contraction types (single and / or dual contraction types) of manipulated human tissue. The contraction-relaxation cycle model can be deployed on an underlying computing device or system, thereby improving its accuracy, classification, and / or prediction when carrying out the drug discovery and development tasks described herein. 【0039】 Furthermore, this disclosure includes certain features that are well understood in the art, and that differ from routine and customary activities, and / or otherwise, it adds non-customary steps that limit this disclosure to specific useful applications, such as systems and methods for classifying the contraction type of manipulated tissue, which can be used, for example, to determine noise-filtered and / or model-fitted waveforms(s) based on manipulated tissue, which can improve downstream tasks such as drug discovery and development. 【0040】 Systems and methods for tracking organizational scaffolding. According to one aspect of the present disclosure, a system is provided for modeling the contraction deflection of a flexible tissue scaffold. The system comprises a bioreactor having a flexible scaffold for attachment to biological tissue, the flexible scaffold being positioned to deflect in response to contraction forces acting thereon. The bioreactor further comprises an imaging device configured to acquire one or more images of the flexible scaffold. The system further comprises a processing unit communicatively coupled to the bioreactor. The processing unit is configured to acquire from the bioreactor a plurality of images of the flexible scaffold at a plurality of time points, the plurality of images capturing the deflection of the flexible scaffold along a first dimension at a plurality of time points, due to contraction forces acting thereon. The processing unit is further configured to fit the plurality of curves to the plurality of images such that each curve extends along the centerline of the flexible scaffold in each of the plurality of images, and to determine a plurality of displacement values ​​from the plurality of curves, each of the plurality of displacement values ​​including a measurement between each of the plurality of curves and a reference line extending along a second dimension perpendicular to the first dimension. The processing unit is further configured to generate a model based on multiple displacement values, which characterizes the contraction forces acting on the flexible scaffold over multiple time points. 【0041】 A further aspect of the present disclosure provides a method for modeling contraction deflection of a flexible tissue scaffold. The method comprises acquiring a plurality of images of the flexible scaffold at a plurality of time points, the plurality of images capturing the deflection of the flexible scaffold along a first dimension resulting from contraction forces acting on it at the plurality of time points. The method further comprises fitting the plurality of curves to the plurality of images such that each curve extends along the centerline of the flexible scaffold in each of the plurality of images, and determining a plurality of displacements from the plurality of curves, wherein each of the plurality of displacement values ​​includes a measurement along the first dimension between each of the plurality of curves and a reference line extending along a second dimension perpendicular to the first dimension. The method further comprises generating a model based on the plurality of displacement values, the model characterizing the contraction forces acting on the flexible scaffold over the plurality of time points. 【0042】 In an additional aspect of the present disclosure, a non-temporary computer-readable medium storing instructions is provided, and when the instructions are executed by the processing unit, the processing unit is caused to acquire a plurality of images of a flexible scaffold at a plurality of time points in time, the plurality of images capturing the deflection of the flexible scaffold along a first dimension resulting from contraction forces acting on it at a plurality of time points in time; to fit a plurality of curves to the plurality of images such that each curve extends along the centerline of the flexible scaffold in each of the plurality of images; to determine a plurality of displacement values ​​from the plurality of curves, each of which includes a measurement along the first dimension between each of the plurality of curves and a reference line extending along a second dimension perpendicular to the first dimension; and to generate a model based on the plurality of displacement values, the model characterizing the contraction forces acting on the flexible scaffold over a plurality of time points in time. 【0043】 In accordance with the foregoing and herein disclosures, this disclosure includes applying certain features or embodiments using or by using certain machines, such as bioreactors. In various embodiments, a bioreactor may include a device configured to grow or manipulate human tissue (e.g., human tissue such as muscle tissue, cardiac tissue, and / or skeletal muscle tissue). Additionally or alternatively, a bioreactor may include a sensor assembly configured to detect one or more functional responses of tissue within the device. 【0044】 Furthermore, the disclosure includes converting or reducing a particular article to a different state or object, for example, converting or reducing the displacement value of a flexible scaffold in a bioreactor detected by an imaging assembly to a different state or object, for example, generating, creating, or otherwise determining the contractile force of manipulated tissue attached to the flexible scaffold tissue. 【0045】 Furthermore, this disclosure includes improvements in computer functions or other technologies, at least because the disclosure herein discloses systems and methods for reducing errors in an underlying computing device by enabling, for example, the accurate tracking, efficient extraction and measurement of deflection of a tissue scaffold using a flexible scaffold, thereby enabling the generation of a model characterizing the contractile force(s) that produced the deflection. The underlying system can be updated with a model, which can be used to encode or calibrate the relationship between contractile force and measured displacement, thereby improving the accuracy of contractile force measurements obtained from such a model. Improvements to such a model also improve the efficiency and performance of the computing system used to generate and deploy such a model, while providing improvements to downstream tasks that utilize such a model. This makes it possible to generate more accurate and error-free outputs, and as a result, to generate high-quality pharmaceuticals such as therapeutic drugs. 【0046】 Furthermore, when the contraction-relaxation cycle model is deployed on the underlying system, it enables the systems and methods of this disclosure to be executed with fewer iterations and use fewer computing resources than the relevant systems and methods of the prior art. In other words, this disclosure describes an improvement in the capabilities of the computer itself or "any other art or technical field" because the increased predictive improvement provided by implementing a model (e.g., a force-displacement model) generated for a particular device and / or tissue scaffold allows any variation in the contraction response of the tissue scaffold to be modeled within the force-displacement model, thereby improving the consistency of the force measurements obtained from the displacement values ​​provided to the model. Thus, using a model (e.g., a force-displacement model) results in fewer computational cycles, or in other cases fewer iterations, compared to the aforementioned systems and methods of the prior art, which has less impact on the underlying computing device. In other words, the systems and methods of this disclosure are an improvement over the prior art, at least because the systems and methods of the prior art require empirical or trial-and-error methods, which may involve real-world testing on human tissues, resulting in and potentially requiring large database and memory utilization as well as processor usage to arrive at similar real-world or simulated results having the same or similar outcomes. On the other hand, the disclosed systems and methods describe the generation and / or use of bioreactors for tissue growth and testing to define a limited set of data specific to the tissue (e.g., human engineered tissue), which requires less memory usage and / or processing utilization compared to conventional methods that use or require large sets of unknown, potentially irrelevant data. Furthermore, the disclosures herein enable the identification and use of high-fidelity datasets, further reducing the need for additional computational cycles. 【0047】 In addition, this disclosure relates to improvements in other arts or technical fields, at least because the systems and methods of this disclosure can be used to improve the efficiency and performance of several downstream drug discovery and development tasks, providing a robust, efficient, and comparable tissue measurement solution. This can be carried out, for example, by a flexible scaffold and an imaging device configured to generate a model based on displacement values, thereby allowing the model to characterize the contractile forces acting on the flexible scaffold. The model can then be used to identify, classify, or define tissues, such as manipulated human tissue. The model can be deployed on an underlying computing device or system, thereby improving accuracy, classification, and / or prediction when carrying out the drug discovery and development tasks described herein. 【0048】 Furthermore, this disclosure includes certain features that are well understood in the art, and that differ from routine and customary activities, and / or otherwise, it adds non-customary steps that limit this disclosure to specific useful applications, such as systems and methods for modeling the contraction deflection of a flexible tissue scaffold and for generating models that characterize the contraction forces acting on a flexible tissue scaffold over multiple time points, which can improve downstream tasks such as drug discovery and development. 【0049】 The advantages will become clearer to those skilled in the art from the following description relating to preferred embodiments illustrated and described by examples. As should be understood, embodiments of the present invention may also be other different embodiments, and their details are modifiable in various ways. Therefore, the drawings and description should be considered illustrative rather than limiting. 【0050】 Hereinafter, embodiments of the present disclosure will be described with reference to the attached drawings, merely as examples. [Brief explanation of the drawing] 【0051】 [Figure 1]This disclosure describes a system for processing functional response data according to the aspects described herein. [Figure 2] The waveforms shown include the functional response of the manipulated tissue obtained from the system of Figure 1 according to the aspects of this disclosure. [Figure 3] A contraction-relaxation cycle model according to an embodiment of this disclosure is illustrated. [Figure 4] An exemplary waveform having a corresponding contraction-relaxation cycle model fitting according to an aspect of this disclosure is shown. [Figure 5A] The effects of each parameter in the contraction-relaxation cycle model according to embodiments of this disclosure are illustrated. [Figure 5B] The effects of each parameter in the contraction-relaxation cycle model according to embodiments of this disclosure are illustrated. [Figure 6] This embodiment of the disclosure shows a parameter estimation model for fitting the parameters of a contraction-relaxation-cycle model. [Figure 7A] The elements of the parameter estimation model shown in Figure 6 according to an embodiment of this disclosure are illustrated as an example. [Figure 7B] The elements of the parameter estimation model shown in Figure 6 according to an embodiment of this disclosure are illustrated as an example. [Figure 7C] The elements of the parameter estimation model shown in Figure 6 according to an embodiment of this disclosure are illustrated as an example. [Figure 7D] The elements of the parameter estimation model shown in Figure 6 according to an embodiment of this disclosure are illustrated as an example. [Figure 8] A method for processing functional response waveforms according to aspects of this disclosure is shown. [Figure 9] This disclosure describes a method for fitting a model to a functional response waveform according to an embodiment of this disclosure. [Figure 10] This disclosure describes a method for training a parameter estimation model using synthetic training data. [Figure 11] Embodiments of this disclosure describe a method for predicting a set of parameter values ​​using the parameter estimation model training method shown in Figure 10. [Figure 12] A method for extracting contraction-relaxation cycle waveforms according to aspects of this disclosure is shown. [Figure 13] This disclosure describes a method for extracting further contraction-relaxation cycle waveforms according to embodiments of this disclosure. [Figure 14] A method for predicting therapeutic effects using a contraction-relaxation cycle model is described according to aspects of this disclosure. [Figure 15A] The functional response of tissue exhibiting periodic contraction behavior, as described in this disclosure, is shown. [Figure 15B] The functional response of tissue exhibiting spontaneous contraction behavior, as described in this disclosure, is shown. [Figure 16] This disclosure describes a system for determining the spontaneous behavior of an manipulated organization. [Figure 17A] The spectral reactions of different shrinkage reactions according to embodiments of this disclosure are shown. [Figure 17B] The spectral reactions of different shrinkage reactions according to embodiments of this disclosure are shown. [Figure 18] This disclosure describes an embodiment of a convolutional neural network for predicting contraction behavior. [Figure 19] The sequential results of implementing the modified Pan-Tompkins algorithm according to embodiments of this disclosure are illustrated. [Figure 20] This invention illustrates adaptive thresholding of waveforms according to an embodiment of this disclosure. [Figure 21] This disclosure provides a method for determining the spontaneous behavior of an manipulated organization. [Figure 22] This disclosure describes a method for obtaining a classification score from a waveform according to an embodiment of this disclosure. [Figure 23] This disclosure describes a method for obtaining a classification score from a waveform according to an embodiment of this disclosure. [Figure 24] The modified Pans-Tompkins method according to embodiments of this disclosure is shown. [Figure 25]A dual contraction-type contraction-relaxation cycle model according to an embodiment of this disclosure is illustrated. [Figure 26] This disclosure describes a system for processing functional response waveforms having different contraction types, according to embodiments of this disclosure. [Figure 27] The following describes the classification of contraction types for three waveforms according to embodiments of this disclosure. [Figure 28] This disclosure describes a method for processing functional response waveforms according to an aspect of this disclosure. [Figure 29] This disclosure describes a method for fitting a model to a functional response waveform according to an embodiment of this disclosure. [Figure 30] This paper describes a method for training a classifier to predict tissue contraction type from functional response waveforms. [Figure 31] Embodiments of this disclosure describe a method for predicting the tissue contraction type of a waveform using a synthetically trained classifier. [Figure 32] This disclosure describes a method for training a parameter estimation model using synthetic training data. [Figure 33] A method for extracting single or dual contraction-relaxation cycle waveforms according to aspects of this disclosure is shown. [Figure 34] Embodiments of this disclosure describe a method for extracting further single or double contraction-relaxation cycles from a waveform. [Figure 35] This disclosure describes a method for predicting perturbation effects according to the aspects described herein. [Figure 36] Figure 1 shows the wells of a bioreactor, such as the bioreactor shown in this disclosure, according to an embodiment of this disclosure. [Figure 37] Exemplary images of tissue scaffolds under different contractile forces according to embodiments of the present disclosure are shown. [Figure 38] This disclosure describes a method for tracking flexible scaffolding according to the embodiments described herein. [Figure 39] An embodiment of the present disclosure shows a one-dimensional vascular enhancement filter. [Figure 40]This shows a plot of data points obtained from a transformed image of a tissue scaffold according to an embodiment of the present disclosure. [Figure 41] The generated model corresponding to the time series of scaffolding deflection values ​​according to the embodiments of this disclosure is shown. [Figure 42] An embodiment of the present disclosure shows a flexible scaffold that flexes due to a predetermined force applied upward by a probe. [Figure 43] The plot of the force-displacement model according to the embodiments of this disclosure is shown. [Figure 44] A method for modeling the contraction deflection value of a flexible tissue scaffold according to aspects of this disclosure is provided. [Figure 45] This disclosure describes an embodiment of a method for fitting multiple curves. [Figure 46] This invention describes an embodiment of a method for fitting a reference line. [Figure 47] An exemplary computing system for carrying out the method of this disclosure, according to one embodiment, is shown. [Modes for carrying out the invention] 【0052】 Technical field This disclosure relates to processing functional response waveforms. More specifically, and not exclusively, this disclosure relates to processing functional response waveforms using a contraction-relaxation cycle model. More specifically, and not exclusively, this disclosure relates to generating a noise-filtered representation of functional response waveforms using a contraction-relaxation cycle model. Again, more specifically, and not exclusively, this disclosure relates to identifying effects associated with perturbations in manipulated tissue based on the noise-filtered functional response waveforms generated using a contraction-relaxation cycle model. 【0053】 In addition, this disclosure relates to detecting tissue behavior. Specifically, although not exclusive, this disclosure relates to detecting spontaneous tissue contraction. Specifically, although not exclusive, this disclosure relates to detecting spontaneous tissue contraction in the functional response waveform of manipulated tissue. 【0054】 Furthermore, this disclosure relates to classifying and processing functional response waveforms. More specifically, but not exclusively, this disclosure relates to classifying and processing functional response waveforms using a contraction-relaxation cycle model. More specifically, but not exclusively, this disclosure relates to generating a noise-filtered representation of functional response waveforms using a contraction-relaxation cycle model based on identified contraction types. Again, more specifically, but not exclusively, this disclosure relates to identifying effects associated with perturbations in manipulated tissue based on the noise-filtered functional response waveforms generated using a contraction-relaxation cycle model. 【0055】 Furthermore, this disclosure relates to the modeling of tissue scaffolds. Specifically, although not exclusive, this disclosure relates to the modeling of contraction deflection of flexible tissue scaffolds. Specifically, although not exclusive, this disclosure relates to the generation of a model characterizing the contraction forces acting on a flexible tissue scaffold over multiple time points. 【0056】 Detailed explanation The ability to accurately and efficiently extract features from functional response waveforms of manipulated tissues is a critical step when using such waveforms for downstream tasks such as drug discovery and development. Existing methods are often limited by the quality and / or quantity of available data. Furthermore, because the functional responses of manipulated tissues can exhibit different response types, fitting a single model to all types of responses is difficult. This disclosure presents a system and method for peak searching and artificial tissue responses to overcome such problems. 【0057】 Figure 1 shows a system 100 for processing functional response data or spontaneous tissue contraction data according to various aspects of the present disclosure. Additionally or alternatively, system 100 also illustrates the modeling of contraction deflection of a flexible tissue scaffold according to aspects of the present disclosure. 【0058】 System 100 comprises a bioreactor 102 and a control unit 104. The bioreactor 102 comprises a device 106 for growing manipulated tissue, a sensor assembly 108, and an interface 110. In one embodiment, the sensor assembly 108 forms part of the device 106, but such elements may be separate in other embodiments. The control unit 104 may comprise a model fitting unit 104-1, a signal processing unit 104-2, and a signal processing unit 112 (which may also be referred to as a signal processor). Additionally or alternatively, the control unit 104 may comprise the signal processing unit 112 (which may alternatively be referred to as a processing unit, signal processor, or processor). The processing unit may include one or more processors, for example, one or more processors described herein with respect to Figure 47. The interface 110 connects the bioreactor 102 and the control unit 104 in a communicative manner so that data can be exchanged between the bioreactor 102 and the control unit 104. In various embodiments, bioreactors are used to grow tissue (for example, manipulated tissue such as human tissue, including, in non-limiting examples, muscle tissue, cardiac tissue, skeletal muscle tissue, or other human tissues 124)). 【0059】 As shown in enlarged portion 106-1 of device 106, device 106, i.e., the substrate, comprises one or more wells, such as well 114; one or more cell culture wells, such as cell culture well 116; a pair of electrodes, including a first electrode 118-1 and a second electrode 118-2; and a pair of elements, e.g., a second scaffold, including a first element 120-1 (e.g., a first scaffold) and a second element 120-2 (e.g., a scaffold). Well 114 is positioned within the cell culture well 116 and has a bottom on device 106, a first end 122-1, and a second end 122-2. Well 114 is configured to grow an engineered tissue 124 from cells seeded therein. Culture medium may be added to the cell culture well 116 to grow and / or maintain the engineered tissue 124. Alternatively referred to as artificial tissue, the manipulated tissue 124 includes manipulated muscle tissue. In one embodiment, the manipulated muscle tissue is manipulated cardiac tissue. For example, the manipulated cardiac tissue is Cellular Dynamics International (CDI) iCell Cardiomyocytes 2 This is generated using a side population of normal human ventricular fibroblasts embedded in a hydrogel composed of fibrin (Sigma-Aldrich), collagen (Sigma-Aldrich), and Matrigel (Corning). In an alternative embodiment, iCell Cardiomyocytes 2 Instead, any other human cardiomyocytes, such as Axol Bioscience Human iPSC-Derived Ventricular Cardiomyocytes or Sigma-Aldrich Human Cardiac Myocytes (HCM), may be used. Additionally, or alternatively, fibrin and / or collagen are omitted from the hydrogel. In alternative embodiments, the manipulated muscle tissue is manipulated skeletal muscle tissue. 【0060】 A pair of electrodes are separated by a gap in which a well 114 is positioned. The pair of electrodes is configured to apply electrical stimulation to a cell culture in one or more wells of the device 106 (e.g., the manipulated tissue 124 in well 114 shown in enlarged section 106-1). While the cell culture matures in the device 106, the pair of electrodes apply stimulation to the cell culture according to an electrical stimulation protocol that lasts for several weeks. After the cell culture has matured, the pair of electrodes may be configured to stimulate the cell culture (e.g., the manipulated tissue 124) at a set frequency, i.e., a pacing frequency. In one embodiment, the frequency at which the cell culture is stimulated, i.e., the pacing frequency, is set by a control unit 104. The control unit 104 may then send a command 126 to a bioreactor 102, causing the bioreactor 102 to stimulate the manipulated tissue(s) in the device 106 at the set pacing frequency. 【0061】 In some exemplary embodiments, the first element 120-1 and the second element 120-2 are arranged across the well 114 such that a gap exists between the bottom of the well 114 and the pair of elements. The first element 120-1 and the second element 120-2 are configured to (a) allow the attachment of the manipulated tissue 124 formed between them, thereby suspending the manipulated tissue 124 above the bottom of the well 114, and (b) deform in response to the contractile force exerted on the pair of elements by the manipulated tissue 124, thereby simulating the physiological environment inherent to the manipulated tissue 124 and / or to allow measurement of the contractile force (e.g., by a sensor assembly 108). For example, a pair of electrodes may apply an electrical stimulus to the manipulated tissue 124 at a frequency of 0.1 Hz. The manipulated tissue 124 contracts in response to this electrical stimulus, causing deformation of at least one of the first element 120-1 and the second element 120-2. By measuring the deformation of one or more pairs of elements, it becomes possible to record the functional response of the manipulated tissue 124 when stimulated at 0.1 Hz. 【0062】 The sensor assembly 108 is configured to detect one or more functional responses of the manipulated tissue within the device 106 (e.g., one or more functional responses of the manipulated tissue 124). In one embodiment, the sensor assembly 108 includes an optical sensor. The optical sensor is configured to detect deformation of a first element 120-1 and / or a second element 120-2 (e.g., resulting from a contraction force exerted on the first element 120-1 and / or the second element 120-2 by the manipulated tissue 124). By detecting the deformation of the first element 120-1 and / or the second element 120-2, it becomes possible to determine one or more functional responses, such as displacement or contraction displacement of the manipulated tissue 124, or contraction force of the manipulated tissue 124. Additionally or alternatively, the optical sensor of the sensor assembly 108 is configured to detect the fluorescence intensity of the manipulated tissue 124. By detecting the fluorescence intensity of the manipulated tissue 124, it becomes possible to determine one or more functional responses, such as a transient calcium reaction or a change in the membrane potential of the manipulated tissue 124. Additionally or alternatively, the optical sensor of the sensor assembly 108 is configured to detect changes in the dimensions of the manipulated tissue 124 over a time frame. By detecting changes in the dimensions of the manipulated tissue 124 over a time frame, it becomes possible to determine one or more functional responses, such as displacement or contraction displacement of the manipulated tissue 124, or contraction force of the manipulated tissue 124. 【0063】 The optical sensor of the sensor assembly 108 is configured to acquire multiple image-based representations of one or more functional responses of tissue (e.g., manipulated tissue 124) over a time frame or period. For example, the optical sensor may be configured to capture images, or frames, of tissue every n seconds, where n is associated with a predetermined frequency (rate) at which images or frames are captured. For example, if n=1, one image of tissue is captured per second. Any preferred value of n, e.g., n={1 / 60, 1 / 50, 1 / 30, 1 / 24, 1 / 12, 1 / 4, 1 / 2, 1, 2}, and similar values ​​may be selected. In one embodiment, the frame rate is determined according to the frequency at which the tissue in the device is stimulated. Thus, a sequence of images or frames of tissue over a time frame will capture one or more functional responses of the tissue over that time frame. 【0064】 In one embodiment, the bioreactor 102 is configured to convert a series of images capturing one or more functional responses of tissue over a time frame into a waveform representation. For example, the sensor assembly 108, interface 110, or other components of the bioreactor 102 may process images within the series of images to extract features relating to the tissue's functional response at a point in time relevant to the image. The features relating to the tissue's functional response extracted from the series of images may then be combined to form a waveform containing one or more functional responses of the tissue over time. A waveform, such as waveform 128, may then be output from the bioreactor 102. In an alternative embodiment, raw image or frame data is output from the bioreactor 102, and another unit, such as a control unit 104 or an image analysis unit (not shown), processes this data to determine time-series functional response data. 【0065】 The time-series functional response data (e.g., waveform 128) produced by the bioreactor 102 provides a high-fidelity, high-information-density representation of the functional response of the manipulated tissue over a predetermined period (e.g., contractile force produced by the first tissue in response to a pacing frequency of 1 Hz over a period of 30 seconds). To obtain such waveform data from the bioreactor 102 for the manipulated tissue 124, a command may be sent to the bioreactor 102 (e.g., from the control unit 104) to begin stimulating the manipulated tissue 124 at the pacing frequency. Alternatively, no electrical stimulation may be applied to observe the spontaneous response of the manipulated tissue 124. The sensor assembly 108 then captures a series of images of the manipulated tissue 124 over a predetermined period. Crucially, the series of images capture the response (e.g., deformation) of the first element 120-1 and / or the second element 120-2 as a result of the contractile response of the manipulated tissue 124. Next, the series of images are processed to extract the responses of the first element 120-1 and / or the second element 120-2 over a predetermined period, and to convert the element responses over time into a time series of functional responses over time. For example, the displacement of the first element 120-1 and / or the second element 120-2 may be used to determine the force of the contraction response of the manipulated tissue 124. The time series (e.g., waveform 128) is then output from the bioreactor 102 for further processing and / or analysis. 【0066】 Tissue within the bioreactor 102 (e.g., manipulated tissue 124) can be periodically administered a drug or compound, and the functional response of the tissue after administration can be recorded in the form of waveforms. Alternatively, the functional response of the tissue in the absence of any external administration regimen can be periodically recorded. In either case, changes in the functional response of the tissue are captured in subsequent waveforms, as illustrated in Figure 2. 【0067】 Figure 2 shows waveform 202, which includes the functional response of the tissue over a predetermined period. 【0068】 Waveform 202 is obtained from a bioreactor such as the bioreactor 102 shown in Figure 1, as described above. Waveform 202 provides a high-fidelity and high-information-density representation of the tissue's functional response over time frames t1 to t2. Waveform 202 contains multiple peaks (alternatively referred to as contraction or contraction-relaxation cycles) corresponding to the tissue's contraction response over time frames (e.g., 30 seconds, 40 seconds, etc.). A magnified view of a single contraction-relaxation cycle (i.e., a peak) within portion 204 of waveform 202 is shown in magnified portion 204-1. 【0069】 Typically, multiple features are extracted from each contraction-relaxation cycle of waveform 202 to characterize the waveform 202, thereby enabling further processing or analysis of the waveform 202. As shown in enlarged section 204-1, features extracted from a single contraction-relaxation cycle include the peak (or single contraction) amplitude 206, the time to peak amplitude 208 (or contraction time), the time to peak decline 210 (or relaxation time), the duration 212 (contraction-relaxation cycle duration or single contraction duration), the maximum development rate 214 (or maximum contraction slope), the maximum decline rate 216 (or maximum relaxation slope), and the passive tension 218. 【0070】 While the above features provide rich characterization of each contraction-relaxation cycle (i.e., of waveform 202), the noise level of the underlying signal often makes accurate extraction or measurement of these features difficult. For example, noise introduced into the underlying waveform as a result of measurement or transmission (e.g., noise introduced by sensor assembly 108) can lead to inaccuracies in identifying the key points necessary to extract the above features. These inaccuracies consequently limit the effectiveness and results of downstream tasks that utilize such features. Furthermore, in vitro functional response data is often limited in quantity, thereby limiting the applicability of such data to training predictive models such as machine learning regression or classification models. 【0071】 This disclosure aims to address some, if not all, of the above problems by introducing a contraction-relaxation cycle model. This model is used to enable noise filtering and the generation of high-fidelity, large-volume synthetic training datasets, as described in more detail below. Such high-fidelity, large-volume synthetic training datasets can then be used to train, update, or otherwise improve artificial intelligence models, such as new or updated machine learning models, to produce more accurate and error-free outputs. Such outputs can be used in the downstream drug discovery, development, and / or manufacture of high-quality pharmaceuticals, such as therapeutic drugs. In some embodiments, this disclosure presents systems and methods for processing functional response waveforms to enable efficient and effective feature extraction. 【0072】 A system and method for processing functional response waveforms. This disclosure presents a system and method for processing functional response waveforms to enable efficient and effective feature extraction. 【0073】 Figure 3 illustrates elements of a contraction-relaxation cycle model according to embodiments of the present disclosure. In some embodiments, the contraction-relaxation cycle model may include a single contraction-type contraction-relaxation cycle model. 【0074】 Figure 3 shows the contraction function 302, the relaxation function 304, and the contraction-relaxation cycle model 306. For brevity, the contraction-relaxation cycle model 306 is also referred to as the single contraction model, single contraction type model, single type model, or single model. The contraction-relaxation cycle model 306 can include a combination or product of the contraction function 302 and the relaxation function 304. The contraction-relaxation cycle model 306 includes a combination or product of the contraction function 302 and the relaxation function 304. Therefore, the contraction-relaxation cycle model 306 independently models the contraction and relaxation responses of the contraction-relaxation cycle in the manipulated tissue. Specifically, the growth rate of the contraction response (i.e., the growth rate of the contraction function 302) and the growth rate or decay rate of the relaxation response (i.e., the growth rate or decay rate of the relaxation function 304) vary independently, thereby enabling the contraction-relaxation cycle model 306 to model a wide variety of different response types. Therefore, the contraction-relaxation cycle model 306 can be used effectively and efficiently across a range of application areas involving various types of functional response waveforms. Additionally or alternatively, the growth rate of the contraction response (i.e., the growth rate of the contraction function 302) and the growth rate or decay rate of a single relaxation response (i.e., the growth rate or decay rate of the relaxation function 304) can vary independently, thereby enabling the single contraction model 306 to model a wide variety of single contraction-relaxation responses. 【0075】 According to one aspect of the present disclosure, a contraction-relaxation cycle model, such as contraction-relaxation cycle model 306, can be fitted to in vitro functional response data of an engineered tissue, such as the contractile force waveform of the engineered cardiac tissue. A noise-filtered representation of the in vitro functional response data is then generated from the contraction-relaxation cycle model, and features are precisely extracted from the noise-filtered representation. 【0076】 The contraction and relaxation functions are logistic (sigmoid) functions, bi-exponential functions, or any other suitable functions. In one embodiment, a logistic function is used to model a functional response waveform based on force, while a bi-exponential function is used to model a functional response waveform based on calcium transients. Generally, a contraction-relaxation cycle model (e.g., contraction-relaxation cycle model 306) uses the contraction function f c (t) and relaxation function f r This includes the product with (t). 【number】 Here, A is the maximum value of the contraction-relaxation cycle model, and B is the shift of the contraction-relaxation cycle model applied along the y-axis. In the embodiment shown in Figure 3, the contraction function 302 is an ascending logistic function of the following form: 【number】 The relaxation function 304 is a descending logistic function of the following form: 【number】 Therefore, the contraction-relaxation cycle model 306 can be expressed as follows: 【number】 Here, t0 and k c This parameterizes the contraction function 302, corresponding to the midpoint of the contraction function 302 and the growth rate of the contraction function 302 (the reciprocal of the logistic growth rate). Parameter t d and k r This parameterizes the relaxation function 304, corresponding to the midpoint of the relaxation function 304 and the growth rate or decay rate of the relaxation function 304 (the reciprocal of the logistic growth rate or decay rate). 【0077】 Therefore, the parameter set θ = [A, B, t0, t d , k c , k r fully describes the contraction-relaxation cycle of the manipulated tissue, thereby enabling the contraction-relaxation cycle model 306 to fit and approximate the contraction-relaxation cycle(s) within the real-world functional response data obtained from the manipulated tissue (e.g., artificial heart tissue). 【0078】 The contraction-relaxation cycle model illustrated in FIG. 3 can be used to model functional responses having a single contraction type, but not all functional responses follow a single contraction type. For example, arrhythmias such as atrial fibrillation or tachycardia can cause ectopic beats and result in an irregular beat pattern. Thus, the functional response waveform can include combinations of both a single contraction type as described in relation to FIG. 3 above, and a dual contraction type as described in relation to FIG. 25 below. 【0079】 FIG. 4 shows three exemplary waveforms to which the corresponding contraction-relaxation cycle model fits. 【0080】 FIG. 4 shows a first waveform 402, a second waveform 404, and a third waveform 406. FIG. 4 further shows a first model-fitted waveform 408, a second model-fitted waveform 410, and a third model-fitted waveform 410. Each of the model-fitted waveforms corresponds to a waveform generated from the contraction-relaxation cycle model to one of the first, second, or third waveforms. For example, the first model-fitted waveform 408 corresponds to a waveform generated from the contraction-relaxation cycle model fitted to the first waveform 402. 【0081】 FIGS. 5A and 5B show the effect of each parameter of the contraction-relaxation cycle model according to embodiments of the present disclosure. 【0082】 The plots in Figures 5A and 5B show a contraction-relaxation cycle model according to equation (4) above, with one fluctuating parameter (e.g., a single contraction-type contraction-relaxation cycle model). In some embodiments, as described above, the parameter changes of the single contraction-relaxation cycle model shown in Figures 5A and 5B can also be applied to the parameter changes of a dual contraction-type model, for example, as described in relation to Figure 25 below. 【0083】 Plot A in Figure 5A shows the contraction-relaxation cycle model with varying values ​​for the maximum parameter A. Plot B in Figure 5A shows the contraction-relaxation cycle model with varying values ​​for the shift parameter B. Plot C in Figure 5A shows the contraction-relaxation cycle model with varying values ​​for the contraction midpoint parameter t0. Plot D in Figure 5A shows the contraction midpoint parameter t d Figure 5B shows a contraction-relaxation cycle model with varying values ​​of k. Plot E in Figure 5B shows the contraction-growth rate parameter k. c The contraction-relaxation cycle model with varying values ​​of k is shown. Plot F in Figure 5B shows the relaxation growth rate parameter k. r This shows a contraction-relaxation cycle model with varying values. 【0084】 The range and variation of shapes that can be captured by both or either of single and / or dual contraction-relaxation cycles means that the model(s) can accurately fit a range of different functional response data. In particular, complex responses can be modeled efficiently and accurately by independently modeling the contraction and relaxation(s) responses, for example, by independently modeling the growth and decay of one or both responses within a given cycle(s). Thus, the contraction-relaxation models(s) of this disclosure can be fitted to a range of different functional response data by identifying the contraction types in the functional response data and determining which contraction type model to fit, and / or additionally or alternatively, by determining the parameters that best fit the model to the underlying data. 【0085】 An example of this process is shown in Figure 26, as described herein. 【0086】 As a further example, the contraction-relaxation model can be fitted using any suitable parameter fitting technique, such as least-squares-based fitting, or machine learning-based fitting, such as those shown in Figure 6 or elsewhere in this specification. 【0087】 Figure 6 shows a parameter estimation model 600 for fitting parameters to a contraction-relaxation cycle model according to an embodiment of the present disclosure. 【0088】 The parameter estimation model 600 includes a machine learning model in the form of a deep neural network. The parameter estimation model 600 comprises a residual layer 602, a convolutional neural network 604, an expanded convolutional network 606, a connected block 608, a long short-term memory (LSTM) network 610, and a fully connected network 612. The parameter estimation model 600 receives an input vector 614 and produces an output vector 616. 【0089】 The input vector 614 contains a sequence (time series) corresponding to a waveform (e.g., waveform 202 shown in Figure 2) that includes the contraction and relaxation responses of an artificial tissue during a single contraction-relaxation cycle. The input vector 614 is fed to the residual layer 602, the convolutional neural network 604, and the inflated convolutional network 606. The residual layer 602 is a two-dimensional (2D) convolutional layer with 64 filters of size 1. The convolutional neural network 604 and the inflated convolutional network 606 are described in more detail below in relation to Figures 7A and 7B. The outputs of the residual layer 602, the convolutional neural network 604, and the inflated convolutional network 606 are concatenated in a concatenation block 608. The connected block 608 includes a deep connected layer that connects the outputs of the residual layer 602, the convolutional neural network 604, and the inflated convolutional network 606, a batch normalization layer that normalizes the output of the deep connected layer, and an exponential linear unit layer that performs an identity operation on positive inputs and α(exp(x)-1) on negative inputs x, where α=1. The output of the connected block 608 (i.e., the output of the exponential linear unit layer) is fed to the LSTM and network 610, which is described in more detail in relation to Figure 7C below. The output of the LSTM and network 610 is fed to the fully connected network 612, which is described in more detail in relation to Figure 7D below. The output vector 616 produced by the fully connected network 612 contains parameter values ​​of a contraction-relaxation-cycle model that is fitted to the waveform represented by the input vector 614. 【0090】 Beneficially, by using expanded convolution and LSTM networks, the parameter estimation model 600 can efficiently and accurately fit parameters to waveforms containing complete contraction-relaxation cycles. This efficiency and accuracy leads to more efficient use of computing resources and improved performance in downstream tasks such as drug discovery and development that utilize the estimated parameters. The parameter estimation model also reduces errors that would otherwise be present in the original waveform, for example, produced by the underlying computing device. Thus, the error-reduced parameter estimation model operates with greater efficiency, reducing additional computational cycles on the underlying computing device (e.g., one or more processors and / or memory of the basic device), and thus saving processor and memory utilization on the device on which the parameter estimation model is run. 【0091】 In one embodiment, the parameter estimation model 600 is trained using a synthetic dataset containing 40,000 training samples and 10,000 validation samples. The synthetic dataset is generated using the method described in relation to Figure 10 below. To train the parameter estimation model 600, a mini-batch gradient descent method with a batch size of 128 is used with an ADAM solver. The ADAM solver has an initial learning rate of 1e-3 and terminates early based on the validation loss. 【0092】 Additionally, or alternatively, in some embodiments, the parameter estimation model 600 in Figure 6 may include a predictive model. In such embodiments, the parameter estimation model 600 may be referred to herein as the predictive model 600. 【0093】 In such an embodiment, the prediction model 600 includes a machine learning model in the form of a deep neural network. The prediction model 600 comprises a residual layer 602, a convolutional neural network 604, an expanded convolutional network 606, a connected block 608, a long short-term memory (LSTM) network 610, and a fully connected network 612. The prediction model 600 receives an input vector 614 and produces an output vector 616. 【0094】 The predictive model 600 is trained for different purposes. By modifying the fully connected network 612, as described in more detail below, the predictive model 600 is trained to either predict contraction types or predict the parameters of a contraction-relaxation cycle model. When predicting model parameters, the predictive model 600 is trained to predict either single contraction type or dual contraction type model parameters. Thus, the predictive model 600 shown in Figure 6 and described in detail in relation to Figures 7A-7D can be trained as a parameter estimation model or a contraction type classifier. 【0095】 The input vector 614 includes a sequence (time series) corresponding to a waveform containing at least one contraction response and at least one relaxation response of the artificial tissue. The waveform represented by the input vector 614 has either a single contraction type or a double contraction type. In the case of a single contraction type, the input vector 614 includes a time series corresponding to a single contraction response. In the case of a double contraction type, the input vector includes a time series corresponding to a double contraction response. 【0096】 The input vector 614 is supplied to the residual layer 602, the convolutional neural network 604, and the inflated convolutional network 606. In one embodiment, the residual layer 602 is a two-dimensional (2D) convolutional layer having 64 filters of size 1. The convolutional neural network 604 and the inflated convolutional network 606 are described in more detail with reference to Figures 7A and 7B below. The outputs of the residual layer 602, the convolutional neural network 604, and the inflated convolutional network 606 are concatenated in a concatenation block 608. The concatenation block 608 includes a depth concatenation layer that concatenates the outputs of the residual layer 602, the convolutional neural network 604, and the inflated convolutional network 606, a batch normalization layer that normalizes the output of the depth concatenation layer, and an exponential linear unit layer that performs an identity operation on positive inputs and α(exp(x)-1) on negative inputs x. In one embodiment, α=1. The output of the connected block 608 (i.e., the output of the exponential linear unit layer) is supplied to the LSTM network 610, which is described in more detail in relation to Figure 7C below. The output of the LSTM network 610 is supplied to the fully connected network 612, which is described in more detail in relation to Figure 7D below. 【0097】 As described in more detail below, the architecture of the fully connected network 612 and its final activation capabilities are modified depending on the application of the predictive model 600. As a result, the output vector 616 produced by the fully connected network 612 contains vectors or scalar scores associated with the contraction type classification when the predictive model 600 is used for contraction type classification. For example, the output vector 616 may contain the probability that the waveform represented by the input vector 614 is a single contraction type or a double contraction type. When the predictive model 600 is used for parameter estimation, the output vector 616 produced by the fully connected network 612 contains vectors of parameter values ​​for a single or double contraction-relaxation cycle model fitted to the waveform represented by the input vector 614. 【0098】 Beneficially, by using expanded convolution and LSTM networks, the predictive model 600 can accurately predict the type of waveform contraction and efficiently fit parameters to waveforms containing either single or double contraction types. These advantages lead to more efficient use of computing resources, as well as improved performance on downstream tasks such as drug discovery and development. 【0099】 In one embodiment, the predictive model 600 is trained for the task of contraction type classification using a training dataset of labeled waveforms. A training dataset of approximately 10,000 training samples and approximately 2,000 validation samples is used. The training dataset includes approximately 6,000 manually labeled single-contraction type contraction-relaxation cycle waveforms and 6,000 manually labeled dual-contraction type contraction-relaxation cycle waveforms. 【0100】 In another embodiment, the predictive model 600 is trained using a synthetic dataset containing 40,000 training samples and 10,000 validation samples. The synthetic dataset is generated using the method described in relation to Figure 29 below. When training the predictive model 600 for systolic type classification, the synthetic dataset contains 25,000 single-systolic type waveforms and 25,000 double-systolic type waveforms. For the systolic type classification task, each waveform in the synthetic dataset is associated with a corresponding label indicating whether the waveform is a single-systolic type or a double-systolic type. When training the predictive model 600 for parameter estimation, the synthetic data contains either single-systolic type waveforms or double-systolic type waveforms, along with the corresponding parameters for each waveform. Thus, for parameter estimation, one model is trained to predict the parameters of single-systolic type waveforms, and another model is trained to predict the parameters of double-systolic type waveforms. 【0101】 For both the classification and parameter estimation tasks in both of the embodiments described above, a mini-batch gradient descent method with a batch size of 128 is used with an ADAM solver to train the predictive model 600. The ADAM solver has an initial learning rate of 1e-3 and terminates early based on the validation loss. 【0102】 Figure 7A shows a convolutional neural network 700 that forms part of the parameter estimation model 600 (e.g., prediction model) of Figure 6, according to an embodiment of the present disclosure. 【0103】 The convolutional neural network 700 corresponds to the convolutional neural network 604 of the parameter estimation model 600 (e.g., prediction model) shown in Figure 6. The convolutional neural network 700 comprises a first inception network 702, a connected block 704, and a second inception network 706. The first inception network 702 comprises a first inception module 708-1, a second inception module 708-2, a third inception module 708-3, a fourth inception module 708-4, and a fifth inception module 708-5. Each of these inception modules has the same architecture, as illustrated in relation to the first inception module 708-1, which comprises a first 2D convolutional layer 710, a batch normalization layer 712, an exponential linear unit layer 714, and a second 2D convolutional layer 716. The connected block 704 comprises a depth connected layer 722, a batch normalization layer 724, and an exponential linear unit layer 726 with α=1. The second inception network 706 has the same architecture as the first inception network 702 (i.e., the same number and configuration of inception modules). 【0104】 Each inception module in the first inception network 702 and the second inception network 706 shares the same architecture (i.e., as illustrated in relation to the first inception module 708-1 shown in Figure 7A) and is parameterized according to the filter size of the second 2D convolutional layer (e.g., the second 2D convolutional layer 716). The filter sizes of the second 2D convolutional layer in the first inception module 708-1, the second inception module 708-2, the third inception module 708-3, the fourth inception module 708-4, and the fifth inception module 708-5 are [3, 5, 7, 9, 11], respectively. The first 2D convolutional layer (e.g., the first 2D convolutional layer 710) comprises 64 filters of size 1. The batch normalization layer 712 includes a standard batch normalization layer, and the exponential linear unit layer 714 includes a standard exponential linear unit layer with α=0.1. 【0105】 Figure 7B shows an expanded convolutional network 730 that forms part of the parameter estimation model 600 (e.g., a prediction model) of Figure 6, according to an embodiment of the present disclosure. 【0106】 The expanded convolutional network 730 corresponds to the expanded convolutional network 606 of the parameter estimation model 600 (e.g., the prediction model) shown in Figure 6. The expanded convolutional network 730 comprises a first expanded convolutional module 730-1, a second expanded convolutional module 730-2, a third expanded convolutional module 730-3, a fourth expanded convolutional module 730-4, a fifth expanded convolutional module 730-5, a sixth expanded convolutional module 730-6, and a final expanded 2D convolutional layer 732. 【0107】 The architecture of each inflated convolutional module is the same and is illustrated in Figure 7B by the architecture of the second inflated convolutional module 730-2, which comprises an inflated 2D convolutional layer 734, a batch normalization layer 736, and an exponential linear unit layer 738 with α=0.1. Each inflated convolutional module is parameterized by the size of the inflated operation performed in the inflated 2D convolutional layer. The inflated sizes of the inflated 2D convolutional layers of the first inflated convolutional module 730-1, the second inflated convolutional module 730-2, the third inflated convolutional module 730-3, the fourth inflated convolutional module 730-4, the fifth inflated convolutional module 730-5, and the sixth inflated convolutional module 730-6 are [1, 2, 4, 8, 16, 32], respectively. 【0108】 The dilation operation is illustrated in the first expansion section 740, which shows the input section 742 to the second dilation convolution module 730-2 and section 744 of the dilation 2D convolution layer 734, where the input corresponds to the output of the exponential linear unit layer of the first dilation convolution module 730-1. Dilation convolution beneficially increases the receptive field of the layer (i.e., the period the layer processes) without increasing the number of parameters. The dilation 2D convolution layer 734 implements this expansion by inserting zeros between each filter element. As shown in section 744 of the dilation 2D convolution layer 734, a dilation size or coefficient of 2 results in a single zero being padded between each filter element (whereas a dilation size or coefficient of 4 would result in three zeros being inserted). 【0109】 The final expanded 2D convolutional layer 732 corresponds to a standard expanded 2D convolutional layer with an expansion size or coefficient of 64. 【0110】 Figure 7C shows a long short-term memory (LSTM) network 746 that forms part of the parameter estimation model 600 (e.g., prediction model) of Figure 6, according to an embodiment of the present disclosure. 【0111】 The LSTM network 746 corresponds to the LSTM network 610 of the parameter estimation model 600 (e.g., prediction model) shown in Figure 6. The LSTM network 746 comprises an LSTM module 748, a first high-density module 750, and a second high-density module 752. The LSTM network 746 comprises a smoothing layer 754, a first bidirectional LSTM layer 756, a first dropout layer 758, a second bidirectional LSTM layer 760, and a second dropout layer 762. The first high-density module 750 comprises a first high-density layer 764, a first exponential linear unit layer 766 with α=1, a first dropout layer 768, a second high-density layer 770, a second exponential linear unit layer 772 with α=1, a second dropout layer 774, and a smoothing layer 776. The architecture of the second high-density module 752 is the same as that of the first high-density module 750. 【0112】 The LSTM module 748 attempts to learn bidirectional long-term dependencies between time steps of waveform (i.e., time-series) data via bidirectional LSTM layers. Therefore, the LSTM module 748 can learn from the complete waveform at each time step. Both the first bidirectional LSTM layer 756 and the second bidirectional LSTM layer 760 have 64 hidden units. The first dropout layer 758 and the second dropout layer 762 help reduce overfitting by randomly setting the input to zero with a probability of 0.1. 【0113】 The first high-density layer 764 and the second high-density layer 770 of the first high-density module 750 have an output size of 256. The first dropout layer 768 and the second dropout layer 774 randomly set the input to zero with a probability of 0.5. As described above, the architecture of the second high-density module 752 is the same as the architecture of the first high-density module 750. 【0114】 Figure 7D shows a fully connected network 778 that forms part of the parameter estimation model 600 (e.g., prediction model) of Figure 6, according to an embodiment of the present disclosure. 【0115】 The fully connected network 778 corresponds to the fully connected network 612 of the parameter estimation model 600 (e.g., the prediction model) shown in Figure 6. The fully connected network 778 comprises a connecting layer 780, a first high-density block 782-1, a second high-density block 782-2, a third high-density block 782-3, a high-density layer 784, and / or an output layer 786. 【0116】 The first high-density block 782-1, the second high-density block 782-2, and the third high-density block 782-3 all have the same architecture. As shown in relation to the first high-density block 782-1, the architecture comprises a high-density layer 786, an exponential linear unit layer 788 with α=1, and a dropout layer 790. The high-density layer of the first high-density block 782-1 and the second high-density block 782-2 has an output size of 512. The high-density layer of the third high-density block 782-3 has an output size of 1024. The dropout layer of the first high-density block 782-1 and the second high-density block 782-2 randomly sets the input to zero with a probability of 0.5. The dropout layer of the third high-density block 782-3 randomly sets the input to zero with a probability of 0.1. 【0117】 In one example, the high-density layer 784 has an output size corresponding to the number of parameters that the parameter estimation model should fit. Setting the output size to 6 gives the complete set of parameters for the contraction-relaxation cycle model described above in relation to Figure 3, θ=[A, B, t0, t d , k c , k r This is learned. Therefore, the output of the high-density layer 784 corresponds to the output of the parameter estimation model (for example, the output vector 616 of the parameter estimation model 600 shown in Figure 6). 【0118】 In one embodiment, the estimated parameters produced by the parameter estimation model are further refined using optimization techniques. As will be described in more detail below in relation to Figure 9, the optimization techniques used include simplex search algorithms such as the Nelder-Mead method. 【0119】 Once a contraction-relaxation cycle model is fitted to data, the waveform can be reconstructed from the model to generate a noise-filtered representation of the waveform. For example, several points (e.g., 100, 200, 500, 1000, etc.) are sampled from the contraction-relaxation cycle model fitted to the data over the duration of the first waveform to which the contraction-relaxation cycle model is fitted. Beneficially, this makes it possible to generate high-resolution waveform data from the contraction-relaxation cycle, which helps ensure that more accurate features of the functional response of the manipulated tissue are extracted. This, in turn, helps improve the accuracy and effectiveness of downstream tasks involving functional response features. The contraction-relaxation cycle model also reduces errors that would otherwise be present in the original waveform, for example, produced by the underlying computing device. Thereafter, the error-reduced contraction-relational cycle model operates with greater efficiency, reducing the additional computational cycles of the underlying computing device (e.g., one or more processors and / or memory of the basic device), and thus saving processor and memory utilization of the device on which the contraction-relational cycle model is run. 【0120】 In another example, when a parameter estimation model 600 (e.g., prediction model 600) formed in part by a fully connected network 778 is used for contraction type classification, the high-density layer 784 has two nodes, one for each contraction type, so that the output of each node corresponds to the probability that the input waveform belongs to the respective contraction type. When the prediction model 600 is used for parameter estimation, the high-density layer 784 has an output size (i.e., the number of nodes) corresponding to the number of parameters that the parameter estimation model should fit. For a single contraction type model, the output size (i.e., the number of nodes in the high-density layer 784) is set to 6, and the complete set of parameters for the single contraction type contraction-relaxation cycle model described above in relation to Figure 3 θ=[A, B, t0, t d , k c , k r Learns ]. For a dual contraction-type model, the output size (i.e., the number of nodes in the high-density layer 784) is set to 11, and the complete set of parameters for the dual contraction-type contraction-relaxation cycle model described above in relation to Figure 25. 【number】 Learn about it. 【0121】 The output of the high-density layer 784 corresponds to the output of the parameter estimation model (for example, the output vector 616 of the prediction model 600 shown in Figure 6). 【0122】 When the prediction model described above (e.g., prediction model 600 in Figure 6) is used for contraction type classification, the predicted contraction type is used to determine which of the two parameter estimation models fits the waveform (as illustrated by system 2600 in Figure 26). In one embodiment, the predicted contraction type provides a type of representation of the state of the manipulated tissue (e.g., disease state, treatment state, etc.), as illustrated in Figure 27. 【0123】 Here, we will explain a method for using the contraction-relaxation cycle model described above to process functional response waveforms. 【0124】 Figure 8 shows a method 800 for processing a functional reaction according to an aspect of the present disclosure. 【0125】 Method 800 includes the steps of acquiring a first waveform in step 802, fitting a model to the first waveform in step 804, and generating a second waveform from the model in step 806. Method 800 further includes an optional step 808 for extracting feature values ​​from the second waveform and an optional step 810 for outputting the feature values. In some embodiments, the extracted feature values ​​may be used to train a machine learning model, which has increased predictive accuracy by using the feature values ​​of the second waveform, as the second waveform itself filters out error-prone data (e.g., noise), which leads to improved predictive accuracy. Predictions may then be generated by inputting a dataset of human tissue into the machine learning model. The predictions may define one or more characteristics of the human tissue, which may correspond to at least one of the one or more feature values. In this way, a highlighted accurate machine learning, or other artificial intelligence model, may be trained or generated using the feature values ​​of a waveform with reduced error (e.g., the second waveform). In one embodiment, method 800 is carried out by the model fitting unit 104-1 of the control unit 104 shown in Figure 1. 【0126】 Generally, Method 800 is used to generate a noise-filtered or noise-suppressed representation of a contraction-relaxation cycle waveform (functional response waveform). In embodiments, the contraction-relaxation cycle waveform is obtained from hardware such as a bioreactor (e.g., bioreactor 102 shown in Figure 1), which may introduce noise into the waveform due to sensor variability, signal transmission, signal conversion, and the like. Thus, the contraction-relaxation cycle waveform contains a potentially noisy representation of the functional response of the manipulated tissue during a single contraction-relaxation cycle. Beneficially, Method 800 efficiently generates a noise-filtered representation of a single contraction-relaxation cycle, thereby improving the accuracy of the extracted features and consequently helping to improve the performance of downstream tasks that utilize such features. 【0127】 In acquisition step 802, a first waveform is acquired that includes the contraction and relaxation responses of the prosthetic tissue during a single contraction-relaxation cycle. 【0128】 The first waveform (i.e., a single contraction-relaxation cycle waveform, a single cycle waveform, a single contraction cycle waveform, a peak, or a functional response waveform) captures the functional response of the prosthetic tissue during a single contraction-relaxation cycle, which includes a contraction period (i.e., the period during which the prosthetic tissue contracts and generates tension) and a relaxation period (i.e., the period during which the prosthetic tissue returns to its normal state or normal length). Thus, the contraction response of the first waveform includes the functional response of the prosthetic tissue during the contraction period of the contraction-relaxation cycle, and the relaxation response includes the functional response of the prosthetic tissue during the relaxation period of the single contraction-relaxation cycle. The functional response is the contractile force of the prosthetic tissue (e.g., the contractile force measured using data acquired from the sensor assembly 108 of the bioreactor 102, in which the prosthetic tissue grows or is maintained). Alternatively, the functional response may be a contraction displacement, a transient calcium response, or a change in membrane potential. 【0129】 Artificial or manipulated tissue includes manipulated muscle tissue such as manipulated cardiac tissue or manipulated skeletal muscle tissue. In one embodiment, the first waveform is obtained from a bioreactor containing the artificial tissue (e.g., bioreactor 102 shown in Figure 1). Thus, the first waveform is obtained from a waveform obtained from the bioreactor, which includes multiple contraction-relaxation cycles of the artificial tissue. In one embodiment, the first waveform is obtained or extracted from the waveform using an extraction method such as method 1200, which is described in more detail below in relation to Figure 12. 【0130】 In fitting step 804, the model is fitted to the first waveform. The model, the contraction-relaxation cycle model, parameterizes the growth of the contraction and relaxation responses independently. Therefore, the model does not assume that the underlying functional response is symmetric. This allows the model to efficiently and accurately fit a wide variety of functional responses from diverse manipulated tissue types. More efficient and accurate model fitting helps generate more accurate features, which in turn leads to the generation of better data. 【0131】 The model uses the contraction function f c (t) and relaxation function f r (t) is included. In one embodiment, the contraction function is an ascending logistic function with a positive growth rate, and the relaxation function is a descending logistic function with a negative growth rate. Thus, the contraction response of the contraction-relaxation cycle is modeled by the ascending logistic function, and the relaxation response of the contraction-relaxation-cycle is modeled by the descending logistic function. The model includes the product of the ascending logistic function and the descending logistic function. 【0132】 The contraction-relaxation cycle model includes multiple parameters associated with the contraction and relaxation responses. These multiple parameters include the maximum value parameter A and the rate of increase parameter k. c , decline rate parameter k r, y-shift parameter B, upward x-shift parameter t0, and downward x-shift parameter t d This includes the relationship of each of these parameters to the overall model, which is shown and described in more detail in relation to Figures 5A and 5B above. 【0133】 In one embodiment, the model is fitted to a first waveform using predictions of parameter values ​​obtained from a machine learning model (e.g., the parameter estimation model shown in Figure 6), as described in more detail in relation to method 900 in Figure 9 below. 【0134】 In step 806, the second waveform is generated from a model fitted to the first waveform such that the second waveform contains a noise-filtered representation of the first waveform. For example, several points (e.g., 100, 200, 500, 1000, etc.) are sampled from the contraction-relaxation cycle model fitted to the data over the duration of the first waveform to which the contraction-relaxation cycle model is fitted. Beneficially, this makes it possible to generate high-resolution waveform data from the contraction-relaxation cycle, which helps ensure that more accurate features of the functional response of the manipulated tissue are extracted. This, in turn, helps improve the accuracy and effectiveness of downstream tasks involving functional response features. 【0135】 Optionally, a second waveform may be output. In one embodiment, outputting the second waveform includes storing or saving the second waveform in persistent storage such as non-volatile memory, non-temporary media, or the like. Additionally or alternatively, outputting the second waveform includes transmitting the second waveform over a network (e.g., a local area network, a wide area network, and the like) or displaying the waveform for user review. 【0136】 In the optional extraction step 808, one or more feature values ​​are extracted from the second waveform. 【0137】 The second waveform is a noise-filtered representation of the first waveform (i.e., a noise-filtered representation of the contraction-relaxation cycle), thus allowing for the extraction of more accurate values ​​of the underlying functional response features. One or more feature values ​​include one or more of the following: single contraction, or peak, amplitude value (i.e., peak amplitude 206 shown in Figure 2), contraction time value (i.e., time to peak amplitude 208 shown in Figure 2), maximum contraction slope value (i.e., maximum development rate 214 shown in Figure 2), relaxation time value (i.e., time to peak decline 210 shown in Figure 2), maximum relaxation slope value (i.e., maximum decline rate 216 shown in Figure 2), and single contraction duration value (i.e., duration 212 shown in Figure 2). 【0138】 Once extracted, one or more feature values ​​can be used as quantitative descriptors of the contraction-relaxation cycle, thus providing a numerical representation of the functional response of the artificial tissue during the contraction-relaxation cycle. As described in more detail in relation to Figure 14, such features can be used in various downstream processing tasks, such as efficacy identification in drug discovery and development. 【0139】 In an optional output step 810, one or more feature values ​​extracted from the second waveform are output. In one embodiment, outputting one or more feature values ​​includes storing or saving one or more feature values ​​in persistent storage such as non-volatile memory, non-temporary media, or the like. Additionally or alternatively, outputting one or more feature values ​​includes transmitting one or more feature values ​​over a network (e.g., a local area network, a wide area network, and the like) or displaying one or more feature values ​​for user review. 【0140】 Figure 9 shows a method 900 for fitting a model to a functional response waveform according to an embodiment of the present disclosure. 【0141】 In one embodiment, method 900 is performed as part of fitting step 804 of method 800. Method 900 includes step 902 of predicting multiple values ​​for multiple parameters and further includes an optional step 904 of optimizing the multiple values. 【0142】 The steps in Method 900 are used to predict the parameter values ​​of the contraction-relaxation cycle model from the first waveform. A trained machine learning model (e.g., the parameter estimation model shown in Figure 6) is used to predict the parameter values ​​so that the fitted model closely approximates the first waveform. 【0143】 In prediction step 902, multiple values ​​for multiple parameters of the model are predicted so that the model fitted to the first waveform contains multiple values ​​for multiple parameters. 【0144】 In the optional optimization step 904, the multiple values ​​determined in the prediction step 902 are optimized by minimizing the error between the first waveform and the model fitted to the first waveform (i.e., using multiple values ​​for multiple parameters of the model). 【0145】 The model determined in prediction step 902 【number】 If multiple values ​​are given for the value, the updated multiple values 【number】 teeth, 【number】 This is determined in step 904, which optimizes to be such that L θ(·) represents the first waveform X1 and the model f fitted to the first waveform according to the set of parameter values ​​θ. θ This is a cost function or loss function that measures the error between the first waveform and the second waveform. A smaller value of L indicates that the model is better fitted to the first waveform. In one embodiment, the cost function L is the root mean square error as follows: 【number】 In one embodiment, this corresponds to the following: 【number】 【0146】 Optimizing θ is a multidimensional problem because the model involves multiple parameters. Therefore, L θ Minimizing requires the simultaneous fitting of multiple parameters. To perform this minimization, in one embodiment, multiple values ​​are optimized using a simplex search algorithm such as the Nelder-Mead method. Beneficially, performing the optimization after obtaining initial predictions of the parameters helps to obtain an accurate model fit while making better use of processing resources, because the optimization process starts with a solution that is expected to be close to the optimal solution. 【0147】 Figure 10 shows Method 1000, which illustrates a method for training a parameter estimation model using synthetic training data according to an aspect of the present disclosure. 【0148】 Method 1000 includes the steps of acquiring multiple waveforms (step 1002), extracting multiple parameter sets (step 1004), determining the parameter set distribution (step 1006), generating a synthetic training dataset (step 1008), and training a predictive model on the synthetic training dataset (step 1010). In one embodiment, Method 1200 is carried out by a control unit 104 or a subunit thereof, as shown in Figure 1. 【0149】 In many situations, the effectiveness of a predictive model (e.g., a machine learning model such as a deep neural network) is limited by the quantity and quality of the available training data. Without large amounts of high-quality training data, predictive models will often be unable to produce adequate outputs. In this disclosure, this problem leads to the fitting of inaccurate contraction-relaxation models, which will then reduce the effectiveness and applicability of such models when actually used for tasks such as drug discovery and development. Method 1000 attempts to address such a problem by generating high-fidelity synthetic training data, thereby enabling the generation of a virtually unlimited amount of data. This helps improve the performance of the predictive model being trained, which in turn improves the accuracy of the model fitted using the predictive model. This improvement in accuracy helps facilitate improvements in downstream tasks that utilize features extracted from such models. 【0150】 In acquisition step 1002, multiple waveforms are acquired. These multiple waveforms represent the functional response of one or more prosthetic tissues during a single contraction-relaxation cycle. 【0151】 The multiple waveforms correspond to real-world or bootstrap data from which the synthetic training dataset will be generated. Each waveform corresponds to a time-series value representing the functional response (e.g., contractile force, calcium transient, etc.) of the artificial or manipulated tissue over a single contraction-relaxation cycle. Thus, each waveform includes a contraction period and a relaxation period and can be parameterized according to the contraction-relaxation model described above in relation to Figure 3. 【0152】 To help ensure that the synthetic training data represents the broadest possible population of contraction-relaxation cycles, multiple waveforms are preferably obtained from various artificial tissues across a range of different conditions. The artificial tissues include one or more manipulated muscle tissues, such as manipulated cardiac tissue and / or manipulated skeletal muscle tissue. Diversity is achieved by obtaining waveforms from artificial tissues across different cell lines, disease states, and treatments. Alternatively, the range of conditions within the multiple waveforms can be limited, thereby allowing the synthetic data and subsequent parameter estimation models to be fine-tuned for specific applications. For example, the multiple waveforms may be restricted to a vehicle-processed waveform to generate a control parameter estimation model, or to a specific tissue type (e.g., manipulated cardiac tissue) to generate a tissue-specific parameter estimation model. Beneficially, this helps improve the performance of the parameter estimation model if it is known which type of waveform the parameter estimation model will be used for. 【0153】 In step 1004, the extraction process, multiple parameter sets are extracted from multiple waveforms. One of these parameter sets characterizes the corresponding waveforms within the multiple waveforms. 【0154】 The parameter set includes parameters for the contraction-relaxation cycle model (as described above, for example, in relation to Figure 3). In one embodiment, the parameter set associated with the waveform includes a maximum parameter value (A), a shift parameter value (B), a contraction midpoint parameter value (t0), and a contraction growth rate parameter value (k c ), relaxation midpoint parameter value (t d ), and the slack growth rate parameter value (k r ) includes. 【0155】 Multiple parameter sets are extracted using either supervised, unsupervised, or semi-supervised methods. According to supervised methods, the parameter sets are manually fitted to each waveform. For example, a first waveform is presented to the user, and the parameter values ​​in the parameter set are adjusted by the user until a second waveform, produced by contraction-relaxation cycle model fitting according to the parameter set, closely matches the waveform. The final parameter set resulting in the closely matching second waveform is then used as one of the parameter sets within the multiple parameter sets associated with the first waveform. According to unsupervised methods, the parameter sets are automatically fitted to the waveform (e.g., using the methods described in relation to Figures 8 and 9 above). In one embodiment, the unsupervised method utilizes a trained machine learning model to predict the parameter set values ​​of the waveform. According to semi-supervised methods, the automatically determined parameter sets obtained according to the unsupervised method are manually reviewed and refined by one or more users. 【0156】 In the determination step 1006, the parameter set distribution is determined from multiple parameter sets. 【0157】 In step 1004 of method 1000, the extracted set of parameters includes multiple values ​​for each parameter of the contraction-relaxation cycle model. For example, if a set of 100 parameters is extracted, 100 parameter values ​​are extracted for each parameter of the contraction-relaxation cycle model. In step 1006, the distribution or distribution of values ​​is determined for each parameter of the model. In one embodiment, the distribution is determined independently for each parameter. Alternatively, a multivariate distribution is determined for the multiple parameters that make up the parameter set. 【0158】 Any suitable method can be used to determine the parameter set distribution, such as histogram-based methods, density estimation methods, and clustering methods. In one embodiment, the parameter set distribution is determined using a kernel density estimation (KDE) method that estimates the parameter set distribution using a kernel and bandwidth parameters. In one embodiment, a standard (Gaussian) kernel is used with a bandwidth selected using either cross-validation or a bandwidth selection method such as Scott's rule or Silverman's rule. 【0159】 Once the parameter set distribution is determined, the parameter set can be obtained by sampling from this distribution (i.e., by sampling from each individual distribution or from the combined distribution). 【0160】 In step 1008, a synthetic training dataset is generated. Each element of the synthetic training dataset contains a synthetic waveform and a corresponding parameter set used to generate the synthetic waveform. The corresponding parameter set is obtained from a parameter set distribution. 【0161】 Synthetic training datasets are generated by repeatedly sampling the parameter sets of the contraction-relaxation cycle model from the parameter set distribution (as described above) and generating waveforms corresponding to each sampled parameter set. In this way, training datasets of any size (e.g., 1000, 10000, 100000 training data elements) can be efficiently generated. Since the parameter set distribution is modeled based on real-world data, the synthetic data will closely approximate actual waveform data. 【0162】 Optionally, noise components are added to each waveform in the synthetic training dataset. The noise components are determined through a uniform distribution determined from multiple waveforms. 【0163】 In training step 1010, the predictive model is trained using a synthetic training dataset. The predictive model is trained to estimate an output parameter set from an input waveform. In one embodiment, the predictive model corresponds to the parameter estimation model 600 described in relation to Figure 6 above. As described above, in one embodiment, training the parameter estimation model 600 involves using a minibatch gradient descent method with a batch size of 128 and an ADAM solver. The ADAM solver has an initial learning rate of 1e-3 and terminates early based on the validation loss. Further details regarding the training of the parameter estimation model 600 performed in training step 1010 are shown above in relation to the description of Figure 6. 【0164】 Figure 11 shows a method 1100 for predicting a set of parameter values ​​for a contraction-relaxation cycle model using a synthetically trained predictive model, according to an embodiment of the present disclosure. 【0165】 Method 1100 includes the steps of acquiring a first waveform 1102 and predicting a first set of parameter values ​​1104. Method 1100 also includes an optional step 1106 of outputting a first set of parameter values. In one embodiment, Method 1100 is carried out by a control unit 104 shown in Figure 1, or by a subunit such as a model fitting unit 104-1. 【0166】 In acquisition step 1102, a first waveform is acquired. The first waveform includes the functional response of the first artificial tissue. 【0167】 In prediction step 1104, a first set of parameter values ​​is predicted from the first waveform using a prediction model trained on a synthetically generated training dataset (as described above in relation to Figure 10). 【0168】 In the optional output step 1106, a first set of parameter values ​​is output. In one embodiment, outputting the first set of parameter values ​​includes storing or saving the first set of parameter values ​​in persistent storage such as non-volatile memory, non-temporary media, or the like. Additionally or alternatively, outputting the first set of parameter values ​​includes transmitting the first set of parameter values ​​over a network (e.g., a local area network, a wide area network, and the like) or displaying the first set of parameter values ​​for user review. 【0169】 The method described in relation to Figures 8-10 above fits a contraction-relaxation cycle model to a waveform containing a single functional response (i.e., a single contraction-relaxation cycle) of the artificial tissue. In practice, the waveform typically contains multiple contraction-relaxation cycles, which may require multiple model fittings. To analyze and process such waveforms, each individual contraction-relaxation cycle is extracted from the waveform before model fitting. 【0170】 Figure 12 shows a method 1200 for extracting a contraction-relaxation cycle waveform according to an embodiment of the present disclosure. 【0171】 Method 1200 includes the steps of acquiring a first waveform 1202, convolving the first waveform with a pulse train 1204, identifying a first position 1206, and extracting a second waveform from the first waveform at the first position 1208. Method 1200 also includes an optional step 1210 of outputting a second waveform. In one embodiment, Method 1200 is carried out by a signal processing unit 104-2 of a control unit 104 shown in Figure 1. 【0172】 Generally, Method 1200 extracts a single contraction-relaxation cycle waveform (i.e., a single cycle, a single contraction cycle, a peak, or a functional response) from a larger waveform containing multiple contraction-relaxation cycles. The larger waveform may be obtained from a hardware device such as a bioreactor (i.e., bioreactor 102 shown in Figure 1) and typically corresponds to a functional response of an artificial tissue under specific conditions. For example, the larger waveform may contain the contractile response of manipulated cardiac tissue stimulated at 1 Hz over a period of 30 seconds. In this example, the larger waveform would contain approximately 30 peaks or 30 contraction-relaxation cycles corresponding to the contractions of the manipulated cardiac tissue in response to the electrical stimulation. Method 1200 provides an efficient and precise mechanism for extracting each contraction-relaxation cycle from the larger waveform, thereby allowing these sub-waveforms to be used for further processing and analysis (e.g., fitting a contraction-relaxation cycle model to these waveforms and extracting relevant features as described in relation to Figure 8 above). 【0173】 In acquisition step 1202, a first waveform is acquired. The first waveform includes multiple functional responses of the artificial tissue stimulated at a first frequency. 【0174】 In one embodiment, the first waveform is obtained from a bioreactor (e.g., bioreactor 102 shown in Figure 1) on which the artificial or manipulated tissue is grown / maintained. As previously mentioned, the artificial tissue includes manipulated muscle tissue such as manipulated cardiac tissue or manipulated skeletal muscle tissue. 【0175】 As described above in relation to Figure 1, electrical stimulation is applied to cell cultures during maturation, and once mature, the electrical stimulation is applied to artificial tissue to simulate the physiological environment specific to the artificial tissue, thereby allowing the functional response of the artificial tissue to this stimulation to be measured. The first waveform acquired in step 1202 includes the functional response of the artificial tissue to stimulation at a first frequency. In one embodiment, the first frequency to which the artificial tissue is stimulated, i.e., the pacing frequency, is 0.1 Hz to 20 Hz. In a further embodiment, the first frequency is 1 Hz to 6 Hz. 【0176】 Optionally, method 1200 includes a step (not shown) of stimulating artificial tissue at a first frequency prior to step 1202 of acquiring a first waveform. For example, an instruction (e.g., instruction 126 in Figure 1) or command is sent to a bioreactor containing artificial tissue (e.g., bioreactor 102 in Figure 1) to cause the bioreactor to stimulate the artificial tissue at a first frequency. 【0177】 In the convolution step 1204, the first waveform is convolved with a pulse train to generate a convolution waveform. The pulse train is generated at a first frequency. 【0178】 A pulse train corresponds to an idealized representation of the functional response of an artificial tissue at a first frequency. As is known, a pulse train, or pulse wave, is a waveform containing non-sinusoidal (rectangular) pulses or waves of duration T with frequency 1 / T0, where T0 is the duration of the pulse train. Thus, the duty cycle of the pulse train is T / T0. Therefore, a pulse train into which a first waveform is convolved contains a sequence of rectangular pulses having duration 1 / f and duration T, where f is the first frequency. 【0179】 Optionally, method 1200 further includes the step (not shown) of generating a pulse train at a first frequency before performing the convolution step 1204. 【0180】 Given a first waveform X1 and a pulse train p, the convolution y=(X1*p) performed in convolution step 1204 is as follows: 【number】 【0181】 In the identification step 1206, a first position associated with the maximum value of the convolutional waveform is identified. The first position corresponds to the expected position of the first contraction-relaxation cycle. 【0182】 The maximum value of the convolutional waveform corresponds to the position where the first waveform and the pulse train are best aligned. Therefore, the position of the maximum value of the convolutional waveform is used to identify the most likely position of a single contraction-relaxation cycle within the first waveform. The position of the maximum value of the convolutional waveform also provides a fixed point from which other contraction-relaxation cycles can be extracted from the first waveform. 【0183】 In extraction step 1208, a second waveform is extracted from a first position of the first waveform. The second waveform includes a first contraction-relaxation cycle and has a first duration proportional to the first frequency. 【0184】 The first position identified in identification step 1206 corresponds to the most likely position of a single contraction-relaxation cycle within the first waveform. Therefore, the second waveform extracted from the first position contains this contraction-relaxation cycle. Since the first waveform corresponds to the functional response of the prosthetic tissue when stimulated at a predetermined, set frequency, the duration, or length, of the second waveform is proportional to this frequency. For example, if the prosthetic tissue is stimulated at 1 Hz, the duration of the first waveform is 1 second, and if the prosthetic tissue is stimulated at 2 Hz, the duration of the first waveform is 0.5 seconds, and so on. 【0185】 Therefore, the second waveform corresponds to a window or subframe within the first waveform having a length corresponding to the first duration. In one embodiment, the second waveform is centered at a first position such that the midpoint of the second waveform is aligned, or substantially aligned, with a first position in the first waveform. 【0186】 In an optional output step 1210, a second waveform is output. In one embodiment, outputting the second waveform includes storing or saving the second waveform in persistent storage such as non-volatile memory, non-temporary media, or the like. Additionally or alternatively, outputting the second waveform includes transmitting the second waveform over a network (e.g., a local area network, a wide area network, and the like) or displaying the waveform for user review. 【0187】 In one embodiment, outputting a second waveform involves causing another process or method of the present disclosure to output a second waveform. For example, the second waveform may be output to method 800 described above, such that acquiring step 802 includes acquiring the second waveform from method 1200. 【0188】 Figure 13 shows a method 1300 for extracting further contraction-relaxation cycles from a waveform according to an embodiment of the present disclosure. 【0189】 Method 1300 includes the steps of identifying a second location (1302) and extracting a third waveform from a first waveform at the second location (1304), and further includes an optional step (1306) of outputting the third waveform. Method 1300 is performed after Method 1200. In particular, Method 1300 may be performed after the step of identifying the first location (1206) and may be performed in parallel with the extraction step (1208) of the second waveform. In one embodiment, Method 1300 is performed by the signal processing unit 104-2 of the control unit 104 shown in Figure 1. 【0190】 Method 1300 is used to extract further contraction-relaxation cycle waveforms from the first waveform. Beneficially, the extraction performed in Method 1300 is efficient and highly parallelized because Method 1300 utilizes prior information about the expected location of contraction-relaxation cycles within the first waveform, thereby enabling the independent extraction of contraction-relaxation cycles. 【0191】 In the identification step 1302, the second position is identified based on the first position and the first frequency. The second position corresponds to the expected position of the second contraction-relaxation cycle. 【0192】 The first position (identified in step 1206 of Method 1200) corresponds to the best alignment between the first waveform and the pulse train. Therefore, the first position can be understood as the most likely position of a contraction-relaxation cycle within the first waveform. Since the first waveform contains the functional response of the prosthesis at a predetermined frequency (i.e., the first frequency), other contraction-relaxation cycles linked to the functional response of the prosthesis are very likely to be located at positions separated from the first position. Thus, the first position can function as a fixed point within the first waveform from which other contraction-relaxation cycle waveforms can be extracted. 【0193】 The second position corresponds to the expected position of the second contraction-relaxation cycle and is separated from the first position by a distance proportional to the first frequency. Specifically, given a first position t1 in the first waveform, the second position t2 is given by t2 = t1 + (a × T0), where a is the step coefficient and T0 = 1 / f is the period of the pulse train. Thus, the next contraction-relaxation cycle can be identified by setting the scaling coefficient a = 1, and the previous contraction-relaxation cycle can be identified by setting the scaling coefficient a = -1. 【0194】 In extraction step 1304, the third waveform is extracted from the second position of the first waveform. The third waveform includes a second contraction-relaxation cycle and has a second duration proportional to the first frequency. 【0195】 The third waveform corresponds to the functional response (i.e., contraction-relaxation cycle) of the artificial tissue when stimulated at a predetermined, set frequency. Therefore, the duration, or length, of the third waveform is proportional to this frequency. For example, if the artificial tissue is stimulated at 1 Hz, the second duration is 1 second, and if the artificial tissue is stimulated at 2 Hz, the second duration is 0.5 seconds, and so on. In one embodiment, the first duration and the second duration are the same. 【0196】 In one embodiment, the third waveform is centered at a second position such that the midpoint of the third waveform is aligned, or substantially aligned, with the second position of the first waveform. 【0197】 In an optional step 1306 of outputting, a third waveform is output. In one embodiment, outputting the third waveform includes storing or saving the third waveform in persistent storage such as non-volatile memory, non-temporary media, or the like. Additionally or alternatively, outputting the third waveform includes transmitting the third waveform over a network (e.g., a local area network, a wide area network, and the like) or displaying the waveform for user review. 【0198】 In one embodiment, outputting a third waveform includes causing another process or method of the present disclosure to output a third waveform. For example, the third waveform may be output to method 800 described above, such that acquiring step 802 includes acquiring the third waveform from method 1300. 【0199】 As described above, method 1300, which extracts further contraction-relaxation cycle waveforms from a first waveform, can be repeated for all contraction-relaxation cycles within the first waveform. Since the extraction performed by method 1300 depends only on the first position and first frequency, further signal processing or analysis is not required to identify the positions of further contraction-relaxation cycles. Thus, method 1300 provides a rapid and efficient method for extracting contraction-relaxation cycles from a waveform. These waveforms can then be further processed, for example, by fitting a model to the waveform to generate a noise-filtered representation of the waveform. 【0200】 Figure 14 shows a method 1400 for predicting therapeutic effects using a contraction-relaxation cycle model according to an aspect of the present disclosure. 【0201】 Method 1400 includes the steps of: acquiring multiple signals 1402; splitting the multiple signals into a first set of waveforms 1404; fitting a model to each of the first set of waveforms 1406; generating a second set of waveforms from the model 1408; extracting first feature values ​​1410; extracting second feature values ​​1412; and determining the effect 1414. Method 1400 also includes an optional step 1416 for outputting the effect. In one embodiment, Method 1400 is carried out by a control unit 104 or a subunit thereof, as shown in Figure 1. 【0202】 Generally, Method 1400 describes the application of a contraction-relaxation cycle model to downstream drug discovery / development tasks. Specifically, the contraction-relaxation cycle model is used to efficiently generate accurate feature values ​​from baseline and perturbation signals of manipulated tissue. Accurately extracting features from these signals allows for the efficient and accurate identification of effects associated with perturbations. 【0203】 In acquisition step 1402, multiple signals are acquired. These multiple signals include a baseline signal and a perturbation signal. The baseline signal includes a first set of functional responses of the manipulated tissue under reference conditions. The perturbation signal includes a second set of functional responses of the manipulated tissue under perturbation conditions with a first perturbation. 【0204】 Baseline and perturbation signals include multiple functional responses (i.e., multiple contraction-relaxation cycles, or peaks) of the manipulated tissue under reference and perturbation conditions. Generally, reference conditions refer to conditions that provide a baseline comparison to perturbation conditions. In embodiments, reference conditions are conditions associated with a control setting or environment. Reference conditions may correspond to manipulated or artificial tissue in a default, natural, or unaltered state (i.e., without drug or agent administration). Alternatively, reference conditions may correspond to vehicle-treated manipulated tissue. Perturbation conditions refer to conditions under which the manipulated tissue is perturbed in some way. Examples of perturbation conditions include administration of a drug or compound (i.e., a perturbant), a disease state, a different cell line, or a physical perturbation applied to the manipulated tissue. Therefore, perturbation conditions may alternatively be referred to as treatment conditions. In the case of perturbations involving drugs or compounds, the conditions are further associated with effects related to the drug or compound, such as mechanism of action or toxicity. Considering the range of different perturbation conditions, the manipulated tissue (e.g., manipulated tissue in a disease state treated with a specific compound) may be associated with more than one perturbation condition. 【0205】 In one embodiment, baseline and perturbation signals are acquired from a bioreactor (e.g., bioreactor 102 in Figure 1) on which the engineered or artificial tissue is grown / maintained. As previously stated, the artificial tissue includes engineered muscle tissue such as engineered cardiac tissue or engineered skeletal muscle tissue. The baseline and perturbation signals are acquired at two different time points. For example, the baseline signal is acquired at a first time point, when the engineered tissue is then perturbed according to a first perturbation (e.g., a first dose of the compound is applied to the engineered tissue), and the perturbation signal is acquired at a second time point after the first time point. The baseline and perturbation signals include the functional response of the engineered tissue when stimulated at a predetermined pacing frequency (e.g., 0.1 Hz, 0.5 Hz, 1 Hz, 2 Hz, etc.). Alternatively, the baseline and perturbation signals include the spontaneous functional response of the engineered tissue in the absence of external stimulation. 【0206】 In the splitting step 1404, the multiple signals are split into a first set of multiple waveforms. Each of the first set of multiple waveforms contains the contraction and relaxation responses of the manipulated tissue during a single contraction-relaxation cycle. 【0207】 Multiple signals are divided into a first set of waveforms using the method 1200 for extracting the contraction-relaxation cycle waveforms described above. Alternatively, multiple signals are divided by manually annotating or extracting regions within the first set of waveforms that correspond to individual contraction-relaxation cycles. 【0208】 The first set of waveforms generated in the splitting step 1404 includes a first subset of waveforms associated with the waveform extracted from the baseline signal and a second subset of waveforms associated with the waveform extracted from the perturbation signal. The subsequent model fitting (described later) is independent of the waveform source (i.e., independent of whether the waveform is the reference for the perturbation waveform), but the identification of whether the waveform corresponds to a reference condition or a perturbation condition is used in the subsequent feature extraction step. 【0209】 In fitting step 1406, the model is fitted to each of the first set of waveforms. The model is a contraction-relaxation cycle model that independently parameterizes the growth of the contraction and relaxation responses of the manipulated tissue during a single contraction-relaxation cycle for each waveform. 【0210】 Step 1406, which involves fitting the model to each waveform, corresponds to step 804 of fitting, which is described in more detail above in relation to Figure 8. Thus, in fitting step 1406, fitting step 804 is repeated for each of the first set of waveforms. The process of fitting the model to the waveforms is described in more detail above in relation to Figures 8 and 9. 【0211】 In step 1408 of the generation process, the second set of waveforms is generated from a model fitted to each of the first set of waveforms. The second set of waveforms includes a set of filtered baseline waveforms associated with the baseline signal and a set of filtered treatment waveforms associated with the treatment signal. 【0212】 The generation step 1408 corresponds to repeatedly applying the generation step 806 (described in more detail above in relation to Figure 8) to the model fitted to each waveform in the first set of waveforms. The process of generating the second waveform from the model fitted to the first waveform is described in more detail above in relation to Figures 8 and 9. 【0213】 In extraction step 1410, the first feature value of the first feature is extracted from multiple filtered baseline waveforms. 【0214】 Multiple filtered baseline waveforms contain noise-filtered or noise-suppressed representations of the contraction-relaxation cycles in the baseline signal. Because the signal is filtered to remove noise, features can be accurately extracted from these waveforms. 【0215】 Step 1410 for extracting a first feature value includes extracting multiple feature values ​​from multiple filtered baseline waveforms such that the first feature value includes multiple feature values ​​or representations of multiple feature values. Thus, the value of the first feature is extracted from each of the multiple filtered baseline waveforms in order to determine the first feature value. In one embodiment, the first feature value includes the mean (mean, median, etc.) of the first feature determined from the multiple filtered baseline waveforms. In a further embodiment, the first feature value includes the maximum, minimum, or distribution of values ​​determined from the multiple filtered baseline waveforms. 【0216】 The first characteristic is one of the following: single contraction, or peak, amplitude (i.e., peak amplitude shown in Figure 2, 206), contraction time (i.e., time to peak amplitude shown in Figure 2, 208), maximum contraction slope (i.e., maximum development rate shown in Figure 2, 214), relaxation time (i.e., time to peak decline shown in Figure 2, 210), maximum relaxation slope (i.e., maximum decline rate shown in Figure 2, 216), or single contraction duration (i.e., duration shown in Figure 2, 212). 【0217】 In extraction step 1412, the second feature value of the first feature is extracted from multiple filtered perturbation waveforms. 【0218】 Step 1412 of extracting the second eigenvalue includes extracting a plurality of eigenvalues from a plurality of filtered perturbation waveforms such that the second eigenvalue includes a plurality of eigenvalues or representations of a plurality of eigenvalues. Therefore, the value of the first feature is extracted from each of the plurality of filtered perturbation waveforms to determine the second eigenvalue. In one embodiment, the second eigenvalue includes an average value (such as an average value, a median value, etc.) of the first feature determined from a plurality of filtered perturbation waveforms. In a further embodiment, the second eigenvalue includes a maximum value, a minimum value, or a distribution of values determined from a plurality of filtered perturbation waveforms. 【0219】 In step 1414 of determination, the effect associated with the first perturbation is determined based on a comparison between the first eigenvalue and the second eigenvalue. 【0220】 The first eigenvalue is a quantitative descriptor of the functional response of the tissue operated under reference conditions. The second eigenvalue is a quantitative descriptor of the functional response of the tissue operated under perturbation conditions with the first perturbation. Therefore, by comparing the first eigenvalue and the second eigenvalue, any change that occurs as a result of the first perturbation in the functional response of the operated tissue, that is, the effect, becomes clear. 【0221】 For example, the first perturbation may correspond to the application of a compound having an unknown physiological effect. By comparing the first eigenvalue corresponding to the peak amplitude of the contraction force waveform of the tissue operated under reference conditions with the second eigenvalue corresponding to the peak amplitude of the contraction force waveform of the tissue operated under perturbation conditions with the application of the compound, an increase in the average peak amplitude becomes clear. Therefore, it can be inferred that the compound has an effect associated with increasing the contraction force of the operated tissue during the contraction-relaxation cycle. Advantageously, since the eigenvalue is determined from a noise-filtered waveform, the effect caused by the difference between the eigenvalues can be more accurately identified, leading to improved treatment and potentially improved patient outcomes. 【0222】 Systems and methods for detecting spontaneous tissue contraction. The functional response waveform includes a time series of values corresponding to measurements of the functional response of tissue over a period of time. Thus, the functional response waveform encodes the changes (e.g., contractility, displacement, etc.) in the functional response of the tissue over a set period of time. In some situations, this change is induced as a result of an external stimulus applied to the tissue. For example, an electrical stimulus can be applied periodically to tissue such as manipulated muscle tissue at a predetermined pacing frequency (e.g., 0.5 Hz, 1 Hz, 2 Hz, etc.). When such an external stimulus is applied, the functional response waveform will typically exhibit periodic behavior that can be utilized when identifying and extracting a single response for feature extraction and downstream analysis tasks. In situations without such a stimulus, or when the tissue exhibits abnormal contraction behavior, such periodicity may not be present within the waveform. Thus, it is more difficult to identify and extract a single response from such “spontaneous” contraction response waveforms. The present disclosure describes systems and methods for efficiently and accurately identifying spontaneous contractions within functional response waveforms that may not exhibit periodic behavior. This enables the processing and analysis of waveforms encoding spontaneous tissue behavior, thereby opening the possibility that such waveforms can be used in the context of downstream tasks such as drug discovery and drug development. 【0223】 As described above, in many situations, the contraction-relaxation cycles within the waveform are caused by external stimuli such as electrical stimuli applied to the tissue at a predetermined pacing frequency (e.g., 1 Hz, 2 Hz, etc.). The regularity of the functional response enables the identification and processing of individual contraction-relaxation cycles either through the identification of repeating patterns or through the incorporation of prior knowledge of the pacing frequency. This is illustrated in FIG. 15A, as described below. 【0224】 FIG. 15A shows the functional response of tissue exhibiting periodic contraction behavior. 【0225】 Figure 15A shows a functional response waveform 1500, and a timeline 1502 of peak locations including the first point 1504, the second point 1506, and the third point 1508. The functional response waveform 1500 contains the functional response (e.g., contractile force) of the tissue over a period of time. A single contractile response (a single peak or contraction) is indicated within the functional response waveform 1500 by a peak within the functional response waveform 1500, i.e., a local maximum value within the functional response waveform 1500. The timeline 1502 of peak locations illustrates the periodicity or regularity of these contractile responses, because the points in the timeline 1502 corresponding to the locations of peaks or contractions in the functional response waveform 1500 are spaced at approximately regular intervals along the timeline 1502. For example, the interval between the contraction associated with the first point 1504 and the contraction associated with the second point 1506 is substantially the same as the interval between the contraction associated with the second point 1506 and the contraction associated with the third point 1508. Therefore, the periodicity of contractions within the functional response waveform 1500 indicates that the tissue exhibits periodic or regular contraction behavior. 【0226】 As previously mentioned, for waveforms exhibiting periodic behavior, periodicity can be leveraged to assist in downstream tasks such as contraction response extraction (i.e., extracting a single contraction-relaxation cycle from the waveform). For example, once known or learned, periodicity can be used to determine the expected interval between single contractions, thereby providing prior information for identifying the relative positions of contraction responses. However, some tissues may exhibit spontaneous behavior in contrast to periodic contraction behavior, as illustrated in Figure 15B. 【0227】 Figure 15B shows the functional response of tissue exhibiting spontaneous contraction behavior. 【0228】 Figure 15B shows a functional response waveform 1510 and a timeline 1512 of peak positions including a first point 1514, a second point 1516, and a third point 1518. The functional response waveform 1510 contains the functional response (e.g., contractile force) of the tissue over a period of time. As shown in Figure 15A, a single contractile response (a single peak or contraction) is indicated within the functional response waveform 1510 by its peak. The timeline 1512 of peak positions illustrates the spontaneity (i.e., lack of periodicity or regularity) of these contractile responses because the points in the timeline 1512 that correspond to the positions of peaks or contractions are irregularly spaced along the timeline 1512. For example, the interval between the contraction associated with the first point 1514 and the contraction associated with the second point 1516 is substantially different from the interval between the contraction associated with the second point 1516 and the contraction associated with the third point 1518. Therefore, the lack of periodicity in contractions within the functional response waveform 1510 indicates that the tissue exhibits spontaneous contraction behavior. Alternatively, spontaneous contraction behavior exhibited by tissue can be periodic but is considered spontaneous because the tissue's contraction response (e.g., by the electronic stimulation described above in relation to Figure 1) does not result from the stimulation of the tissue. In either case, there is no prior information that can be used to identify and extract individual contractions from the functional response waveform. 【0229】 The irregularity in the tissue contraction behavior observed in Figure 15B may be due to several factors. For example, in manipulated tissue, the irregularity may be due to external stimuli no longer applied to the manipulated tissue. Alternatively, the irregularity or spontaneousness may be due to one or more conditions of the tissue, such as conditions induced by a disease state, compounds or drugs applied to the tissue, or similar. 【0230】 In the absence of a stimulus applied to the tissue to induce contraction, prior knowledge of periodicity, i.e., the fundamental characteristics of regular behavior, cannot be used to identify individual contraction responses (i.e., the frequency at which the tissue is stimulated cannot be used to identify individual contraction-relaxation cycles or individual "peaks" within a waveform). According to aspects of this disclosure, different methods may be used to determine the spontaneous behavior of tissue (such as manipulated or artificial tissue), as shown in Figure 16. 【0231】 Figure 16 shows a system 1600 for determining the spontaneous behavior of an manipulated organization, according to an aspect of this disclosure. 【0232】 System 1600 comprises a periodicity classifier 1602, a decision unit 1604, a spontaneous contraction classifier 1606, and a profile generator 1608. A waveform 1610 containing the functional response of the manipulated tissue over a period of time is provided as input to the periodicity classifier 1602. The periodicity classifier 1602 generates a classification score 1612 based on the waveform 1610. The decision unit 1604 determines whether the classification score 1612 indicates any periodic contractions present in the waveform 1610. If the classification score 1612 indicates that no periodic contractions are present in the waveform 1610, the waveform 1610 is provided as input to the spontaneous contraction classifier 1606. The spontaneous contraction classifier 1606 generates a classification score 1614 based on the waveform 1610. The profile generator 1608 generates a behavioral profile 1616 of the manipulated tissue based on the classification score 1614. In one embodiment, the periodic classifier 1602 includes a spectral transformation process 1618 and a predictive model 1620. In one embodiment, the spontaneous contraction classifier 1606 includes a transformation process 1622 and a thresholding operation 1624. 【0233】 System 1600 corresponds to a hierarchical method for determining the spontaneous behavior of the manipulated tissue. By determining the spontaneous behavior of the manipulated tissue as encoded within the functional response waveform, the behavior and contraction location data can be used to perform many downstream analytical tasks (e.g., feature extraction, assay development, etc.). The hierarchy of classifiers (i.e., periodic classifier 1602 and spontaneous contraction classifier 1606) is used to identify the overall contraction behavior of the manipulated tissue, and then the local contraction behavior of the manipulated tissue. By using the hierarchy of classifiers, a more efficient use of computing resources is achieved because the overall classifier (i.e., periodic classifier 1602) first filters out waveforms known to exhibit periodic or regular behavior. Thus, only waveforms that are not predicted to have periodic or regular behavior are fed to the local classifier (i.e., spontaneous classifier 1606). Furthermore, this means that the local classifier (i.e., spontaneous classifier 1606) can be tailored to waveforms that may exhibit spontaneous behavior, thereby providing an improved identification of spontaneous behavior (and the location of spontaneous contraction). 【0234】 Waveform 1610 includes the functional response of the manipulated or artificial tissue over a period of time (e.g., over 10 seconds, 20 seconds, 30 seconds, etc.). As described above with respect to Figure 1, the functional response may include changes in the contractile force of the manipulated tissue over the period, or displacement or contractile displacement of the manipulated tissue over the period. Although not shown in Figure 16, waveform 1610 is obtained either directly or indirectly from a bioreactor, such as the bioreactor 102 shown in Figure 1. In one embodiment, the manipulated tissue includes an artificial heart or artificial muscle tissue such as skeletal muscle tissue. 【0235】 A periodic classifier 1602 (alternatively referred to as the first classifier, frequency-based global classifier, or global classifier) ​​is used to predict the overall characteristics of waveform 1610. In particular, the periodic classifier 1602 is configured to determine whether waveform 1610 contains any periodic contractions, such as those described in relation to Figure 15A above. The periodic classifier 1602 includes any suitable predictive model that can predict from the time series input whether the time series input contains regular or periodic responses (peaks or contractions). In one embodiment, the periodic classifier 1602 includes a spectral transform process 1618 and a predictive model 1620. The spectral transform process 1618 applies a spectral transform, such as a Fourier transform or similar, to waveform 1610 to generate a spectral waveform. The spectral waveform contains a frequency-based representation of waveform 1610. The spectral waveform is then used by the predictive model 1620 to determine a classification score 1612 (i.e., the presence or absence of any regular periodic contractions within the waveform 1610). Beneficially, the spectral waveform encodes key features of the tissue's contraction behavior, which helps improve the classification performance of the predictive model 1620. This is illustrated by the exemplary spectral responses shown in Figures 17A and 17B. 【0236】 Figures 17A and 17B show spectral reactions of different shrinkage reactions according to embodiments of the present disclosure. 【0237】 Spectral responses are obtained from functional response waveforms using a spectral transform process involving a Fourier transform. Each waveform is associated with an engineered or artificial tissue exhibiting a different contractile response. 【0238】 The first functional response waveform 1702 is obtained from manipulated tissue exhibiting normal contractile behavior. That is, the contractile response encoded in the first spectral response 1702 follows a periodic (regular) pattern of contraction-relaxation cycles. As can be seen, the power is strongest at 1 Hz (i.e., the pacing frequency) and decreases as the harmonics increase. The second functional response waveform 1704 is obtained from manipulated tissue exhibiting an abnormal contractile response. That is, several double contractions (or double pulsations) are present. This is seen in the change in harmonic power in the second spectral response 1714. The third functional response waveform 1706 is obtained from manipulated tissue exhibiting a contractile response with increased force. That is, the power in the third spectral response 1716 shifts from the harmonics to the primary frequency (1 Hz). The fourth functional response waveform 1708 is obtained from tissue exhibiting a spectral response with reduced contractile force. That is, the power of the fourth spectral response 1708 shifts away from the primary frequency (1 Hz) to the harmonics (1718). The fifth functional response waveform 1710 is obtained from tissue that does not exhibit a periodic or regular contraction response. That is, the power of the fifth spectral response 1710 is strongest at 0 Hz (1720). 【0239】 Therefore, the use of spectral waveforms provides a compact and descriptive representation of tissue contraction behavior, which can help improve the discriminative performance of classifiers or predictive models tasked with identifying different contraction behaviors. 【0240】 Referring back to FIG. 16, the prediction model 1620 determines the classification score 1612 using the spectral waveform (i.e., the frequency-based representation of waveform 1610). The classification score 1612 represents the contraction behavior of the tissue associated with waveform 1610 and indicates whether waveform 1610 includes regular or periodic contractions. Thus, the classification score 1612 is a binary classification, which takes one score or value (e.g., "+1") indicating that waveform 1610 includes periodic contractions and another score or value (e.g., "0" or "-1") indicating that waveform 1610 does not include periodic contractions. Alternatively, the prediction model 1620 is a multi-class classifier that predicts the contraction behavior type from the spectral waveform. For example, the prediction model 1620 can be trained to assign the spectral waveform to one of the contraction response types illustrated in FIGS. 17A and 17B (e.g., a classification score of "0" indicating no periodic contractions, a classification score of "1" indicating normal periodic contractions, a classification score of "2" indicating abnormal contraction behavior, etc.). 【0241】 The prediction model 1620 is a trained machine learning model such as a trained neural network, support vector machine, random forest, or the like. In one embodiment, the prediction model 1620 is a convolutional neural network such as that shown in FIG. 18. 【0242】 FIG. 18 shows a convolutional neural network 1800 for predicting contraction behavior according to an embodiment of the present disclosure. 【0243】 The neural network 1800 comprises a convolutional network 1802, a long short-term memory (LSTM) network 1804, and a fully connected network 1806. An input vector 1808 is received by the convolutional network 1802, and an output vector 1810 is produced by the fully connected network 1806. The convolutional network 1802 includes a first block 1812, a second block 1814, a third block 1816, a fourth block 1818, and a fifth block 1820. Each block includes a two-dimensional (2D) convolutional layer, a batch normalization layer, and optionally a rectified linear unit layer. This is illustrated in Figure 18 by the 2D convolutional layer 1822, batch normalization layer 1824, and rectified linear unit layer 1826 of the first block 1812. The LSTM network 1804 includes a smoothing layer 1828, a bidirectional LSTM layer 1830, and a dropout layer 1832. The fully connected network 1806 includes a fully connected layer 1834, a dropout layer 1836, and a softmax layer 1838. 【0244】 The input vector 1808 contains a sequence of values ​​corresponding to spectral waveforms (such as the spectral waveforms shown in Figures 17A and 17B). 【0245】 The convolutional network 1802 includes a sequence of blocks having a similar architecture. The first block 1812 includes a 2D convolutional layer 1822 with four filters of size 3, a batch normalization layer 1824 (normalizing the input via recentering and rescaling), and a rectified linear unit layer 1826. The second block 1814 and the third block 1816 include the same architecture, namely a 2D convolutional layer with seven filters of size 3, a batch normalization layer, and a rectified linear unit layer. The fourth block 1818 includes a 2D convolutional layer with sixteen filters of size 3, a batch normalization layer, and a rectified linear unit layer. The fifth block 1820 includes a 2D convolutional layer with thirty-two filters of size 3, and a batch normalization layer (i.e., the fifth block 1820 does not include an optional rectified linear unit layer). 【0246】 The LSTM network 1830 comprises a flattening layer 1828 that flattens the spatial dimension of the output of the fifth block 1820 to a single dimension, a bidirectional LSTM layer 1830 with eight hidden units, and a dropout layer 1832 that helps reduce overfitting by randomly setting the input to zero with a probability of 0.3. 【0247】 The fully connected network 1806 includes a fully connected layer 1834 with an output size set to the number of contraction behaviors to predict (i.e., two layers are used when predicting periodic and aperiodic contractions), a dropout layer 1836 that randomly sets the input to zero with a probability of 0.05, and a softmax layer 1838 that applies a softmax function to the output of the dropout layer 1836. 【0248】 The output vector 1810 corresponds to a probability vector with a size equal to the number of predicted bloat behaviors. Therefore, the output vector 1810 is used to determine the classification score (i.e., the classification score 412 described in relation to Figure 16 above). For example, it can be used to identify the maximum value via a threshold. 【0249】 In one embodiment, the neural network 1800 is trained using a synthetic dataset containing 40,000 training samples and 10,000 validation samples. Each element of the synthetic training dataset is created by generating a synthetic waveform that includes either periodic or aperiodic contractions, and then applying a spectral transform (Fourier transform) to generate a spectral waveform. The synthetic dataset comprises 25,000 periodic waveforms and 25,000 aperiodic waveforms. Each waveform is associated with a corresponding label indicating whether the waveform is periodic or aperiodic. The neural network 1800 is trained using mini-batch gradient descent with a batch size of 128 and an ADAM solver. The ADAM solver has an initial learning rate of 1e-3 and terminates early based on the validation loss. 【0250】 Referring again to Figure 16, the decision unit 1604 uses the classification score 1612 obtained from the periodicity classifier 1602 to determine whether the waveform 1610 contains periodic contractions and therefore whether a second classifier (i.e., the spontaneous contraction classifier 1606) should be called. The decision unit 1604 uses a decision rule to determine whether to pass the waveform 1610 to the spontaneous contraction classifier 1606 (as indicated by the black circles in Figure 16) or to take no further action. The decision rule determines whether the classification score 1612 indicates that there are no periodic contractions in the waveform 1610 (for example, if the classification score 1612 takes a value such as "0", this indicates that the periodic classifier 1602 classified the waveform 1610 as containing no periodic contractions). If there are no periodic contractions, the waveform 1610 is input to the spontaneous contraction classifier 1606. 【0251】 The spontaneous contraction classifier 1606 (alternatively referred to as the local classifier, second classifier, or peak detector) is used to generate a classification score 1614 indicating whether the waveform 1610 contains any spontaneous contractions of the manipulated tissue over a period of time. It is possible that no tissue contraction response is observed within the waveform 1610 (i.e., the tissue did not perform induced or spontaneous contractions). Therefore, both classification scores 1612 and 1614 would indicate that there are no induced or spontaneous contractions within the waveform 1610. Alternatively, the waveform 1610 may contain one or more spontaneous contractions (as illustrated in Figure 15B) rather than induced contractions. The spontaneous contraction classifier 1606 uses the waveform 1610 to determine which of these two behaviors is exhibited by the tissue within the waveform 1610. Therefore, the spontaneous contraction classifier 1606 is any suitable predictive model or trained machine learning model, such as a trained neural network, support vector machine, random forest, or similar. 【0252】 In one embodiment, the spontaneous contraction classifier 1606 includes a conversion process 1622 and a thresholding operation 1624. Generally, the conversion process 1622 generates a converted waveform from the waveform 1610, and then the thresholding operation 1624 is applied to the converted waveform. The conversion process 1622 reduces noise while simultaneously enhancing peaks in the waveform 1610. This helps improve the performance of the thresholding operation 1624, which then identifies any peaks that exceed one or more thresholds. If there are any peaks in the converted waveform that exceed one or more thresholds (thresholds), the waveform 1610 includes spontaneous contractions of the manipulated tissue; otherwise, if there are no thresholds, the waveform 1610 does not include any spontaneous contractions of the manipulated tissue. 【0253】 The conversion process 1622 includes any preferred signal processing operation that can enhance peaks in the waveform 1610, such as peak sharpening or peak filtering. In one embodiment, the conversion process 1622 includes a Pan-Tompkins algorithm. In a further embodiment, the conversion process 1622 includes a modified Pan-Tompkins algorithm. The Pan-Tompkins algorithm was developed to detect the QRS complex wave (i.e., Q wave, R wave, and S wave) of an electrocardiogram (ECG) signal. The Pan-Tompkins algorithm includes a sequence of filters applied to the waveform to enhance the frequency components (i.e., peaks) of the waveform while removing noise. Generally, the Pan-Tompkins algorithm includes a noise reduction process and a subsequent enhancement process. The noise reduction process applies a bandpass filter (i.e., a low-pass filter followed by a high-pass filter) to the input waveform. The enhancement process includes a differential, squared, and integral filter. The output of the noise reduction process is provided to a derivative operation that provides slope information. The squaring operation enhances the peak of the derivative operation's output, and the integral filter applies a moving average to the output of the squaring operation. As described in more detail below, the modified Pan-Tompkins algorithm of this disclosure incorporates a steady-state waveform transformation into the noise reduction process and replaces the squaring filter with a rectification operation. Thus, the modified Pan-Tompkins algorithm produces a transformed waveform with better noise reduction characteristics and improved peak enhancement. This helps to improve the identification of spontaneous collapse, which in turn helps to improve the performance of downstream tasks that utilize such information for tasks such as drug discovery and development. 【0254】 Figure 19 illustrates the sequential results of implementing the modified Pan-Tompkins algorithm according to an embodiment of the present disclosure. 【0255】 Figure 19 shows the results of applying the steps of the modified Pan-Tompkins algorithm of this disclosure to waveform 1902. Figure 19 shows the low-pass filtered waveform 1904, the high-pass filtered waveform 1906, the wavelet-transformed waveform 1908, the differentiated waveform 1910, the rectified waveform 1912, and the integrated waveform 1914. The modified Pan-Tompkins algorithm is described as performing each of the steps described below, but those skilled in the art will understand that in some embodiments the steps can be combined and / or omitted. For example, the algorithm may include performing a steady wavelet transform and rectification operation, or a steady wavelet transform, differentiation operation, and rectification operation. In some embodiments, a normalization operation is applied to the output of the modified Pan-Tompkins algorithm to scale the waveform to a consistent range of values ​​along the y-axis. 【0256】 Waveform 1902 corresponds to waveform 1610 shown in Figure 16 and includes the functional response (e.g., force) of an manipulated tissue over a period of time. The results of applying the noise reduction process of the modified Pan-Tompkins algorithm are illustrated in Figure 19 by the low-pass filtered waveform 1904, the high-pass filtered waveform 1906, and the wavelet-transformed waveform 1908. Low-pass filtered waveform 1904 corresponds to the result of applying a low-pass filter to waveform 1902. The low-pass filter transmits portions of waveform 1902 with frequencies below a predetermined cutoff frequency and attenuates portions with frequencies above the predetermined cutoff frequency. Thus, the low-pass filter helps remove major wiring mismatches from waveform 1902. In one embodiment, the low-pass filter comprises a two-dimensional Gaussian kernel with a standard deviation of 2. High-pass filtered waveform 1906 corresponds to the result of applying a high-pass filter to low-pass filtered waveform 1904. Therefore, the high-pass filtered waveform 1906 corresponds to the band-pass filtered waveform, as it is the result of both low-pass and high-pass filtering of waveform 1902. The high-pass filter transmits portions of the low-pass filtered waveform 1904 that have frequencies above a predetermined cutoff frequency and attenuates portions that have frequencies below the predetermined cutoff frequency. Thus, the high-pass filter helps to perform baseline alignment while eliminating drift. In one embodiment, the high-pass filter comprises an elliptic filter having an order of 8, a ripple of 0.5 dB, an attenuation of 40 dB, and edge frequencies of 1 and 20. 【0257】 Wavelet-transformed waveform 1908 is the result of applying a stationary wavelet transform to high-pass filtered waveform 1906. As described above, the modified Pan-Tompkins algorithm of this disclosure performs a stationary wavelet transform as an additional step in the noise reduction process of the Pan-Tompkins algorithm. Beneficially, the stationary wavelet transform is shift-invariant and helps reduce noise in the waveform while preserving important transition features (changes) in the waveform that may be required for downstream tasks such as feature extraction. The stationary wavelet transform is an extension of the wavelet transform in which the wavelet coefficients are not subtracted at all stages. In one embodiment, the wavelet transform includes a discrete stationary wavelet transform (1D) using a five-stage decomposition and a Daubecchies 4 (db4) wavelet. 【0258】 The results of the augmentation process using the modified Pan-Tompkins algorithm are shown in Figure 19, with the differentiated waveform 1910, the rectified waveform 1912, and the integrated waveform 1914. Differentiated waveform 1910 is the result of performing a differential operation on the wavelet-transformed waveform 1908. Differential operations are used to highlight rapid changes (i.e., contractions). Numerical gradients are calculated with uniform spacing between points in all directions. In one embodiment, the differential operation is performed on the transfer function H(z)=(1 / 8T)(-z 2 -2z -1 +2z 1 +z 2This corresponds to a five-point derivative having ), where T is the sampling period. Instead of squaring, the modified Pan-Tompkins algorithm of this disclosure applies a rectification operation to the differentiated waveform 1910 to produce a rectified waveform 1912. The rectification operation clips the differentiated waveform 1910 so that any negative parts of the differentiated waveform 1910 are removed. Thus, the rectification operation enhances the dominant peaks in the waveform. The integrated waveform 1914 corresponds to the result of applying a moving window integration operation to the rectified waveform 1912. The moving window integration operation applies a moving average filter (i.e., a sliding window filtering operation) to the rectified waveform 1912. Thus, the moving window integration operation removes short-duration artifacts from the rectified waveform 1912. In one embodiment, the moving average filter is 【number】 It features a moving average filter calculated over a sliding window, where n is the length of the waveform 1610. 【0259】 As can be seen in the example in Figure 19, the output of the modified Pan-Tompkins algorithm (integrated waveform 1914) includes a denoised version of the input waveform (waveform 1902) with an enhanced representation of the contraction-relaxation cycles. By reducing noise in the waveform and simultaneously enhancing the peaks (i.e., contraction-relaxation cycles), the identification of spontaneous contractions in the transformed waveform is improved. This, in turn, helps improve the performance of downstream tasks that utilize such information for tasks such as drug discovery and development. 【0260】 Referring again to Figure 16, the thresholding operation 1624 identifies any peaks in the transformed waveform produced by the transformation process 1622, i.e., threshold overruns. The presence of peaks in the transformed waveform indicates that waveform 1610 contains spontaneous contractions. That is, the classification score 1614 indicates that waveform 1610 contains one or more spontaneous contractions if any portion of the transformed waveform meets the thresholding criteria defined by the thresholding operation 1624. Similarly, the absence of threshold overruns in the transformed waveform indicates that waveform 1610 does not contain any spontaneous contractions (or contractions induced by the hierarchical classification system used by system 1600). That is, if no portion of the transformed waveform meets the thresholding criteria defined by the thresholding operation 1624, the classification score 1614 indicates that waveform 1610 does not contain spontaneous contractions. 【0261】 Thresholding operation 1624 involves one or more adaptive thresholds. An adaptive threshold is a threshold based on one or more characteristics or features of the signal (waveform) to which the adaptive threshold is applied. Therefore, the adaptive threshold will vary depending on the statistical characteristics of the waveform. In one embodiment, two adaptive thresholds are applied: a uniform adaptive threshold and a dynamic adaptive threshold. The uniform adaptive threshold remains constant over the duration of the waveform 1610, while the dynamic adaptive threshold varies over the duration. This is illustrated in Figure 20. 【0262】 Figure 20 illustrates adaptive thresholding of a waveform according to an embodiment of the present disclosure. 【0263】 Figure 20 shows a plot of waveform 2002 (for example, the transformed representation of waveform 1610 obtained from transformation process 1622 shown in Figure 16) along with a uniform adaptive threshold 2004 and a dynamic adaptive threshold 2006. Figure 20 further shows the first point 2008, the second point 2010, the third point 2012, the fourth point 2014, and the fifth point 2016, all of which are points on waveform 2002. The first point 2008 and the second point 2010 exceed both the uniform adaptive threshold 2004 and the dynamic adaptive threshold 2006. The third point 2012 and the fourth point 2014 exceed the uniform adaptive threshold 2004 but not the dynamic adaptive threshold 2006. The fifth point 2016 exceeds the dynamic adaptive threshold 2006 but not the uniform adaptive threshold 2004. 【0264】 Thresholding operations (such as thresholding operation 1624 shown in Figure 16) determine one or both of the uniform adaptive threshold 2004 and the dynamic adaptive threshold 2006 for the waveform 2002 and use them to identify threshold overruns (local maximum values). These threshold overruns correspond to spontaneous contractions encoded within the waveform 2002. For example, if the thresholding operation requires both thresholds to be exceeded for a spontaneous contraction to be identified, then the first point 20020 and the second point 2010 on the waveform 2002, among the exemplary points highlighted in Figure 20, would be identified as locations of spontaneous contractions. 【0265】 The uniform adaptive threshold 2004 is constant over the duration of the waveform 2002 and represents a lenient threshold (i.e., more points in the waveform 2002 are identified as potential spontaneous contractions compared to a more conservative threshold). The uniform adaptive threshold 2004 is determined based on the statistical characteristics of the waveform 2002. The statistical characteristics are calculated over the entire duration of the waveform 2002, or substantially the entire duration. The statistical characteristics include one or more of the mean of the waveform 2002, the median of the waveform 2002, and the average of the maximum and minimum values ​​of the waveform 2002. For example, the uniform adaptive threshold may be set as the mean of the waveform. In one embodiment, statistical characteristics are weighted according to a weighting coefficient α = (0,1). Thus, the weighting coefficient is used to control the tolerance of the threshold. In one embodiment, the weighting coefficient is selected according to a manual fine-tuning process, thereby allowing the user to select a suitable value for α (i.e., a "holdout set" waveform) across a range of waveforms during the training or calibration phase. The value is then fixed when used in a system such as the system 1600 shown in Figure 16. 【0266】 The dynamic adaptive threshold 2006 fluctuates over the duration of the waveform 2002 and represents a conservative threshold (i.e., fewer points in the waveform 2002 are identified as potential spontaneous contractions compared to a more lenient threshold). Unlike the uniform adaptive threshold 2004, the dynamic adaptive threshold 2006 is determined for each subregion of the waveform 2002, thereby causing the dynamic adaptive threshold 2006 to fluctuate over the duration. In one embodiment, the dynamic adaptive threshold 2006 is calculated for each time point in the waveform 2002. Around each time point in the waveform 2002, a window is identified (e.g., a window containing 3, 5, or 10 points, or a window containing 0.5 seconds, 1 second, or 2 seconds, etc.), and a local threshold for that time point is determined using the portion of the waveform within the window. The local threshold is determined using the statistical properties of the portion of the waveform in the same manner as described in relation to the uniform adaptive threshold 2004 (e.g., mean, median, etc.). With respect to the uniform adaptive threshold 2004, in one embodiment, the statistical characteristics are weighted according to a weighting coefficient α = (0,1). In one embodiment, the weighting coefficient is selected according to a manual fine-tuning process, thereby allowing the user to select a suitable value for α (i.e., a "holdout set" waveform) across a range of waveforms during the training or calibration phase. The value is then fixed when used in a system such as the system 1600 shown in Figure 16. 【0267】 Utilizing both the uniform adaptive threshold 2004 and the dynamic adaptive threshold 2006 helps improve the accuracy of identifying individual spontaneous contractions within the waveform 2002 by reducing the number of false positives (e.g., the fourth point 2014 and the fifth point 2016). This improvement in accuracy can consequently lead to improved performance in downstream tasks such as contraction extraction, feature extraction, and use in drug discovery and development tasks. 【0268】 Referring again to Figure 16, the spontaneous contraction classifier 1606 determines a classification score 1614 based on the presence or absence of any spontaneous contractions in the waveform. For example, the classification score may be a binary value indicating the presence (e.g., +1) or absence (e.g., -1) of any spontaneous contractions in the waveform 1610. Alternatively, the classification score may be a vector of spontaneous contraction locations in the waveform 1610, where an empty vector indicates that no spontaneous contractions were identified in the waveform 1610. 【0269】 The profile generator 1608 generates a behavioral profile 1616 of the manipulated tissue based on the classification score 1614. In one embodiment, the behavioral profile 1616 includes an indication (e.g., a binary identifier) ​​of whether the waveform 1610 contains any spontaneous contractions. Additionally or alternatively, the behavioral profile 1616 includes a summary of spontaneous contractions within the waveform 1610, such as the number of spontaneous contractions and the average amplitude of the spontaneous contractions. Additionally or alternatively, the behavioral profile 1616 includes one or more spontaneous contractions within the waveform 1610 (or their locations within the waveform 1610). The behavioral profile 1616 can then be output for further processing. For example, the behavioral profile 1616 can be assigned to the waveform 1610 as a label or as the location of spontaneous contractions within the waveform 1610. 【0270】 In one embodiment, the behavior profile 1616 is used to extract one or more features related to spontaneous contraction from the waveform 1610. For example, features such as those described in relation to Figure 2 above are extracted around each spontaneous contraction location in the waveform 1610. These features then form a feature vector that serves as a descriptor of the tissue's (spontaneous) contraction response. This feature vector, or its transformation or summary, can then be used to carry out downstream drug discovery and development tasks such as estimating the effect of a compound or identifying a disease state. 【0271】 In another embodiment, the behavior profile 1616 is used as a feature vector to describe the conditions or states of the manipulated tissue in which the waveform 1610 is generated. That is, the behavior profile 1616, or summary statistics associated with the behavior profile, such as the number of spontaneous contractions, mean contractile force, etc., can be used as the phenotype of the manipulated tissue. Measuring changes in the behavior profile of the manipulated tissue under control and perturbation conditions (e.g., therapeutic conditions including drugs or compounds, disease states, etc.) (determined using system 1600) can be used to identify the effects associated with the perturbation conditions. 【0272】 Next, we will describe a method for determining the spontaneous behavior of an manipulated organization using the system components, processes, and operations described above. 【0273】 Figure 21 shows a method 2100 for determining the spontaneous behavior of an manipulated organization, according to an aspect of this disclosure. 【0274】 Method 2100 includes the steps of acquiring a first waveform 2102, applying a first classifier 2104, applying a second classifier 2106, and generating a behavior profile 2108. Method 2100 further includes an optional step 2110 for assigning the behavior profile and an optional step 2112 for outputting the behavior profile. In one embodiment, Method 2100 is carried out by the signal processing unit 112 of the control unit 104 shown in Figure 1, or another subunit. 【0275】 In acquisition step 2102, a first waveform (e.g., waveform 410 shown in Figure 16) is acquired. The first waveform includes the functional response of the manipulated tissue over a period of time. In one embodiment, the first waveform is acquired directly from the bioreactor in which the manipulated tissue is held (e.g., bioreactor 102 shown in Figure 1). Alternatively, the first waveform is acquired via an intermediate unit that converts observed measurements acquired from the bioreactor into a functional response waveform. The functional response of the manipulated tissue measured in the first waveform may be contractile force, displacement, calcium transient response, or similar. 【0276】 The manipulated tissue corresponds to any muscle-facilitating tissue or muscle tissue. Muscle tissue is grown within a bioreactor device (e.g., device 106 of bioreactor 102 shown in Figure 1) from cells seeded therein, such as induced pluripotent stem cells (iPSCs). In one embodiment, the artificial tissue is artificial heart tissue. More specifically, the manipulated tissue may be manipulated human heart tissue grown from human iPSC-derived cardiomyocytes and ventricular fibroblasts. Alternatively, the manipulated tissue is manipulated skeletal muscle tissue. 【0277】 In step 2104 of the application, a first classifier is applied to a first waveform, thereby generating a first classification score (for example, a periodic classifier 1602 applied to waveform 1610 shown in Figure 16 to generate a classification score 1612). The first classification score indicates whether the waveform contains periodic contractions of the manipulated tissue over a period of time. For this reason, the first classifier is alternatively referred to as a periodic classifier, a frequency-based global classifier, or a global classifier. 【0278】 The first classifier predicts the overall characteristics of the waveform acquired in step 2102, i.e., whether the waveform contains any periodic (regular or induced) contractions, such as those described in relation to Figure 15A above. The first classifier includes any suitable predictive model that can predict from the time series input whether the time series input contains regular or periodic responses (peaks or contractions). 【0279】 The first classifier includes any suitable predictive model that can predict from the time series input whether the time series input contains regular or periodic responses (peaks or contractions). In one embodiment, as described in more detail with reference to Figure 22 below, the first classifier applied in step 2104 includes a spectral transformation process and a predictive model (e.g., the spectral transformation process 1618 and predictive model 1620 of the periodic classifier 1602 shown in Figure 16). 【0280】 In step 2106 of the application, a second classifier is applied to the waveform to generate a second classification score. The second classifier is applied if the first classification score indicates that no periodic contractions are present in the waveform (for example, if classification score 412 indicates that no induced or periodic contractions are present in waveform 1610, the spontaneous contraction classifier 1606 is applied to waveform 1610 shown in Figure 16). Thus, in one embodiment, method 2100 further includes a step of determining whether the waveform contains one or more periodic contractions based on the classification score obtained in step 2104 of the application, prior to step 2106 of the application. Step 2106 of applying the second classifier is performed if periodic contractions are not identified as present in the waveform (otherwise, method 2100 terminates by returning, for example, a preferred indication or notification that the waveform is periodic). The second classification score determined by the second classifier indicates whether the waveform contains spontaneous contractions of the manipulated tissue over a period of time. The second classifier is alternatively referred to as a spontaneous contraction classifier, local classifier, or peak detection. 【0281】 The second classification score generated by the second classifier indicates whether the waveform acquired in the 2102 acquisition steps contains any spontaneous contractions of the manipulated tissue over the period. It is possible that no tissue contraction response is observed in the waveform (i.e., the tissue did not exhibit induced or spontaneous contractions). In such situations, both the first and second classification scores would indicate the absence of either induced or spontaneous contractions in the waveform. Alternatively, the waveform may contain one or more spontaneous contractions but no induced contractions (as illustrated in Figure 15B). The second classifier uses the waveform acquired in the 2102 acquisition steps to determine which of these two behaviors is exhibited by the tissue in the waveform. The second classification score can be a binary value indicating the presence (e.g., +1) or absence (e.g., -1) of any spontaneous contractions in the waveform. Alternatively, the second classification score can be a vector of spontaneous contraction locations in the waveform, where an empty vector indicates that no spontaneous contractions are identified in the waveform. 【0282】 As will be described in more detail with reference to Figure 23 below, in one embodiment, the second classifier includes a transformation process and a thresholding operation (e.g., the transformation process 1622 and thresholding operation 1624 of the spontaneous contraction classifier 1606 shown in Figure 16). 【0283】 In the generation step 2108, a behavioral profile of the manipulated tissue during the period is generated based on a second classification score. In one embodiment, the behavioral profile includes an indication (e.g., a binary identifier) ​​of whether the waveform acquired in the acquisition step 2102 includes any spontaneous contractions. Additionally or alternatively, the behavioral profile includes one or more spontaneous contractions associated with one or more portions of the waveform associated with the spontaneous contraction locations. 【0284】 In the optional assignment step 2110, the behavior profile is assigned to the waveform acquired in the acquisition step 2102. For example, the behavior profile is assigned as a label to the waveform, so that the waveform and label can be used for further processing or classification (e.g., within a drug discovery or development system). Additionally or alternatively, if the behavior profile includes the locations of spontaneous contractions within the waveform, the spontaneous contractions can be assigned to the waveform by identifying the locations of spontaneous contractions within the waveform (e.g., using metadata, position vectors, or similar). 【0285】 In an optional step 2112 of outputting, a behavior profile is output. In one embodiment, outputting a behavior profile includes storing or saving the behavior profile in persistent storage such as non-volatile memory, non-temporary media, or the like. Additionally or alternatively, outputting a behavior profile includes transmitting the behavior profile over a network (e.g., a local area network, a wide area network, and the like) or displaying the behavior profile for user review. In one embodiment, the behavior profile is output together with the waveform. 【0286】 Figure 22 shows a method 2200 for obtaining a classification score from a waveform according to an embodiment of the present disclosure. 【0287】 In one embodiment, Method 2200 is performed in step 2104, in which a first classifier is applied in Method 2100 to generate a first classification score. Thus, Method 2200 includes a step performed by a frequency-based global classifier (periodic classifier or global classifier) ​​according to one embodiment (e.g., a periodic classifier 1602 shown in Figure 16). Method 2200 includes a step 2202 to transform a waveform to generate a transformed waveform, a step 2204 to apply the transformed waveform to a predictive model, and a step 2206 to obtain a classification score from the predictive model. 【0288】 In step 2202 of the transformation, a spectral transformation process is applied to the waveform (i.e., the waveform acquired in step 2102 of the acquisition of method 2100) to generate a spectral waveform (e.g., the spectral transformation process 418 applied to waveform 1610 as part of the periodic classifier 1602 in Figure 16). The spectral transformation process is any suitable spectral transformation, such as a Fourier transform, that produces a frequency-based representation of the waveform. Exemplary spectral waveforms are shown in Figures 17A and 17B. 【0289】 In step 2204, the spectral waveform generated in step 2202, which is converted, is input to the prediction model. The prediction model is a trained machine learning model, such as a trained neural network, support vector machine, random forest, or similar. In one embodiment, the prediction mode is a convolutional neural network, such as the one shown in Figure 18 and described in more detail above. 【0290】 In acquisition step 2206, the classification score is obtained from a predictive model based on the spectral waveform. The classification score represents the contractile behavior of the tissue associated with the waveform (i.e., associated with the spectral waveform) and indicates whether the waveform includes regular or periodic contractions. In one embodiment, the classification score is binary, taking one score or value (e.g., "+1") indicating that the waveform includes periodic contractions and another score or value (e.g., "0" or "-1") indicating that the waveform does not include periodic contractions. 【0291】 Figure 23 shows a method 2300 for obtaining a classification score from a waveform according to an embodiment of the present disclosure. 【0292】 In one embodiment, Method 2300 is carried out in step 906, which involves applying a second classifier to generate a second classification score in Method 900. Thus, Method 2300 includes steps carried out by a local classifier according to one embodiment (a spontaneous contraction classifier or peak detector) (e.g., a spontaneous contraction classifier 1606 shown in Figure 16). Method 2300 includes steps 2302, which involves transforming a waveform to generate a transformed waveform; step 2304, which involves determining one or more adaptive thresholds; and step 2306, which involves determining a classification score based on one or more adaptive thresholds applied to the transformed waveform. 【0293】 In the conversion step 2302, the conversion process is applied to the waveform (i.e., the waveform acquired in step 2102 of the acquisition method 2100 shown in Figure 21) to generate a converted waveform (e.g., the conversion process 1622 applied to the waveform 1610 as part of the spontaneous contraction classifier 1606 shown in Figure 16). The conversion process includes any preferred signal processing technique to reduce noise while simultaneously enhancing peaks in the waveform. In one embodiment, the conversion process includes a Pan-Tompkins algorithm. In further embodiments, the conversion process includes a modified Pan-Tompkins algorithm, as described in more detail with reference to Figure 24 below. 【0294】 In the determination step 2304, one or more adaptive thresholds are determined based on the transformed waveform. Therefore, the values ​​of one or more adaptive thresholds depend on the transformed waveform. The adaptive thresholds include one or more uniform adaptive thresholds and dynamic adaptive thresholds. A uniform adaptive threshold is constant over the duration of the waveform, while a dynamic adaptive threshold varies over the duration. Figure 20 shows exemplary uniform and dynamic adaptive thresholds as described above. 【0295】 In the determination step 2306, a classification score is determined based on the adaptive thresholds applied to the transformed waveform. Therefore, the thresholding operation with the adaptive threshold(s) determined in the determination step 2304 is performed in the determination step 2306 (for example, the thresholding operation 1624 of the spontaneous contraction classifier 1606 shown in Figure 16). The second classification score determined in the determination step 2306 indicates a waveform containing spontaneous contraction if one or more parts of the transformed waveform exceed at least one of the adaptive thresholds. Alternatively, the second classification score indicates a waveform containing spontaneous contraction if one or more parts of the transformed waveform exceed both the uniform threshold and the dynamic adaptive threshold. For example, if no part of the transformed waveform exceeds the adaptive threshold(s), a classification score of "0" is determined, but if at least one part of the transformed waveform exceeds the adaptive threshold(s), a classification score of "+1" is determined (i.e., a classification score indicating the presence of one or more spontaneous contractions in the waveform). Alternatively, the classification score includes a vector of positions within the waveform corresponding to identified spontaneous contractions (where an empty vector indicates that no spontaneous contractions were identified). 【0296】 Figure 24 shows a modified Pans-Tompkins 2400 according to an embodiment of the present disclosure. 【0297】 The modified Pan-Tompkins method 2400 includes a noise reduction process 2402 and a signal enhancement process 2404. The noise reduction process 2402 includes a step 2406 of filtering using a low-pass filter, a step 2408 of filtering using a high-pass filter, and a step 2410 of transforming using a steady-state wavelet transform. The signal enhancement process 2404 includes a step 2412 of differentiating, a step 2414 of rectifying, and a step 2416 of integrating. The modified Pan-Tompkins method 2400 further includes an optional step 2418 of normalizing the output of the signal enhancement process 2404. In one embodiment, the modified Pan-Tompkins method 2400 is performed in step 2302 of transforming the waveform in method 2300. 【0298】 The modified Pan-Tompkins algorithm is described as performing each of the steps described below, but those skilled in the art will understand that in some embodiments the steps can be combined and / or omitted. For example, the algorithm may include performing a steady wavelet transform and rectification operation, or a steady wavelet transform, differentiation operation and rectification operation. 【0299】 The noise reduction process 2402 of the modified Pan-Tompkins algorithm 2400 generates a noise-filtered representation of the waveform. The noise reduction process 2402 includes applying a bandpass filter to the waveform (e.g., filtering step 2406 and filtering step 2408), and, in contrast to the standard Pan-Tompkins algorithm, then transforms the output of the bandpass filter using a stationary wavelet transform 2410. 【0300】 In filtering step 2406, a low-pass filter is applied to the waveform to generate a low-pass filtered waveform. The low-pass filter helps remove major wire mismatches from the waveform. An exemplary low-pass filtered waveform is shown by low-pass filtered waveform 1904 in Figure 19, which corresponds to the result of applying a low-pass filter to waveform 1902. In one embodiment, the low-pass filter comprises a two-dimensional Gaussian kernel with a standard deviation of 2. 【0301】 In filtering step 2408, a high-pass filter is applied to the low-pass filtered waveform (generated in filtering step 2406) to generate a high-pass filtered waveform. The high-pass filter helps to perform baseline alignment while removing drift. An exemplary high-pass filtered waveform is shown by the high-pass filtered waveform 1906 in Figure 19, which corresponds to the result of applying a high-pass filter to the low-pass filtered waveform 1904. In one embodiment, the high-pass filter comprises an elliptic filter with order 8, ripple of 0.5 dB, attenuation of 40 dB, and edge frequencies of 1 and 20. 【0302】 In the conversion step 2410, a steady wavelet transform is applied to the high-pass filtered waveform (generated in step 2408 of filtering) to generate the converted waveform. The steady wavelet transform helps to improve the denoising performed in the noise reduction process 2402, thereby improving the classification performance of the spontaneous collapse classifier. An exemplary converted waveform is shown by the wavelet-transformed waveform 1908 in Figure 19, which corresponds to the result of applying a steady wavelet transform to the high-pass filtered waveform 1906. In one embodiment, the wavelet transform includes a discrete steady wavelet transform (1D) using a five-step decomposition and a Daubechies 4 (db4) wavelet. 【0303】 The signal enhancement process 2404 of the modified Pan-Tompkins algorithm 2400 produces an enhanced representation of the noise-filtered waveform produced by the noise reduction process 2402. In contrast to the standard Pan-Tompkins algorithm, the squaring operation is replaced by a rectifying step 2414. 【0304】 In the differentiation step 2412, the noise reduction waveform obtained from the noise reduction process 2402 is differentiated using a differential operation to generate a differentiated waveform. The differential operation is used to highlight rapid changes (i.e., contractions). An example of a differentiated waveform is shown by the differentiated waveform 710 in Figure 19, which corresponds to the result of differentiating the wavelet-transformed waveform 1908. In one embodiment, the differential operation is performed using the transfer function H(z)=(1 / 8T)(-z 2 -2z -1 +2z 1 +z 2 This corresponds to the 5-point derivative with ), where T is the sampling period. 【0305】 In step 2414, the differentiated waveform is rectified, thereby generating the rectified waveform. That is, in step 2414, the differentiated waveform is clipped so that any negative portion of the differentiated waveform is removed. Thus, the rectification operation enhances the dominant peaks in the waveform. An example of the rectified waveform is shown by the rectified waveform 1912 in Figure 19, which corresponds to the result of rectifying the differentiated waveform 1910. 【0306】 In the integration step 2416, a moving window integration operation is applied to the rectified waveform to generate the integrated waveform. The moving window integration operation removes short-duration artifacts from the rectified waveform. An example of the integrated waveform is shown by the integrated waveform 1914 in Figure 19, which corresponds to the result of applying a moving window integration operation to the rectified waveform 1912. In one embodiment, the moving average filter is, 【number】 It features a moving average filter calculated over a sliding window, where n is the length of the waveform 1610. 【0307】 In the optional normalization step 2418, the output of the signal enhancement process 2404 (e.g., the integrated waveform produced in step 2416) is normalized to scale the waveform values ​​to a set range along the y-axis (e.g., 0 to 1). 【0308】 A system and method for classifying the type of shrinkage of manipulated tissue. This disclosure presents a system and method for classifying and processing functional response waveforms to enable efficient and effective extraction of features across a range of contraction types. 【0309】 Figure 25 illustrates a dual contraction-relaxation cycle model according to an embodiment of the present disclosure. 【0310】 Figure 25 shows the single contraction model 2508 and the double contraction model 2510 (alternatively referred to as the double contraction-relaxation cycle model, the double type model, or the double model), as described in detail above in relation to Figure 3. D (t) is a single contraction model, 【number】 This includes combinations of the following. Each of the single contraction models has its own set of parameters. 【number】 Defined by, this means that the overall double contraction model 2510 is parameterized by 【number】 It is parameterized by . In this model, the y-axis shift B is shared across both of a single model, although in some embodiments, B can vary between models. By modeling a functional response having a dual contraction type using the dual contraction model 2510, the individual components of the contraction response can be modeled and therefore varied independently. This makes it possible to efficiently and effectively model a functional response of the dual contraction type, and therefore improve the performance of downstream tasks that utilize such a model. 【0311】 Figure 26 shows a system 2600 for processing functional response waveforms having different contraction types, according to an embodiment of the present disclosure. 【0312】 System 2600 comprises a contraction type classifier 2602, a single contraction model 2604, a dual contraction model 2606, and a waveform generator 2608. System 2600 receives an input waveform 2610 and produces an output waveform 2612 containing a noise-filtered representation of the input waveform 2610. The input waveform 2610 includes at least one contraction response and at least one relaxation response of the manipulated tissue. The input waveform 2610 has a length corresponding to the expected contraction-relaxation cycle of the manipulated tissue and includes either a single contraction of the manipulated tissue (i.e., the input waveform is a single contraction type) or a dual contraction of the manipulated tissue (i.e., the input waveform is a dual contraction type). To identify the relevant contraction type, the input waveform 2610 is classified by the contraction type classifier 2602 into a predicted contraction type 2614 (alternatively referred to as the predicted contraction-relaxation cycle type, predicted type, or predicted cycle type). Based on the predicted contraction type 2614, either a single contraction model 2604 or a dual contraction model 2606 is fitted to the input waveform 2610, resulting in either a fitted single model 2616 or a fitted dual model 2618. The waveform generator 2608 utilizes the fitted model (i.e., fitted single model 2616 or fitted dual model 2618) to generate an output waveform 2612 containing a noise-filtered representation of the input waveform 2610. For example, the waveform generator 2608 samples several points (e.g., 100, 200, 500, 1000, etc.) from the fitted model (i.e., fitted single model 2616 or fitted dual model 2618) over a period of the input waveform 2610 to generate the output waveform 2612. Beneficially, this makes it possible to generate high-resolution waveform data, which helps ensure that more accurate features of the functional response of the manipulated tissue are extracted. This, in turn, helps improve the accuracy and effectiveness of downstream tasks involving functional response features. 【0313】 In one embodiment, the input waveform 2610 is extracted from a larger waveform of the functional response using a method such as that described in relation to Figure 33 below. 【0314】 The fitted single model 2616 and fitted dual model 2618 include parameter values ​​for the associated contraction-relaxation cycle model. In one embodiment, the parameter values ​​of the fitted single model 2616 or fitted dual model 2618 are further refined using optimization techniques. As described in more detail below in relation to Figure 29, the optimization techniques used include simplex search algorithms such as the Nelder-Mead method. 【0315】 In one embodiment, one or more feature values ​​are extracted from the output waveform 2612 and output. Examples of one or more feature values ​​extracted from the output waveform 2612 include one or more of the following: single contraction, or one or more of the following: peak amplitude value (i.e., peak amplitude 206 shown in Figure 2), contraction time value (i.e., time to peak amplitude 208 shown in Figure 2), maximum contraction slope value (i.e., maximum development rate 214 shown in Figure 2), relaxation time value (i.e., time to peak decline 210 shown in Figure 2), maximum relaxation slope value (i.e., maximum decline rate 216), and single contraction duration value (i.e., duration 212). For waveforms corresponding to a dual contraction type, the feature may include two values ​​for each feature extraction. For example, a first peak amplitude value associated with a first peak in the dual contraction type waveform, and a second peak amplitude value associated with a second peak in the dual contraction type waveform. 【0316】 Single or dual contraction-relaxation models can be fitted using any suitable parameter fitting technique, such as least-squares-based fitting, or machine learning-based fitting, such as that shown in Figure 6. 【0317】 Figure 27 shows the classification of contraction types for three waveforms according to embodiments of the present disclosure. 【0318】 Figure 27 shows the baseline waveform 2702, the first perturbation waveform 2704, and the second perturbation waveform 2706. The waveforms include the functional response (contractile force) of the artificial tissue under baseline conditions and two perturbation conditions. Examples of perturbation conditions include treatment of the tissue with a drug or compound, different cell lines, different disease states, physical perturbations of the artificial tissue, changes to a platform or bioreactor containing the artificial tissue, and similar. The baseline waveform 2702, the first perturbation waveform 2704, and the second perturbation waveform 2706 are associated with baseline classification vector 2708, the first perturbation classification vector 2710, and the second perturbation classification vector 2712, respectively. The baseline classification vector 2708 includes a first classification value 2714, which includes multiple classification values ​​associated with the functional response waveform in baseline waveform 2702, e.g., the contractile type classification of the relevant functional response waveform 2716 in baseline waveform 2702. The first perturbation classification vector 2710 includes multiple classification values ​​associated with functional response waveforms within the first perturbation waveform 2704, such as the second classification value 2718 and the third classification value 2720. The second classification value 2718 and the third classification value 2720 include contraction type classifications of the associated functional response waveforms 2722 and 2724 within the first perturbation waveform 2704. The second perturbation classification vector 2712 includes multiple classification values ​​associated with functional response waveforms within the second perturbation waveform 2706. 【0319】 The classification vector (e.g., baseline classification vector 2708) is obtained by extracting multiple functional response waveforms from each waveform (as described in more detail below in relation to Figure 33) and then using a predictive model (e.g., predictive model 600 described above in relation to Figures 6 and 7) to determine the contraction type of each functional response waveform. The contraction type is either a single contraction type or a double contraction type. As shown in Figure 27, the first classification value 2714 and the second classification value 2718 are single contraction types (as indicated by the light gray shading) because the predictive model determines that the associated functional response waveforms 2716 and 2722 contain a single contraction. In contrast, the third classification value 2720 is a double contraction type (as indicated by the dark gray shading) because the predictive model determines that the associated functional response waveform 2724 contains a double contraction. 【0320】 As described above, the waveforms include the functional response of the prosthesis under baseline conditions and two perturbation conditions. Therefore, any change in the distribution of contraction types within the classification vectors indicates a change in the behavior of the prosthesis resulting from the perturbation of the prosthesis. In the example shown in Figure 27, baseline waveform 2702 includes the functional response of the prosthesis under reference or control conditions, first perturbation waveform 2704 includes the functional response of the prosthesis under a first dose of isoproterenol, and second perturbation waveform 2706 includes the functional response of the prosthesis under a second dose of isoproterenol. As can be seen by comparing the first perturbation classification vector 2710 and the second perturbation classification vector 2712 with the baseline classification vector 2708, the contraction behavior of the prosthesis changes with increasing doses of isoproterenol. As most clearly shown in the second perturbation classification vector 2712, isoproterenol induces regular double contractions of the prosthesis that are not present in the behavior of the prosthesis under control conditions. 【0321】 Here, we will describe a method for using the above-mentioned system and model for classifying and processing functional response waveforms. 【0322】 Figure 28 shows a method 2800 for processing a functional reaction according to an aspect of the present disclosure. 【0323】 Method 2800 includes the steps of acquiring a first waveform 2802, determining a predicted contraction type 2804, fitting a model to the first waveform based on the predicted contraction type 2806, and generating a second waveform from the model 2808. Method 2800 further includes an optional step 2810 for extracting feature values ​​from the second waveform and an optional step 2812 for outputting feature values. In one embodiment, Method 2800 is performed by a model fitting unit 104-1 of a control unit 104 shown in Figure 1. 【0324】 Generally, Method 2800 is used to generate a noise-filtered or noise-suppressed representation of a contraction-relaxation cycle waveform (functional response waveform), which may include either a single contraction-type response or a dual contraction-type response. In embodiments, the contraction-relaxation cycle waveform is obtained from hardware such as a bioreactor (e.g., bioreactor 102 shown in Figure 1), which may introduce noise into the waveform due to sensor variability, signal transmission, signal conversion, and the like. Thus, the contraction-relaxation cycle waveform contains a potentially noisy representation of the functional response of the manipulated tissue during a single contraction-relaxation cycle. Beneficially, Method 2800 efficiently classifies the contraction type of the waveform and efficiently generates a noise-filtered representation of the contraction-relaxation cycle, thereby improving the accuracy of the extracted features and thereby helping to improve the performance of downstream tasks that utilize such features. 【0325】 In acquisition step 2802, a first waveform is acquired. The first waveform includes at least one contraction response and at least one relaxation response of the manipulated, i.e., artificial tissue. The first waveform has a predetermined length corresponding to the expected length of the contraction-relaxation cycle of the manipulated tissue. 【0326】 The first waveform (or first functional response waveform) captures the functional response of the artificial tissue. The functional response is the contractile force of the artificial tissue (e.g., the contractile force measured using data acquired from the sensor assembly 108 of the bioreactor 102, in which the artificial tissue grows or is maintained). Alternatively, the functional response may be a contractile displacement, a transient calcium response, or a change in membrane potential. 【0327】 Artificial or manipulated tissue includes manipulated muscle tissue such as manipulated cardiac tissue or manipulated skeletal muscle tissue. In one embodiment, the first waveform is obtained from a bioreactor containing the artificial tissue (e.g., bioreactor 102 shown in Figure 1). Thus, the first waveform is obtained from a waveform obtained from the bioreactor, which includes multiple single or double contraction-relaxation cycles of the artificial tissue. In one embodiment, the first waveform is obtained or extracted from the waveform using an extraction method such as method 3300, which is described in more detail below in relation to Figure 33. 【0328】 A predetermined length of the first waveform, corresponding to the expected length of the contraction-relaxation cycle, is between 0.05 seconds and 10 seconds, more specifically between 0.15 seconds and 1 second. In one embodiment, the predetermined length is proportional to the frequency at which the manipulated tissue is stimulated when the first waveform is recorded. For example, if the manipulated tissue is stimulated at 1 Hz, the predetermined length is 1 second, since the expected length of the contraction-relaxation cycle of the manipulated tissue is 1 second. 【0329】 In the determination step 2804, the predicted contraction type among multiple contraction types is determined for the first waveform. The multiple contraction types include single contraction types and dual contraction types. 【0330】 The predicted deflation type is determined by the classifier based on the first waveform input to the classifier. As described in more detail above in relation to Figures 6 and 7A-7D, the classifier includes a trained machine learning model, such as a trained neural network. In one embodiment, the classifier includes a convolutional neural network, an expanded convolutional neural network, and a long short-term memory network. Further architectural details of the classifier are given above in relation to Figures 6 and 7A-7D. 【0331】 In fitting step 2806, the model is fitted to the first waveform based on the predicted contraction type. The model parameterizes the growth of at least one contraction response in the manipulated tissue independently of the growth of at least one relaxation response in the manipulated tissue. Therefore, the model does not assume that the underlying functional response is symmetric. This allows the model to efficiently and accurately fit a wide variety of functional responses from diverse manipulated tissue types. More efficient and accurate model fitting helps generate more accurate features, which in turn leads to the generation of better data. 【0332】 The model fitted to the first waveform is selected based on the predicted contraction type. If the predicted contraction type is a single contraction type, a single contraction type model is fitted (as illustrated and described in relation to Figure 3 above). If the predicted contraction type is a double contraction type, a double contraction type model is fitted (as illustrated and described in relation to Figure 25 above). 【0333】 A single contraction type model uses the first contraction function, f c (t), and the first relaxation function, f r(t) is included. A dual contraction-type model includes a second contraction function and a second relaxation function, in addition to the first contraction function and the first relaxation function. In one embodiment, the contraction function is an ascending logistic function with a positive growth rate, and the relaxation function is a descending logistic function with a negative growth rate. Thus, the contraction response of a single or dual contraction-relaxation cycle is modeled by an ascending logistic function, and the relaxation response of a single or dual contraction-relaxation cycle is modeled by a descending logistic function. A single contraction-type model includes the product of the first ascending logistic function and the first descending logistic function. A dual contraction-type model includes a combination (i.e., an additive combination) of two single contraction-type models. 【0334】 The contraction-relaxation cycle model includes multiple parameters associated with at least one contraction response and at least one relaxation response. These multiple parameters include at least one maximum parameter A and at least one rate of increase parameter k. c , at least one rate of decline parameter k r , y-shift parameter B, at least one ascending x-shift parameter t0, and at least one descending x-shift parameter t d This includes the relationship between each of these parameters and the overall single or dual contraction-relaxation cycle models, which is shown and described in more detail in relation to Figures 3 and 25 above. 【0335】 In one embodiment, the model is fitted to a first waveform using predictions of parameter values ​​obtained from a machine learning model (e.g., the predictive model 600 shown in Figure 6), as described in more detail in relation to method 2900 in Figure 29 below. 【0336】 In step 2808, the second waveform is generated from a model fitted to the first waveform such that the second waveform contains a noise-filtered representation of the first waveform. For example, over the duration of the first waveform, several points (e.g., 100, 200, 500, 1000, etc.) are sampled from the model fitted in step 2806. Beneficially, this makes it possible to generate high-resolution waveform data from the contraction-relaxation cycle, which helps ensure that more accurate features of the functional response of the manipulated tissue are extracted. This, in turn, helps improve the accuracy and effectiveness of downstream tasks involving functional response features. 【0337】 Optionally, a second waveform may be output. In one embodiment, outputting the second waveform includes storing or saving the second waveform in persistent storage such as non-volatile memory, non-temporary media, or the like. Additionally or alternatively, outputting the second waveform includes transmitting the second waveform over a network (e.g., a local area network, a wide area network, and the like) or displaying the waveform for user review. 【0338】 In the optional extraction step 2810, one or more feature values ​​are extracted from the second waveform. 【0339】 The second waveform is a noise-filtered representation of the first waveform (i.e., a noise-filtered representation of a single or double contraction-relaxation cycle), thus allowing for the extraction of more accurate values ​​of the underlying functional response features. One or more feature values ​​include at least one contraction-relaxation cycle, or one or more of the following: peak, amplitude value (i.e., peak amplitude 206 shown in Figure 2), contraction time value (i.e., time to peak amplitude 208 shown in Figure 2), maximum contraction slope value (i.e., maximum development rate 214 shown in Figure 2), relaxation time value (i.e., time to peak decline 210 shown in Figure 2), maximum relaxation slope value (i.e., maximum decline rate 216 shown in Figure 2), and contraction-relaxation cycle duration value (i.e., duration 212 shown in Figure 2). 【0340】 Once extracted, one or more feature values ​​can be used as quantitative descriptors for single or dual contraction-relaxation cycles, thus providing a numerical representation of the functional response of the artificial tissue. As described in more detail in relation to Figure 35, such features can be used in various downstream processing tasks, such as efficacy identification in drug discovery and development. 【0341】 In an optional output step 2812, one or more feature values ​​extracted from the second waveform are output. In one embodiment, outputting one or more feature values ​​includes storing or saving one or more feature values ​​in persistent storage such as non-volatile memory, non-temporary media, or the like. Additionally or alternatively, outputting one or more feature values ​​includes transmitting one or more feature values ​​over a network (e.g., a local area network, a wide area network, and the like) or displaying one or more feature values ​​for user review. 【0342】 Figure 29 shows a method 2900 for fitting a model to a functional response waveform according to an embodiment of the present disclosure. 【0343】 In one embodiment, method 2900 is performed as part of fitting step 906 of method 900. Method 2900 includes step 2902 of predicting multiple values ​​for multiple parameters and further includes an optional step 2904 of optimizing the multiple values. 【0344】 The steps of Method 2900 are used to predict the parameter values ​​of a model from a first waveform. The model is either a single-collapse type model (Figure 3) or a dual-collapse type model (Figure 25). Which model is fitted by Method 2900 depends on the predicted collapse type determined in step 2804 of Method 2800. A trained machine learning model (e.g., the parameter estimation model shown in Figure 6) is used to predict the parameter values ​​so that the fitted model closely approximates the first waveform. 【0345】 In prediction step 2902, multiple values ​​for multiple parameters of the model are predicted so that the model fitted to the first waveform contains multiple values ​​for multiple parameters. These multiple values ​​are predicted by either a single contraction type model or a dual contraction type model. 【0346】 In the optional optimization step 2904, the multiple values ​​determined in the prediction step 2902 are optimized by minimizing the error between the first waveform and the model fitted to the first waveform (i.e., using multiple values ​​for multiple parameters of the model). 【0347】 The model determined in prediction step 2902 【number】 If multiple values ​​are given for the value, the updated multiple values 【number】 teeth, 【number】 This is determined in step 2904, which optimizes to be such that L θ (·) represents the first waveform X1 and the model f fitted to the first waveform according to the set of parameter values ​​θ. θ This is a cost function or loss function that measures the error between the first waveform and the second waveform. A smaller value of L indicates that the model is better fitted to the first waveform. In one embodiment, the cost function L is the root mean square error as follows: 【number】 In one embodiment, this corresponds to the following: 【number】 【0348】 Optimizing θ is a multidimensional problem because the model involves multiple parameters. Therefore, L θ Minimizing requires the simultaneous fitting of multiple parameters. To perform this minimization, in one embodiment, multiple values ​​are optimized using a simplex search algorithm such as the Nelder-Mead method. Beneficially, performing the optimization after obtaining initial predictions of the parameters helps to obtain an accurate model fit that helps avoid local minima while making better use of processing resources, because the optimization process starts with a solution that is expected to be close to the optimal solution. 【0349】 Figure 30 shows a method for training a classifier to predict tissue contraction type from functional response waveforms. 【0350】 Method 3000 includes the steps of acquiring multiple waveforms (step 3002), extracting a first set of multiple parameter sets (step 3004), extracting a second set of multiple parameter sets (step 3006), determining the distribution of the multiple parameter sets (step 3008), generating a synthetic training dataset (step 3010), and training a classifier on the synthetic training dataset (step 3012). In one embodiment, Method 3000 is carried out by a control unit 104 or a subunit thereof, as shown in Figure 1. 【0351】 In many situations, the effectiveness of a predictive model (e.g., a machine learning model such as a deep neural network) is limited by the quantity and quality of the available training data. Without a large amount of high-quality training data, predictive models will often be unable to produce adequate outputs. In this disclosure, this problem leads to inaccurate identification of single or dual types of voids, which subsequently reduces the effectiveness and applicability of actually using such identifiers for tasks such as drug discovery and development. Method 3000 attempts to address such a problem by generating high-fidelity synthetic training data, thereby enabling the generation of a virtually unlimited amount of data. This helps improve the performance of the predictive model being trained, and subsequently improves the model's predictive accuracy. This improvement in accuracy helps drive improvements in downstream tasks that utilize classifications obtained from such models. 【0352】 In acquisition step 3002, multiple waveforms are acquired. The multiple waveforms include the functional response of one or more manipulated tissues. Each of the multiple waveforms has a predetermined length corresponding to the expected length of the contraction-relaxation cycle of the manipulated tissue. 【0353】 In extraction step 3004, a first set of parameters is extracted from a first subset of multiple waveforms associated with a single contraction type. The first parameter set of the first set of parameters characterizes the first waveform of the first subset of waveforms. 【0354】 In extraction step 3006, a second set of parameters is extracted from a second subset of the multiple waveforms associated with the dual contraction type. The second set of parameters from the second set of parameters characterizes the second waveform from the second subset of waveforms. 【0355】 In the determination step 3008, multiple parameter set distributions are determined. These multiple parameter set distributions include a first parameter set distribution determined from a first set of multiple parameter sets, and a second parameter set distribution determined from a second set of multiple parameter sets. 【0356】 In step 3010, a synthetic training dataset is generated. Each element of the synthetic training dataset includes a synthetic waveform and the corresponding tissue contraction type associated with the synthetic waveform. The synthetic waveform is generated using the parameter set distributions of multiple parameter set distributions associated with the corresponding tissue contraction type. 【0357】 In training step 3012, the classifier is trained using a synthetic training dataset. The classifier, trained using the synthetic training dataset, determines the predicted tissue contraction type for the input waveform. 【0358】 Figure 31 shows a method 3100 for predicting the tissue contraction type of a waveform using a synthetically trained classifier, according to an embodiment of the present disclosure. 【0359】 Method 3100 includes the steps of acquiring a first waveform 3102 and predicting the tissue contraction type 3104. Method 3100 also includes an optional step 3106 for outputting the tissue contraction type. In one embodiment, Method 3100 is carried out by a control unit 104 shown in Figure 1, or by a subunit such as a model fitting unit 104-1. 【0360】 In acquisition step 3102, a first waveform is acquired. The first waveform includes the functional response of the first prosthesis. The first waveform has a length corresponding to the expected length of the contraction-relaxation cycle of the first prosthesis. 【0361】 In prediction step 3104, the tissue contraction type is predicted from the first waveform using a classifier trained on a synthetically generated training dataset (as described above in relation to Figure 30). 【0362】 In the optional output step 3106, the tissue shrinkage type is output. In one embodiment, outputting the tissue shrinkage type includes storing or saving the tissue shrinkage type in persistent storage such as non-volatile memory, non-temporary media, or the like. Additionally or alternatively, outputting the tissue shrinkage type includes transmitting the tissue shrinkage type over a network (e.g., a local area network, a wide area network, and the like) or displaying the tissue shrinkage type for user review. 【0363】 Figure 32 shows a method 3200 for training a parameter estimation model using synthetic training data, according to one aspect of the present disclosure. 【0364】 Method 3200 includes the steps of acquiring multiple waveforms (step 3202), extracting multiple parameter sets (step 3204), determining the parameter set distribution (step 3206), generating a synthetic training dataset (step 3208), and training a predictive model on the synthetic training dataset (step 3210). In one embodiment, Method 3200 is carried out by the control unit 104 or its subunits shown in Figure 1. 【0365】 In many situations, the effectiveness of a predictive model (e.g., a machine learning model such as a deep neural network) is limited by the quantity and quality of the available training data. Without large amounts of high-quality training data, predictive models will often be unable to produce adequate outputs. In this disclosure, this problem leads to the fitting of inaccurate contraction-relaxation models, which will then reduce the effectiveness and applicability of such models when actually used for tasks such as drug discovery and development. Method 3200 attempts to address such a problem by generating high-fidelity synthetic training data, thereby enabling the generation of virtually unlimited amounts of data. This helps improve the performance of the predictive model being trained, which in turn improves the accuracy of the model fitted using the predictive model. This improvement in accuracy helps facilitate improvements in downstream tasks that utilize features extracted from such models. 【0366】 In acquisition step 3202, multiple waveforms are acquired. The multiple waveforms represent the functional responses of one or more artificial tissues. All of the multiple waveforms contain a common contraction type among a given set of contraction types. For example, all of the waveforms are either single-contraction type waveforms or dual-contraction type waveforms. Thus, the model trained by method 3200 is trained to predict the parameters of either a single-contraction type model or a dual-contraction type model, depending on the common contraction type of the multiple waveforms. 【0367】 The multiple waveforms correspond to real-world or bootstrap data from which the synthetic training dataset will be generated. Each waveform corresponds to a time series of values ​​representing the functional response (e.g., contractile force, calcium transient, etc.) of the artificial or manipulated tissue over one single contraction-relaxation cycle or one double contraction-relaxation cycle. Thus, each waveform includes at least one contraction period and at least one relaxation period and can be parameterized according to either the single or double contraction-relaxation model described above in relation to Figures 3 and 25. 【0368】 To help ensure that the synthetic training data represents the broadest possible population of contraction-relaxation cycles, multiple waveforms are preferably obtained from various artificial tissues across a range of different conditions. The artificial tissues include one or more manipulated muscle tissues, such as manipulated cardiac tissue and / or manipulated skeletal muscle tissue. Diversity is achieved by obtaining waveforms from artificial tissues across different cell lines, disease states, and treatments. Alternatively, the range of conditions within the multiple waveforms can be restricted, thereby allowing the synthetic data and subsequent parameter estimation models to be fine-tuned for specific applications. For example, the multiple waveforms may be restricted to a vehicle-processed waveform to generate a control parameter estimation model, or to a specific tissue type (e.g., manipulated cardiac tissue) to generate a tissue-specific parameter estimation model. Beneficially, this helps improve the performance of the parameter estimation model if it is known which class of waveforms the parameter estimation model will be used for. 【0369】 In step 3204, the extraction process, multiple parameter sets are extracted from multiple waveforms. One of these parameter sets characterizes the corresponding waveforms within the multiple waveforms. 【0370】 The parameter set includes parameters for either a single or dual contraction-relaxation cycle model (for example, as described above in relation to Figures 3 and 25). As described above, which model is used depends on the common contraction type of the multiple waveforms acquired in step 3202 acquired above. In one embodiment, the parameter set associated with a waveform includes at least one maximum parameter value (A), a shift parameter value (B), at least one contraction midpoint parameter value (t0), and at least one contraction growth rate parameter value (k c ), at least one relaxation midpoint parameter value (t d ), and at least one relaxation growth rate parameter value (k r ) includes. 【0371】 Multiple parameter sets are extracted using either supervised, unsupervised, or semi-supervised methods. According to supervised methods, the parameter sets are manually fitted to each waveform. For example, a first waveform is presented to the user, and the parameter values ​​within the parameter set are adjusted by the user until a second waveform, produced by contraction-relaxation cycle model fitting according to the parameter set, closely matches the waveform. The final parameter set resulting in the closely matching second waveform is then used as one of the parameter sets within the multiple parameter sets associated with the first waveform. According to unsupervised methods, the parameter sets are automatically fitted to the waveform (e.g., using the methods described in relation to Figures 9 and 10 above). In one embodiment, the unsupervised method utilizes a trained machine learning model to predict the parameter set values ​​of the waveform. According to semi-supervised methods, the automatically determined parameter sets obtained according to the unsupervised method are manually reviewed and refined by one or more users. 【0372】 In the determination step 3206, the parameter set distribution is determined from multiple parameter sets. 【0373】 The multiple parameter sets extracted in step 3204 of method 3200 include multiple values ​​for each parameter of the (single or dual) contraction-relaxation cycle model. For example, if a parameter set of 100 is extracted, 100 parameter values ​​are extracted for each parameter of the contraction-relaxation cycle model. In step 3206, the distribution or distribution of values ​​is determined for each parameter of the model. In one embodiment, the distribution is determined independently for each parameter. Alternatively, a multivariate distribution is determined for the multiple parameters that make up the parameter set. 【0374】 Any suitable method can be used to determine the parameter set distribution, such as histogram-based methods, density estimation methods, and clustering methods. In one embodiment, the parameter set distribution is determined using a kernel density estimation (KDE) method that estimates the parameter set distribution using a kernel and bandwidth parameters. In one embodiment, a standard (Gaussian) kernel is used with a bandwidth selected using either cross-validation or a bandwidth selection method such as Scott's rule or Silverman's rule. 【0375】 Once the parameter set distribution is determined, the parameter set can be obtained by sampling from this distribution (i.e., by sampling from each individual distribution or from the combined distribution). 【0376】 In step 3208, a synthetic training dataset is generated. Each element of the synthetic training dataset contains a synthetic waveform and a corresponding parameter set used to generate the synthetic waveform. The corresponding parameter set is obtained from a parameter set distribution. 【0377】 Synthetic training datasets are generated by repeatedly sampling the parameter sets of the contraction-relaxation cycle model from the parameter set distribution (as described above) and generating waveforms corresponding to each sampled parameter set. In this way, training datasets of any size (e.g., 1000, 10000, 100000 training data elements, etc.) can be efficiently generated. Since the parameter set distribution is modeled based on real-world data, the synthetic data will closely approximate actual waveform data. The synthetic waveform is either a single contraction-type waveform (as illustrated and described in relation to Figure 3) or a double contraction-type waveform (as illustrated and described in relation to Figure 25). 【0378】 Optionally, noise components are added to each waveform in the synthetic training dataset. The noise components are determined through a uniform distribution determined from multiple waveforms. 【0379】 In training step 3210, the predictive model is trained using a synthetic training dataset. The predictive model is trained to estimate an output parameter set from an input waveform. In one embodiment, the predictive model corresponds to the parameter estimation model (i.e., predictive model 600) described in relation to Figure 6 above. As described above, in one embodiment, training the parameter estimation model involves using a minibatch gradient descent method with a batch size of 128 and an ADAM solver. The ADAM solver has an initial learning rate of 1e-3 and terminates early based on the validation loss. Further details regarding the training of the parameter estimation model performed in training step 3210 are shown above in relation to the description in Figure 6. 【0380】 Figure 33 shows a method 3300 for extracting a single or dual contraction-relaxation cycle waveform according to one aspect of the present disclosure. 【0381】 Method 3300 includes the steps of acquiring a first waveform 3302, convolving the first waveform with a pulse train 3304, identifying a first position 3306, and extracting a second waveform from the first waveform at the first position 3308. Method 3300 further includes an optional step 3310 of outputting a second waveform. In one embodiment, Method 3300 is carried out by a signal processing unit 104-2 of a control unit 104 shown in Figure 1. 【0382】 Generally, Method 3300 extracts single or double contraction-relaxation cycle waveforms from a larger waveform containing multiple contraction-relaxation cycles. The larger waveform may be obtained from a hardware device such as a bioreactor (i.e., bioreactor 102 shown in Figure 1) and typically corresponds to the functional response of an artificial tissue under specific conditions. For example, the larger waveform may contain the contractile response of manipulated cardiac tissue stimulated at 1 Hz over a period of 30 seconds. In this example, the larger waveform would contain approximately 30 peaks or 30 contraction-relaxation cycles corresponding to the contractions of the manipulated cardiac tissue in response to the electrical stimulation. Method 3300 provides an efficient and precise mechanism for extracting each (single or double) contraction-relaxation cycle from the larger waveform, thereby allowing these sub-waveforms to be used for further processing and analysis (e.g., fitting models to these waveforms and extracting relevant features as described in relation to Figure 28 above). 【0383】 In acquisition step 3302, a first waveform is acquired. The first waveform includes multiple functional responses of the artificial tissue stimulated at a first frequency. 【0384】 In one embodiment, the first waveform is obtained from a bioreactor (e.g., bioreactor 102 shown in Figure 1) on which the artificial or manipulated tissue is grown / maintained. As previously mentioned, the artificial tissue includes manipulated muscle tissue such as manipulated cardiac tissue or manipulated skeletal muscle tissue. 【0385】 As described above in relation to Figure 1, electrical stimulation is applied to cell cultures during maturation, and once mature, the electrical stimulation is applied to artificial tissue to simulate the physiological environment specific to the artificial tissue, thereby allowing the functional response of the artificial tissue to this stimulation to be measured. The first waveform acquired in step 3302 includes the functional response of the artificial tissue to stimulation at a first frequency. In one embodiment, the first frequency to which the artificial tissue is stimulated, i.e., the pacing frequency, is 0.1 Hz to 20 Hz. In a further embodiment, the first frequency is 1 Hz to 6 Hz. 【0386】 Optionally, method 3300 includes a step (not shown) of stimulating artificial tissue at a first frequency prior to step 3302 of acquiring a first waveform. For example, an instruction (e.g., instruction 126 in Figure 1) or command is sent to a bioreactor containing artificial tissue (e.g., bioreactor 102 in Figure 1) to cause the bioreactor to stimulate the artificial tissue at a first frequency. 【0387】 In the convolution step 3304, the first waveform is convolved with a pulse train to generate a convolution waveform. The pulse train is generated at a first frequency. 【0388】 A pulse train corresponds to an idealized representation of the functional response of an artificial tissue at a first frequency. As is known, a pulse train, or pulse wave, is a waveform containing non-sinusoidal (rectangular) pulses or waves of duration T with frequency 1 / T0, where T0 is the duration of the pulse train. Thus, the duty cycle of the pulse train is T / T0. Therefore, a pulse train into which a first waveform is convolved contains a sequence of rectangular pulses having duration 1 / f and duration T, where f is the first frequency. 【0389】 Optionally, method 3300 further includes the step (not shown) of generating a pulse train at a first frequency before performing the convolution step 3304. 【0390】 Given a first waveform X1 and pulse train p, the convolution y=(X1*p) performed in convolution step 3304 is as follows: 【number】 【0391】 In the identification step 3306, a first position associated with the maximum value of the convolutional waveform is identified. The first position corresponds to the expected position of the first (single or double) contraction-relaxation cycle. 【0392】 The maximum value of the convolutional waveform corresponds to the position where the first waveform and the pulse train are best aligned. Therefore, the position of the maximum value of the convolutional waveform is used to identify the most likely position of a single contraction-relaxation cycle within the first waveform. The position of the maximum value of the convolutional waveform also provides a fixed point from which other contraction-relaxation cycles can be extracted from the first waveform. 【0393】 In extraction step 3308, a second waveform is extracted from a first position of the first waveform. The second waveform contains a first contraction-relaxation cycle and has a first duration proportional to the first frequency. The second waveform is either a single contraction type (illustrated in Figure 3) or a double contraction type (illustrated in Figure 25). 【0394】 The first position identified in identification step 3306 corresponds to the most likely position of one (single or double) contraction-relaxation cycle within the first waveform. Thus, the second waveform extracted from the first position contains this contraction-relaxation cycle. Since the first waveform corresponds to the functional response of the prosthetic tissue when stimulated at a predetermined, set frequency, the duration, or length, of the second waveform is proportional to this frequency. For example, if the prosthetic tissue is stimulated at 1 Hz, the duration of the first waveform is 1 second; if the prosthetic tissue is stimulated at 2 Hz, the duration of the first waveform is 0.5 seconds, and so on. 【0395】 Therefore, the second waveform corresponds to a window or subframe within the first waveform having a length corresponding to the first duration. In one embodiment, the second waveform is centered at a first position such that the midpoint of the second waveform is aligned, or substantially aligned, with a first position in the first waveform. 【0396】 In an optional output step 3310, a second waveform is output. In one embodiment, outputting the second waveform includes storing or saving the second waveform in persistent storage such as non-volatile memory, non-temporary media, or the like. Additionally or alternatively, outputting the second waveform includes transmitting the second waveform over a network (e.g., a local area network, a wide area network, and the like) or displaying the waveform for user review. 【0397】 In one embodiment, outputting a second waveform involves causing another process or method of the present disclosure to output a second waveform. For example, the second waveform may be output to method 2800 described above, such that acquiring step 2802 includes acquiring the second waveform from method 3300. 【0398】 Figure 34 shows a method 3400 for extracting further single or double contraction-relaxation cycles from a waveform according to an embodiment of the present disclosure. 【0399】 Method 3400 includes a step 3402 for identifying a second location, a step 3404 for extracting a third waveform from a first waveform at the second location, and further includes an optional step 3406 for outputting the third waveform. Method 3400 is performed after Method 3300. In particular, Method 3400 may be performed after the step 3306 for identifying the first location and may be performed in parallel with the step 3308 for extracting the second waveform. In one embodiment, Method 3400 is performed by the signal processing unit 104-2 of the control unit 104 shown in Figure 1. 【0400】 Method 3400 is used to extract further waveforms containing single or double contraction-relaxation cycles from a first waveform. Beneficially, the extraction performed in Method 3400 is efficient and highly parallelized because Method 3400 leverages prior information regarding the expected locations of single or double contraction-relaxation cycles within the first waveform, thereby enabling the independent extraction of contraction-relaxation cycles. 【0401】 In the identification step 3402, the second position is identified based on the first position and the first frequency. The second position corresponds to the expected position of the second contraction-relaxation cycle. 【0402】 The first position (identified in step 3306 of method 3300) corresponds to the best alignment between the first waveform and the pulse train. Thus, the first position can be understood as the most likely position of a contraction-relaxation cycle within the first waveform. Since the first waveform contains the functional response of the prosthesis at a predetermined frequency (i.e., the first frequency), other contraction-relaxation cycles linked to the functional response of the prosthesis are very likely to be located at positions separated from the first position. Thus, the first position can function as a fixed point within the first waveform from which other contraction-relaxation cycle waveforms can be extracted. 【0403】 The second position corresponds to the expected position of the second contraction-relaxation cycle (either single or double), and is separated from the first position by a distance proportional to the first frequency. Specifically, given a first position t1 in the first waveform, the second position t2 is given by t2 = t1 + (a × T0), where a is the step coefficient and T0 = 1 / f is the period of the pulse train. Thus, the next single or double contraction-relaxation cycle can be identified by setting the scaling coefficient a = 1, and the previous contraction-relaxation cycle can be identified by setting the scaling coefficient a = -1. 【0404】 In extraction step 3404, the third waveform is extracted from the second position of the first waveform. The third waveform includes a second (single or double) contraction-relaxation cycle and has a second duration proportional to the first frequency. 【0405】 The third waveform corresponds to the functional response of the artificial tissue (i.e., a single or double contraction-relaxation cycle) when stimulated at a predetermined, set frequency. Therefore, the duration, or length, of the third waveform is proportional to this frequency. For example, if the artificial tissue is stimulated at 1 Hz, the second duration is 1 second, and if the artificial tissue is stimulated at 2 Hz, the second duration is 0.5 seconds, and so on. In one embodiment, the first duration and the second duration are the same. 【0406】 In one embodiment, the third waveform is centered at a second position such that the midpoint of the third waveform is aligned, or substantially aligned, with the second position of the first waveform. 【0407】 In an optional step 3406 of outputting, a third waveform is output. In one embodiment, outputting the third waveform includes storing or saving the third waveform in persistent storage such as non-volatile memory, non-temporary media, or the like. Additionally or alternatively, outputting the third waveform includes transmitting the third waveform over a network (e.g., a local area network, a wide area network, and the like) or displaying the waveform for user review. 【0408】 In one embodiment, outputting a third waveform involves causing another process or method of the present disclosure to output a third waveform. For example, the third waveform may be output to method 2900 described above, such that acquiring step 2802 includes acquiring the third waveform from method 3400. 【0409】 As described above, method 3200, which extracts further single or double contraction-relaxation cycle waveforms from a first waveform, can be repeated for all contraction-relaxation cycles in the first waveform. Since the extraction performed by method 3400 depends only on the first position and first frequency, further signal processing or analysis is not required to identify the positions of further contraction-relaxation cycles. Thus, method 3400 provides a rapid and efficient method for extracting single or double contraction-relaxation cycles from a waveform. These waveforms can then be further processed, for example, by fitting a model to the waveform to generate a noise-filtered representation of the waveform. 【0410】 Figure 35 shows a method 3500 for predicting perturbation effects according to an aspect of this disclosure. 【0411】 Method 3500 includes the steps of: acquiring multiple signals 3502; splitting the multiple signals into a first set of multiple waveforms 3504; determining the predicted contraction type 3506; fitting a model to each of the first set of multiple waveforms 3508; generating a second set of multiple waveforms from the model 3510; extracting a first set of feature values ​​3512; extracting a second set of feature values ​​3512; and determining the effect 3516. In one embodiment, Method 3500 is carried out by a control unit 104 or a subunit thereof shown in Figure 1. 【0412】 Generally, Method 3500 describes the application of single and dual contraction-relaxation cycle models to downstream drug discovery / development tasks. Specifically, contraction-relaxation cycle models are used to efficiently generate accurate feature values ​​from baseline and perturbation signals of manipulated tissue. Accurately extracting features from these signals allows for the efficient and accurate identification of effects associated with perturbations. 【0413】 In acquisition step 3502, multiple signals are acquired. These multiple signals include baseline signals and perturbation signals. Baseline signals include a first set of functional responses of the manipulated tissue under reference or baseline conditions. Perturbation signals include a second set of functional responses of the manipulated tissue under a first set of perturbation conditions. 【0414】 Baseline and perturbation signals include multiple functional responses (i.e., multiple contraction-relaxation cycles, or peaks) of the manipulated tissue under a first set of reference and perturbation conditions. Generally, reference conditions refer to conditions that provide a baseline comparison to perturbation conditions. In embodiments, reference conditions are conditions associated with a control setting or environment. Reference conditions may correspond to manipulated or artificial tissue in a default, natural, or unaltered state (i.e., without drug or agent administration). Alternatively, reference conditions may correspond to vehicle-treated manipulated tissue. Perturbation conditions refer to conditions under which the manipulated tissue is perturbed in some way. Examples of perturbation conditions include administration of a drug or compound (i.e., a perturbant), a disease state, a different cell line, a physical perturbation applied to the manipulated tissue, or a change in the environment of the manipulated tissue. Thus, perturbation conditions may alternatively be referred to as treatment conditions. In the case of perturbations involving drugs or compounds, the conditions are further associated with effects related to the drug or compound, such as mechanism of action or toxicity. Considering the range of different perturbation conditions, the manipulated tissue (e.g., manipulated tissue in a disease state treated with a specific compound) may be associated with more than one perturbation condition. 【0415】 In one embodiment, baseline and perturbation signals are acquired from a bioreactor (e.g., bioreactor 102 in Figure 1) on which the engineered or artificial tissue is grown / maintained. As previously stated, the artificial tissue includes engineered muscle tissue such as engineered cardiac tissue or engineered skeletal muscle tissue. The baseline and perturbation signals are acquired at two different time points. For example, the baseline signal is acquired at a first time point, when the engineered tissue is then perturbed according to a first perturbation (e.g., a first dose of the compound is applied to the engineered tissue), and the perturbation signal is acquired at a second time point after the first time point. The baseline and perturbation signals include the functional response of the engineered tissue when stimulated at a predetermined pacing frequency (e.g., 0.1 Hz, 0.5 Hz, 1 Hz, 2 Hz, etc.). Alternatively, the baseline and perturbation signals include the spontaneous functional response of the engineered tissue in the absence of external stimulation. 【0416】 In the splitting step 3504, the multiple signals are split into a first plurality of waveforms. Each of the first plurality of waveforms includes at least one contraction response and at least one relaxation response of the manipulated tissue during a single contraction-relaxation cycle. 【0417】 Multiple signals are divided into a first set of waveforms using the method 3300 for extracting the contraction-relaxation cycle waveforms described above. Alternatively, multiple signals are divided by manually annotating or extracting regions within the first set of waveforms that correspond to individual contraction-relaxation cycles. 【0418】 The first set of waveforms generated in the splitting step 3504 includes a first subset of waveforms associated with the waveform extracted from the baseline signal and a second subset of waveforms associated with the waveform extracted from the perturbation signal. The subsequent model fitting (described later) is independent of the waveform source (i.e., independent of whether the waveform is the reference for the perturbation waveform), but the identification of whether the waveform corresponds to a reference condition or a perturbation condition is used in the subsequent feature extraction step. 【0419】 In the determination step 3506, for each of the first multiple waveforms, the predicted contraction type from among multiple contraction types is determined. The multiple contraction types include single contraction types and double contraction types. 【0420】 The predicted bloat type is determined by the classifier based on the waveform input to the classifier. As described in more detail above in relation to Figures 6 and 7A-7D, the classifier includes a trained machine learning model, such as a trained neural network. In one embodiment, the classifier includes a convolutional neural network, an expanded convolutional neural network, and a long short-term memory network. Further architectural details of the classifier are shown above in relation to Figures 6 and 7A-7D. 【0421】 In fitting step 3508, the model is fitted to each of the first set of waveforms based on the corresponding predicted contraction type. For example, if the waveform is of a single contraction type, a single contraction type model is fitted in fitting step 3508. Alternatively, if the waveform is of a dual contraction type, a dual contraction type model is fitted in fitting step 3508. The model independently parameterizes the growth of at least one contraction response and at least one relaxation response in the manipulated tissue. 【0422】 Step 3508, which fits the model to each waveform, corresponds to step 2806 of fitting, which is described in more detail above in relation to Figure 28. Thus, in fitting step 3508, fitting step 2806 is repeated for each of the first set of waveforms. The process of fitting the model to the waveforms is described in more detail above in relation to Figures 28 and 29. 【0423】 In the generation step 3510, the second set of waveforms is generated from a model fitted to each of the first set of waveforms. The second set of waveforms includes a set of filtered baseline waveforms associated with the baseline signal and a set of filtered perturbation waveforms associated with the perturbation signal. 【0424】 The generation step 3510 corresponds to repeatedly applying the generation step 908 (described in more detail above in relation to Figure 28) to the model fitted to each waveform in the first set of waveforms. The process of generating the second waveform from the model fitted to the first waveform is described in more detail above in relation to Figures 28 and 29. 【0425】 In extraction step 3512, the first feature value of the first feature is extracted from multiple filtered baseline waveforms. 【0426】 Multiple filtered baseline waveforms contain noise-filtered or noise-suppressed representations of the contraction-relaxation cycles in the baseline signal. Because the signal is filtered to remove noise, features can be accurately extracted from these waveforms. 【0427】 Step 3512 for extracting a first feature value includes extracting multiple feature values ​​from multiple filtered baseline waveforms such that the first feature value includes multiple feature values ​​or representations of multiple feature values. Thus, the value of the first feature is extracted from each of the multiple filtered baseline waveforms in order to determine the first feature value. In one embodiment, the first feature value includes the mean (mean, median, etc.) of the first feature determined from the multiple filtered baseline waveforms. In a further embodiment, the first feature value includes the maximum, minimum, or distribution of values ​​determined from the multiple filtered baseline waveforms. 【0428】 The first characteristic is one of the following: single contraction, or peak, amplitude (i.e., peak amplitude shown in Figure 2, 206), contraction time (i.e., time to peak amplitude shown in Figure 2, 208), maximum contraction slope (i.e., maximum development rate shown in Figure 2, 214), relaxation time (i.e., time to peak decline shown in Figure 2, 210), maximum relaxation slope (i.e., maximum decline rate shown in Figure 2, 216), or single contraction duration (i.e., duration shown in Figure 2, 212). 【0429】 In one embodiment, the first feature is the total number of single-collapse type waveforms or the total number of dual-collapse type waveforms (as described above in relation to Figure 27). Thus, the first feature value corresponds to the number or distribution of single and / or dual-collapse type waveforms within a plurality of filtered baseline waveforms. 【0430】 In step 3514, the second feature value of the first feature is extracted from multiple filtered perturbation waveforms. 【0431】 Step 3514 for extracting a second feature value includes extracting multiple feature values ​​from multiple filtered perturbation waveforms such that the second feature value includes multiple feature values ​​or representations of multiple feature values. Thus, the values ​​of the first feature are extracted from each of the multiple filtered perturbation waveforms in order to determine the second feature value. In one embodiment, the second feature value includes the mean (mean, median, etc.) of the first feature determined from the multiple filtered perturbation waveforms. In a further embodiment, the second feature value includes the maximum, minimum, or distribution of values ​​determined from the multiple filtered perturbation waveforms. 【0432】 As described above, in one embodiment, the first feature is the total number of single-contraction type waveforms or the total number of double-contraction type waveforms, such that the second feature value corresponds to the number or distribution of single and / or double-contraction type waveforms within a plurality of filtered perturbation baseline waveforms. 【0433】 In the determination step 3516, the effect associated with the first perturbation is determined based on a comparison between the first and second feature values. 【0434】 The first feature value is a quantitative descriptor of the functional response of the manipulated tissue under reference conditions. The second feature value is a quantitative descriptor of the functional response of the manipulated tissue under perturbed conditions accompanied by the first perturbation. Therefore, by comparing the first and second feature values, any changes, i.e., effects, resulting from the first perturbation in the functional response of the manipulated tissue are revealed. An example of such a comparison is shown above in relation to Figure 27. 【0435】 As a further example, the first perturbation may correspond to the application of a compound with an unknown physiological effect. In this example, a comparison of a first feature value corresponding to the peak amplitude of the contractile force waveform of the manipulated tissue under reference conditions with a second feature value corresponding to the peak amplitude of the contractile force waveform of the manipulated tissue under perturbed conditions with the application of the compound reveals an increase in the mean peak amplitude. Therefore, it can be inferred that the compound has an effect associated with increasing the contractile force of the manipulated tissue during the contraction-relaxation cycle. Beneficially, since the feature values ​​are determined from noise-filtered waveforms, effects resulting from differences between feature values ​​can be more accurately identified, leading to improved treatment and potentially improved patient outcomes. 【0436】 Systems and methods for tracking organizational scaffolding. Figure 36 shows well 3602 of a bioreactor, such as the bioreactor 102 shown in Figure 1, according to an embodiment of the present disclosure. 【0437】 Well 3602 contains artificial tissue 3604 (or manipulated tissue) attached to the first flexible scaffold 3606 and the second flexible scaffold 3608. Figure 36 further shows an imaging device 3610 configured to acquire one or more images of the artificial tissue 3604, the first flexible scaffold 3606, and / or the second flexible scaffold 3608. In one embodiment, well 3602, artificial tissue 3604, the first flexible scaffold 3606, the second flexible scaffold 3608, and the imaging device 3610 correspond to well 114, manipulated tissue 124, the first scaffold 120-1, the second scaffold 120-2, and the imaging device or optical sensor of the sensor assembly 108 shown in Figure 1, respectively. 【0438】 As described above in relation to Figure 1, the artificial tissue 3604 (or manipulated tissue) grows in well 3602 from cells seeded in well 3602. During maturation, the artificial tissue 3604 attaches to a first flexible scaffold 3606 and / or a second flexible scaffold 3608. Both the first flexible scaffold 3606 and the second flexible scaffold 3608 are arranged across well 3602, thereby allowing the artificial tissue 3604 to attach to them. The first flexible scaffold 3606 and the second flexible scaffold 3608 comprise flexible elements that are arranged to bend or deform in response to forces applied thereon. For example, the first flexible scaffold 3606 is arranged to deform in a first direction in response to a contractile force F1 applied to the first flexible scaffold 3606 by the artificial tissue 3604. Similarly, the second flexible scaffold 3608 is positioned to deform in a second direction in response to a contractile force F2 exerted on the second flexible scaffold 3608 by the artificial tissue 3604. In further examples, the first flexible scaffold 3606 and the second flexible scaffold 3608 are positioned to flex in response to a predetermined force exerted on them by a probe or other instrument. In one embodiment, the first flexible scaffold 3606 and the second flexible scaffold 3608 are formed from a flexible polymer wire such as poly(octamethylene memaleate (anhydrous) citrate) (POMaC) wire. In such embodiments, the flexible scaffold may be referred to as a wire or flexible wire. 【0439】 The imaging device 3610 is configured to detect deflection or deformation of the first flexible scaffold 3606 and / or the second flexible scaffold 3608 (e.g., resulting from a contractile force exerted on the flexible scaffold by the artificial tissue 3604 or probe). The imaging device 3610 is configured to acquire multiple image-based representations of one or more deflections of the first flexible scaffold 3606 and / or the second flexible scaffold 3608 over a time frame or period. For example, the imaging device 3610 may be configured to capture an image or frame of tissue or a region of tissue attached to the tissue scaffold every n seconds, where n is associated with a predetermined frequency (rate) at which images or frames are captured. For example, if n=1, one image is captured per second. Any preferred value for n, e.g., n = {1 / 60, 1 / 50, 1 / 30, 1 / 24, 1 / 12, 1 / 4, 1 / 2, 1, 2}, and similar values ​​can be selected. In one embodiment, the frame frequency is determined according to the frequency at which the tissue within the device is stimulated. Thus, an image or sequence of frames of tissue over a time frame captures one or more deflections of the first flexible scaffold 3606 and / or the second flexible scaffold 3608 over a time frame. Images(s) captured by the imaging device 3610 can be output from the device or bioreactor to a control unit or processing unit (for example, images(s) captured by the sensor assembly 108 can be output to the control unit 104 shown in Figure 1). 【0440】 Figure 37 shows exemplary images of tissue scaffolds under different contractile forces according to embodiments of the present disclosure. 【0441】 Figure 37 shows the first image 3702 and the second image 3704 acquired from the bioreactor's imaging device (e.g., imaging device 210 shown in Figure 36). The first image 3702 captures the first deflection of the tissue scaffold 3706 at a first time point t1. The second image 3704 captures the second deflection of the tissue scaffold 3708 at a second time point t2. To understand the difference in deflection, the deflection is shown relative to a common baseline (vertical dashed line in both images). The change in deflection between t1 and t2 may be used to encode or otherwise characterize the contractile force acting on the tissue scaffold (e.g., due to tissue attached to the tissue scaffold, or due to a probe or other instrument). 【0442】 This disclosure relates to flexible scaffold tracking, which enables the accurate and efficient extraction and measurement of deflection of tissue scaffolds, thereby enabling the generation of models characterizing the contractile force(s) that produce the deflection. Beneficially, such models can encode the contractile response of the manipulated tissue, thereby providing quantification of the functional response of the manipulated tissue. Alternatively, such models can be used to encode or calibrate the relationship between contractile force and measured displacement, thereby improving the accuracy of contractile force measurements obtained from such models. Improvements to such models provide improvements to downstream tasks that utilize such models, while also improving the efficiency and performance of the computing systems used to generate and deploy such models. 【0443】 Generally, flexible scaffolding tracking can begin by acquiring an image taken at a first time point from the flexible scaffolding device or bioreactor (e.g., from bioreactor 102 in Figure 1). The image captures the deflection of the flexible scaffolding along a first dimension, resulting from the (contraction) forces acting on it at the first time point. A curve is fitted to the image so that the curve extends along or approximately along the centerline of the flexible scaffolding in the image. Displacement values ​​are determined between the curve and a baseline extending along a second dimension perpendicular to the first dimension. A model is generated based on the displacement values, thereby characterizing the (contraction) forces acting on the flexible scaffolding at the first time point. This method is illustrated in Figure 38 and described in further detail below. 【0444】 Figure 38 shows a method for tracking a flexible scaffold according to an aspect of this disclosure. 【0445】 Figure 38 shows an image region 3802 of an image containing the flexible scaffolding 3804 taken at a first time point. Image region 3802 captures the deflection of the flexible scaffolding 3804 along a first dimension 3806, resulting from the contraction force acting on the flexible scaffolding 3804 at the first time point. Figure 38 also shows a second dimension 3808 perpendicular to the first dimension 3806. In one embodiment, the first dimension 3806 corresponds to the vertical axis of image region 3802, and the second dimension 3808 corresponds to the horizontal axis of image region 3802. A curve 3810 extending along the centerline of the flexible scaffolding 3804 is shown. Curve 3810 intersects the first edge 3812 of image region 3802 at the first intersection 3814 and the second edge 3816 of image region 3802 at the second intersection 3818. Reference line 3820 extends along the second dimension 3808 between the first intersection 3814 and the second intersection 3818. The measurement 3822 taken between curve 3810 and reference line 3820 corresponds to the displacement value of the flexible scaffold 3804 at the first time point. The measurement 3822 is determined at reference point 3824 along the second dimension 3808. The reference point 3824 is determined such that the orthogonal distance between the curve 3810 and the reference line 3820 along the first dimension 3806 is maximized at the reference point 3824 (i.e., the distance between the first point 3826 on the curve 3810 at reference point 3824 along the second dimension 3808 and the second point 3828 on the reference line 3820 at reference point 3824 along the second dimension 3808 corresponds to the maximum orthogonal distance between the curve 3810 and the reference line 3820 along the first dimension 3806). Figure 38 further shows a predetermined expected range 3830 in which the maximum orthogonal distance between the curve 3810 and the reference line 3820 is expected. The predetermined expected range 3830 includes a portion of the length of the image region 3802 along the second dimension 3808 (e.g., 60% of the height of the image region 3802). As described in more detail below, the displacement values ​​determined from measurement 3822 are used to generate a model characterizing the contractile force acting on the flexible scaffold 3804 (for example, due to contraction of manipulated tissue attached to the flexible scaffold 3804, or due to a predetermined force acting on the flexible scaffold 3804 by a probe or other instrument). 【0446】 Image region 3802 includes an image acquired from the device or bioreactor, or a portion of an image (e.g., acquired from an imaging device of the bioreactor, such as imaging device 210 shown in Figure 36). Image region 3802 includes a grayscale bright-field image or a fluorescence image. As clearly shown in Figure 36, image region 3802 captures one of the tissue scaffolds of the device or bioreactor. Image region 3802 may be captured such that only the portion of the device containing the relevant tissue scaffold is captured within image region 3802. Alternatively, image region 3802 may be cropped from a larger image of the entire device / well or tissue. In such a situation, image region 3802 may be cropped manually or automatically (e.g., by cropping the larger image into a predetermined bounding box having coordinates known to correspond to the portion of the larger image containing the tissue scaffold). 【0447】 As described above, curve 3810 extends along the centerline, or approximate centerline, of the flexible scaffold 3804 captured within the image region 3802. To determine the centerline of the flexible scaffold 3804 within the image region 3802, one or more image processing operations are used to enhance the structure and / or appearance of the flexible scaffold 3804 within the image region 3802. As can be seen from the exemplary image in Figure 37, the flexible scaffold (e.g., tissue scaffold 3706 or tissue scaffold 3708) appears as a vascular-like structure in the image. Therefore, image processing and / or filtering operations may be used to enhance the region of the image containing the vascular-like structure. 【0448】 In one embodiment, the image region 3802 is transformed using a filtering operation to generate a transformed image region. The filtering operation determines whether a region or portion of the image region 3802 (i.e., a pixel or a neighborhood of a pixel) contains a flexible scaffold 3804. Thus, the transformed image region contains a transformed representation of the image region 3802, and each pixel value within the transformed image region corresponds to the possibility that the pixel value lies on the flexible scaffold. The filtering operation includes a vascular enhancement filter, such as a multiscale vascular enhancement filter (e.g., a Hessian-based flange vascular filter) or a diffusion filter (e.g., a coherence-enhanced diffusion filter). In an embodiment, the filtering operation includes convolving a one-dimensional kernel onto one dimension of the image region 3802, as illustrated in Figure 39. 【0449】 Figure 39 shows a one-dimensional vascular enhancement filter according to an embodiment of the present disclosure. 【0450】 Figure 39 shows a plot 3902 of pixel intensity values ​​taken between points A and A' along a cross-section of the image region 3802 shown in Figure 38. Figure 39 further shows a one-dimensional kernel 3904 of the response values, a convolution operation 3906, and a plot 3908. The one-dimensional kernel 3904 has a shape corresponding to the approximate cross-sectional shape of the flexible scaffold. The shape of the one-dimensional kernel 3904 is square or approximately square. Convolving the one-dimensional kernel 3904 onto the pixel intensity values ​​shown in plot 3902 via the convolution operation 3906 yields the response values ​​shown in plot 3908. The response values ​​shown in plot 3908 correspond to the pixel intensity values ​​of the row of pixels from A to A' in the transformed representation of the image region. 【0451】 As can be seen from the reaction value plot 3908, the pixel position corresponding to the approximate center point of the flexible scaffold contains the maximum reaction value as a result of the convolution operation 3906. Therefore, repeating the one-dimensional convolution for all rows of the image region 3802 shown in Figure 38 yields a transformed image with the maximum pixel intensity approximately along the centerline of the flexible tissue scaffold. Thus, the pixel in each row of the transformed image with the maximum intensity value (within the row) is selected as the most likely estimate of the pixel along the centerline of the tissue scaffold. Repeating this for each row in the transformed image produces multiple data points corresponding to the most likely positions along the centerline of the flexible scaffold 3804 in the image region 3802. As shown in Figure 40 and described later, a curve can be fitted to the multiple data points to determine the approximate centerline of the flexible scaffold 3804. 【0452】 Figure 40 shows a plot 4002 of data points obtained from the transformed image of the tissue scaffold according to an embodiment of the present disclosure. 【0453】 Plot 4002 includes multiple data points (represented as black dots in plot 4002) corresponding to pixel locations having the maximum local intensity (i.e., the maximum intensity within a row of pixels). Plot 4002 further includes a curve 4004 that fits the multiple data points, a first neighborhood 4006 around a first outlier point 4008, and several further neighborhoods 4010-4014. 【0454】 Curve 4004 corresponds to a quadratic curve fitted to multiple data points (i.e., the locations of the largest pixel values ​​in the transformed image along the first dimension, as described above in relation to Figure 39). In the embodiment, curve 4004 is fitted to the multiple data points using a least-squares method such that the sum of squared residuals, i.e., the deviation between the multiple data points and the corresponding set of data points on the curve, is minimized. Once fitted, curve 4004 corresponds to the approximate centerline of the flexible scaffold. 【0455】 In one embodiment, one or more outlier points are removed from the plurality of data points before the curve 4004 is fitted. Advantageously, this helps improve the accuracy of the curve fitting process, and thus results in a centerline that more closely corresponds to the centerline of the tissue scaffold. A data point of the plurality of data points is considered an outlier point if the outlier point meets or exceeds a predetermined outlier threshold value. 【0456】 In one embodiment, the predetermined outlier threshold value includes a neighborhood distance threshold value. In the example shown in plot 4002, since the distance between the first outlier point 4008 and each of the plurality of data points exceeds a predetermined distance threshold value η, the first outlier point 4008 is determined to be an outlier point. That is, the other data points from the plurality of data points do not exist within the first neighborhood 4006 around the first outlier point 4008. The predetermined distance threshold value η is relative to the expected width of the flexible scaffold. For example, if the flexible scaffold has an expected width of w pixels or μm (for example, 50 μm, 75 μm, 100 μm, etc.), the distance threshold value is set as η = αw, where α is a parameter that can take a value between 1 and 2. In one embodiment, α = 1.5. Additionally, or alternatively, a data point is considered an outlier point if less than a predetermined number of other data points are within the neighborhood of that data point. In the example shown in plot 4002, the data points within each of the plurality of further neighborhoods 4010 - 4014 are visually distinguishable as outlier points, but nevertheless the neighborhoods overlap. For example, the data points associated with neighborhoods 4012 and 4014 are within neighborhood 4010. Therefore, if less than m other data points are within the neighborhood (defined by the distance threshold value η) around a data point, the data point can be considered an outlier point. Here, m is a predetermined value such that 0 < m < 10, or 1 < m < 15. In one embodiment, m = 2. 【0457】 Additionally, or alternatively, a predetermined outlier threshold includes a pixel intensity threshold such that the pixel value at the image position corresponding to the outlier point exceeds a predetermined intensity threshold. In one embodiment, the predetermined intensity threshold is three times the median absolute deviation (MAD) of all pixel intensities for pixels present along the tissue scaffold. 【0458】 Referring again to Figure 38, the curve 3810 fitted to the flexible scaffolding 3804 as described above is used to determine the measurement 3822 corresponding to the displacement value of the flexible scaffolding 3804 at the first time point. 【0459】 The displacement value encodes the deflection of the flexible scaffolding 3804 at a first time point relative to the expected position of the flexible scaffolding 3804 in a stationary state. That is, the displacement or deflection of the flexible scaffolding 3804 at a given time point corresponds to the difference between the position of the flexible scaffolding 3804 at a given time point and the expected position of the flexible scaffolding 3804 in a stationary state. The expected position of the flexible scaffolding 3804 in a stationary state is modeled by a baseline 3820, and the distance between the curve 3810 (extending along the centerline of the flexible scaffolding 3804) and the baseline 3820 is used to determine the displacement value of the flexible scaffolding 3804 at a first time point. 【0460】 The baseline 3820 (alternatively referred to as the stationary line, baseline, or expected stationary position) is determined using the intersection points of the curve 3810 and the boundary or edge of the image region 3802, thereby extending the baseline 3820 between the intersection points. In Figure 38, the first intersection point 3814 is determined by identifying the intersection point between the curve 3810 and the first edge 3812 of the image region 3802. The second intersection point 3818 is determined by identifying the intersection point between the curve 3810 and the second edge 3816 of the image region 3802, where the second edge 3816 of the image region 3802 is on the opposite side of the first edge 3812 of the image region 3802. The baseline 3820 is then fitted so that it extends between the first intersection point 3814 and the second intersection point 3818. In one embodiment, the reference line 3820 defines a second dimension 3808 (i.e., defines the angle / direction of the second dimension 3808 with respect to the axis or multiple axes of the image region 3802). In some cases, the reference line 3820 may be aligned with or substantially aligned with an image axis (e.g., the vertical axis), thereby causing the second dimension 3808 to be aligned with or substantially aligned with an image axis. In some other cases, the reference line 3820 may not be aligned with any image axis, thereby causing the second dimension 3808 to be substantially unaligned with any image axis. 【0461】 The measurement 3822 used to determine the displacement value of the flexible scaffolding 3804 at the first time point is then calculated between the curve 3810 and the reference line 3820. The measurement 3822 is determined at the reference point 3824 along the second dimension 3808. As described above, the reference point 3824 is determined such that the orthogonal distance along the first dimension 3806 (perpendicular to the second dimension 3808) between the curve 3810 and the reference line 3820 is maximized at the reference point 3824 (i.e., the distance between the first point 3826 on the curve 3810 at the reference point 3824 along the second dimension 3808 and the second point 3828 on the reference line 3820 at the reference point 3824 along the second dimension 3808 corresponds to the maximum orthogonal distance between the curve 3810 and the reference line 3820 along the first dimension 3806). Alternatively, the displacement value is determined using a measurement corresponding to the estimated area of ​​the shape bounded by curve 3810 and baseline 3820, i.e., the shape bounded by the path from the first intersection 3814 to the second intersection 3818 along curve 3810 and closed by the path from the second intersection 3818 to the first intersection 3814 along baseline 3820. 【0462】 In one embodiment, the reference point determined using the process described above is replaced by a predetermined reference point if the reference point lies outside a predetermined expected range 3830. As previously stated, the predetermined expected range 3830 corresponds to the portion of the image region 3802 in which the reference point 3824 is expected to reside, i.e., the portion of the image...

Claims

[Claim 1] A method for predicting therapeutic effect, wherein the method is carried out in a device comprising one or more processors and memory, and the method is Acquiring a plurality of signals, including a baseline signal and a perturbation signal, by one or more processors, wherein the baseline signal includes a first plurality of functional responses of the manipulated tissue under reference conditions, and the perturbation signal includes a second plurality of functional responses of the manipulated tissue under perturbation conditions with a first perturbation. The division of the plurality of signals into a first plurality of waveforms by one or more processors, wherein each of the first plurality of waveforms includes the contraction and relaxation responses of the manipulated tissue during a single contraction-relaxation cycle. Fitting a model to each of the first plurality of waveforms using one or more processors, wherein the model independently parameterizes the growth of the contraction and relaxation responses of the manipulated tissue during the single contraction-relaxation cycle of each waveform. The process involves generating a second plurality of waveforms from the model fitted to each of the first plurality of waveforms using one or more processors, wherein the second plurality of waveforms includes a plurality of filtered baseline waveforms associated with the baseline signal and a plurality of filtered perturbation waveforms associated with the perturbation signal. The one or more processors extract the first feature value of the first feature from the plurality of filtered baseline waveforms, The one or more processors extract the second feature value of the first feature from the plurality of filtered perturbation waveforms, A method comprising determining, by one or more processors, an effect associated with the first perturbation based on a comparison of the first feature value and the second feature value. [Claim 2] The method according to claim 1, wherein the first feature is one of the following: single contraction amplitude, contraction time, maximum contraction slope, relaxation time, maximum relaxation slope, and single contraction duration. [Claim 3] The first feature value includes the average value of the first feature determined from the plurality of filtered baseline waveforms. The method according to claim 1, wherein the second feature value includes the average value of the first feature determined from the plurality of filtered perturbation waveforms. [Claim 4] The aforementioned contraction reaction is modeled by an ascending logistic function with a positive growth rate. The method according to claim 1, wherein the relaxation response is modeled by a falling logistic function having a negative growth rate. [Claim 5] The method according to claim 4, wherein the model includes the product of the ascending logistic function and the descending logistic function. [Claim 6] The method according to claim 1, wherein the model includes a plurality of parameters associated with the contraction response and the relaxation response. [Claim 7] The method according to claim 6, wherein the plurality of parameters include a maximum value parameter, an increase rate parameter, a decrease rate parameter, a y-shift parameter, an increase x-shift parameter, and a decrease x-shift parameter. [Claim 8] The step of fitting the model to each of the first plurality of waveforms includes, for the first waveform among the first plurality of waveforms, The method according to claim 6, comprising predicting the plurality of values ​​for the plurality of parameters of the model by one or more processors such that the model fitted to the first waveform includes a plurality of values ​​for the plurality of parameters. [Claim 9] The method according to claim 8, wherein the plurality of values ​​are predicted using a machine learning model comprising an inception model, a trained convolutional neural network having an expanded convolution, and a trained long short-term memory neural network. [Claim 10] The step of fitting the model to each of the first plurality of waveforms is performed with respect to the first waveform among the first plurality of waveforms. The method of claim 8, further comprising optimizing the plurality of values ​​by one or more processors by minimizing the error between the first waveform and the model fitted to the first waveform using the plurality of values ​​for the plurality of parameters of the model. [Claim 11] The method according to claim 10, wherein the plurality of values ​​are optimized using a simplex search algorithm. [Claim 12] The method according to claim 1, wherein the plurality of signals are obtained from a bioreactor containing the manipulated tissue. [Claim 13] The method according to claim 1, wherein the manipulated tissue is one of manipulated cardiac tissue and manipulated skeletal muscle tissue. [Claim 14] A computer-readable medium storing instructions, wherein when the instructions are executed by one or more processors of the device, the one or more processors of the device Acquiring a plurality of signals, including a baseline signal and a perturbation signal, wherein the baseline signal includes a first plurality of functional responses of the manipulated tissue under reference conditions, and the perturbation signal includes a second plurality of functional responses of the manipulated tissue under perturbation conditions with a first perturbation. Dividing the aforementioned plurality of signals into a first plurality of waveforms, wherein the first plurality of waveforms Each waveform includes the contraction and relaxation responses of the manipulated tissue during a single contraction-relaxation cycle, and is divided into: Fitting a model to each of the first plurality of waveforms, wherein the model independently parameterizes the growth of the contraction and relaxation responses of the manipulated tissue during the single contraction-relaxation cycle of each waveform, To generate a second plurality of waveforms from the model fitted to each of the first plurality of waveforms, wherein the second plurality of waveforms includes a plurality of filtered baseline waveforms associated with the baseline signal and a plurality of filtered perturbation waveforms associated with the perturbation signal. Extracting the first feature value of the first feature from the plurality of filtered baseline waveforms, From the plurality of filtered perturbation waveforms, the second feature value of the first feature is extracted, A computer-readable medium that determines the effect associated with the first perturbation based on a comparison of the first feature value and the second feature value. [Claim 15] It is a system, One or more processors, It includes a memory for storing instructions, When the instruction is executed by one or more processors, the one or more processors will be instructed to: Acquiring a plurality of signals, including a baseline signal and a perturbation signal, wherein the baseline signal includes a first plurality of functional responses of the manipulated tissue under reference conditions, and the perturbation signal includes a second plurality of functional responses of the manipulated tissue under perturbation conditions with a first perturbation. Dividing the plurality of signals into a first plurality of waveforms, wherein each of the first plurality of waveforms includes the contraction and relaxation responses of the manipulated tissue during a single contraction-relaxation cycle, Fitting a model to each of the first plurality of waveforms, wherein the model independently parameterizes the growth of the contraction and relaxation responses of the manipulated tissue during the single contraction-relaxation cycle of each waveform, To generate a second plurality of waveforms from the model fitted to each of the first plurality of waveforms, wherein the second plurality of waveforms includes a plurality of filtered baseline waveforms associated with the baseline signal and a plurality of filtered perturbation waveforms associated with the perturbation signal. Extracting the first feature value of the first feature from the plurality of filtered baseline waveforms, From the plurality of filtered perturbation waveforms, the second feature value of the first feature is extracted, A system that determines the effect associated with the first perturbation based on a comparison of the first feature value and the second feature value.

Images