Diagnosis engine for diagnosing preterm birth risk

Abstract

A diagnosis system for predicting a preterm birth risk includes an artificial intelligence engine comprising a plurality of first classifiers, each having as first input of complex impedances measured at a plurality of different frequencies from cervical tissues during pregnancy and optionally as a second input at least one of preterm history, pregnancy stage, age, pH level, etc., and a Meta classifier configured to output an indication of the preterm birth risk, with as a third input being the outputs of each of the first classifiers, wherein the first and the second inputs are pre-processed using a plurality of different pre-processing methods, each the same or different for each classifier, and wherein the third inputs are post-processed using a plurality of different post-processing methods, each the same or different for the outputs of each first classifiers.

Description

DIAGNOSIS ENGINE FOR DIAGNOSING PRETERM BIRTH RISKBACKGROUNDTechnical Field

[0001] This disclosure relates to methods of training a diagnosis engine for diagnosing preterm birth risk.Description of the Related Art

[0002] Preterm birth occurs when an expectant mother delivers her baby prematurely. It is a significant problem and can cause severe health problems to infants. It is the leading cause of infant mortality up to age 5, and for those who survive it can cause severe life-long health problems for the baby. Accurate and early diagnosis and treatment of preterm birth can aid in delaying the delivery of the child by applying effective treatment to the mothers. It has been suggested that electrical impedance spectroscopy (EIS) techniques might be used for detecting preterm birth. Details for the EIS techniques can be explained with the reference U.S. patent application Ser. No. 16 / 065,209 titled “Apparatus and methods for determining force applied to the tip of a probe” filed on Dec. 21, 2016, and U.S. patent publication number 2023 / 0157625 titled “Systems and methods for diagnosing preterm birth risk” filed on November 22, 2021, the entire disclosures of which are hereby incorporated by reference, for all purposes, as if fully set forth herein.BRIEF SUMMARY

[0003] One or more embodiments of the present disclosure can detect data including the impedance of cervical tissues at various frequencies, and provide information for predicting the preterm birth from the data. The embodiments presented herein also provide early diagnosis and treatment for preterm birth.BRIEF DESCRIPTION OF THE SEVERAL VIEWS OF THE DRAWINGS

[0004] FIG. 1 illustrates a block diagram illustrating of a system for providing information for diagnosing preterm birth risk with a cervix according to some embodiments of the present disclosure.DETAILED DESCRIPTION

[0005] Hereinafter, embodiments will be described in detail with reference to the drawings. In the following description, known configurations may be omitted. In addition, the following embodiments may be modified into various other forms, and the scope of the technical spirit of the present disclosure is not limited to the following examples. Rather, these embodiments are provided so that the present disclosure will be more thorough and complete, and will fully convey the scope of the technical spirit of the present disclosure to those skilled in the art.

[0006] It is to be understood that the disclosure herein is not intended to limit the scope to the described embodiments, but includes various modifications, equivalents, and / or alternatives of the embodiments. In the description of the drawings, like reference numerals refer to like elements throughout the description of drawings.

[0007] Conditional language used herein, such as, among others, “can,” “could,” “might,” “may,” “e.g.,” and the like, unless specifically stated otherwise, or otherwise understood within the context as used, is generally intended to convey that certain embodiments include, while other embodiments do not include, certain features, elements and / or steps. Thus, such conditional language is not generally intended to imply that features, elements and / or steps are in any way selected (or required) for one or more embodiments or that one or more embodiments necessarily include logic for deciding, with or without author input or prompting, whether these features, elements and / or steps are included or are to be performed in any particular embodiment. The terms “comprising,” “including,” “having,” and the like are synonymous and are used inclusively, in an open-ended fashion, and do not exclude additional elements, features, acts, operations, and so forth. Also, the term “or” is used in its inclusive sense (and not in its exclusive sense) so that when used, for example, to connect a list of elements, the term “or” means one, some, or all of the elements in the list. In addition, the articles “a,” “an,” and “the” as used in this application and the appended claims are to be construed to mean “one or more” or “at least one” unless specified otherwise. Except as specifically defined herein, all technical and scientific terms used herein are to be given as broad a commonly understood meaning as possible while maintaining claim validity.

[0008] As used herein, a phrase referring to “at least one of’ a list of items refers to any combination of those items, including single members. As an example, “at least one of: A, B, or C” is intended to cover: A, B, C, A and B, A and C, B and C, and A, B, and C. Conjunctive language such as the phrase “at least one of X, Y and Z,” unless specifically stated otherwise, isotherwise understood with the context as used in general to convey that an item, term, etc., may be at least one of X, Y or Z. Thus, such conjunctive language is not generally intended to imply that certain embodiments require at least one of X, at least one of Y and at least one of Z to each be present.

[0009] Similarly, while operations may be depicted in the drawings in a particular order, it is to be recognized that such operations need not be performed in the particular order shown or in sequential order, or that all illustrated operations be performed, to achieve beneficial results. Further, the drawings may schematically depict one more example process in the form of a flowchart. However, other operations that are not depicted can be incorporated in the example methods and processes that are schematically illustrated. For example, one or more additional operations can be performed before, after, simultaneously, or between any of the illustrated operations. Additionally, the operations may be rearranged or reordered in other implementations. In certain circumstances, multitasking and parallel processing may be advantageous. Moreover, the separation of various system components in the implementations described above should not be understood as requiring such separation in all implementations, and it should be understood that the described program components and systems can generally be integrated together in a single software product or packaged into multiple software products. Additionally, other implementations are within the scope of the following claims. In some cases, the actions recited in the claims can be performed in a different order and still achieve beneficial results.

[0010] Terms such as “first” and “second” may be used to modify various elements regardless of order and / or importance. Those terms are only used for the purpose of differentiating a component from other components.

[0011] When an element (e.g., a first constituent element) is referred to as being “operatively or communicatively coupled to” or “connected to” another element (e.g., a second constituent element), it should be understood that each constituent element is directly connected or indirectly connected via another constituent element (e.g., a third constituent element). However, when an element (e.g., a first constituent element) is referred to as being “directly coupled to” or “directly connected to” another element (e.g, a second constituent element), it should be understood that there is no other constituent element (e.g, a third constituent element) interposed therebetween.

[0012] The expression “configured to” as used in the present disclosure can refer to, for example, “suitable for,” “having the capacity to,” “designed to,” “adapted to,” “made to,” or“capable of’ depending on the situation. The term “configured to (or set to)” may not necessarily mean “specifically designed to” in hardware. Instead, in some circumstances, the expression “a device configured to” may mean that the device “is able to~” with other devices or components. For example, “a sub-processor configured to (or set to) execute A, B, and C” may be implemented as a processor dedicated to performing the operation (e.g., an embedded processor), or a generic-purpose processor (e.g., a central processor unit (CPU) or an application processor) that can perform the corresponding operations.

[0013] Hereinafter, the present disclosure will be described in detail with reference to the accompanying drawings. Embodiments will be described in detail with reference to the accompanying drawings so that those skilled in the art can easily embody the present disclosure. The present disclosure may, however, be embodied in many different forms and should not be construed as limited to the embodiments set forth herein. In order to clearly illustrate the present disclosure, parts not related to the description are omitted, and like elements are denoted by like reference numerals throughout the specification.

[0014] The above Figure l is a block diagram illustrating an apparatus or a system for providing information for diagnosing preterm birth risk according to some embodiments of the present disclosure. The diagnosis engine may comprise an algorithm, an Al (Artificial Intelligence) model, or a machine learning model. Within the context of the present disclosure, these terms (algorithm, Al model, machine learning model) can be used for referring to the same concept. The diagnosis engine may be trained with scientific discipline that is concerned with the design and development of algorithms that allow computers to learn based on data. The model can be improved through learning (e.g., by applying data to the model for “training” the algorithms). The data for training the Al model may be thus often referred to as “training data” or “learning data.” Examples of learning types may include supervised learning, unsupervised learning, semi-supervised learning, self-supervised learning, multi-instance learning, inductive learning, reinforcement learning, transudative learning, learning to learn, etc. The diagnosis engine / system may be trained with correlation between data including the plurality of impedances of the cervical tissues and the preterm birth risk.

[0015] In an embodiment, the diagnosis system may be trained with the input data. Input data may include at least one of spectral impedances, pH level, metabolites, temperature, pregnancy stage, and human microbiome, and the result corresponding the preterm birth risk, the diagnosis engine may be trained with correlation between input data, and the preterm birth risk. The diagnosis engine is organized into a taxonomy based on one or more outcomes of themodel and is trained by applying training data to the model. The model is adjusted (e.g., improved) based on how it responds to the training data. Multiple sets of training data may be applied to the same model so that the model may be repeatedly improved.

[0016] In an embodiment, the diagnosis engine may be trained with the supervised learning. Input data may include at least one of spectral impedances, pH level, metabolites, temperature, pregnancy stage, and human microbiome, and the result corresponding the preterm birth risk. The result may be, for example, one of high risk, medium risk, or low risk. Another example, the result may be numerical values of preterm birth risk. The input data may have been acquired from many patients. For example, the impedances may include impedances of cervical tissues of patients. The Input data may further comprise cervical tissue data of a patient, pregnancy stage, and other factors, etc. In an embodiment, the spectral impedances include impedances at different frequencies.

[0017] Within the context of the present disclosure, the two terms, “system” and “apparatus,” refer to the same concept and are used interchangeably.Input

[0018] Referring to above figure, Input 1, 2, . . . , N may be measured from one pregnant mother. For example, input 1 is the complex impedance at frequency 1 impedance, input 2 is the complex impedance at frequency 2, input 3 is history of preterm birth, input 3 4 is the pH level, etc.Pre-Process

[0019] The inputs (input 1, 2, . . . , N) may be pre-processed. Each input may be preprocessed. Pp(ppl) may be same as or different from another. The methodology of preprocessing may comprise at least one of or any combination of:Filtering or smoothing (capable of processing complex data e.g., R +jX), where j is the square root of -1;Linear or non-linear filters;Frequency selective filters (lowpass, highpass, bandpass);Isotropic filters;Averaging filters (including weighted averaging filters);Median filters, or alpha-trimmed filters ;Grouping inputs with weighted additions: e.g., inputl = al * xl + a2 * x2 + . . ., where the coefficients “a” are complex constants, and the “x”s are the raw inputs;Integration of inputs over time (this is equivalent to summing an input over time);Differentiation of inputs e.g., first derivative, second derivative, etc.);Using the history of preterm birth to process; andGeneralization of the impedances over one or more different frequencies (between 1..M), where each frequency can be greater than or less than 1 Hz. There can also be a number of impedances at the same frequency (between 1..N). Here M and N are positive integer constants.

[0020] The purpose of the preprocessing for the first classifiers is to remove noise and unwanted characteristics from the data that do not contribute to preterm risk accuracy.

[0021] After preprocessing, selecting / sorting / screening the preprocessed data to obtain data to be inputted into the Al model. Or by preprocessing, data to be inputted into the Al model may be selected, sorted or screened.

[0022] In an embodiment, some inputs are pre-processed and the other ones are not pre- processed. That is, said other ones may be directly inputted into the first classifiers without modificationFirst Classifier

[0023] Each of the first classifiers receives a plurality of pre-processed inputs that have been preprocessed to improve the overall performance of preterm risk accuracy.

[0024] Each of the first classifier receives the same or different groups of preprocessed inputs. For example, classifier 1 receives pre-processed input 1 to 3, and classifiers 2 and 3 receives pre-processed input 1, 3, 4, etc.

[0025] In an embodiment, each one of the first classifier may comprise at least one of or any combination of:Decision tree - the decision tree has a “roots,” “nodes” and “leaves.” The roots of the tree are the training datasets and they lead to specific nodes which denote a test attribute. Nodes often lead to other nodes, and a node that doesn’t lead onward is called a “leaf.” The decision tree can classify the data it is given by following the training data into sub-nodes until it reaches the conclusion on preterm birth;Random forest algorithm - the random forest builds different decision trees and connects them to gain more accurate results. These can be used for both classification and regression;Support vector machine (SVM) algorithm - SVM works by plotting each piece of data on a chart (in N dimensional space where N = the number of datapoints). Then, the algorithm classifies the datapoints by finding the hyperplace that separates each class. There can be more than one hyperplane, but in the end the result can predict the likelihood of preterm birth;Linear regression - it works by plotting pieces of data on a chart with the X-axis is the independent variable and the Y-axis is the dependent variable. The data points are then plotted out in a linear fashion to determine their relationship and forecast possible future data, which can be the likelihood of preterm birth;Logistic regression algorithm - it uses a binary value (0 / 1) to estimate values from a set of independent variables. The output of logistic regression is either 1 or 0, yes or no. Logistic regression is only useful when the dependent variable is categorical, either yes or no, which makes it useful for the diagnosis of preterm birth;Clustering and K-mean clustering - a clustering function works with a predetermined number of clusters and plots all the data regardless of the cluster. It then plots a randomly-selected piece of data as the centroid for each cluster. From there, it sorts the remaining data points into clusters based on their proximity to each other and the centroid data point for each cluster. The clusters can be defined as the likelihood of preterm birth; andNeural network algorithm - neural network algorithm is a term for a collection of Al algorithms that mimic the functions of a human brain. In unsupervised and supervised learning, it can be used for classification and pattern recognition. Neural networks can be used to predict the likelihood of preterm birth with a higher degree of sophistication, but also have some drawbacks such as “over learning.”

[0026] They each take the preprocessed inputs, and produce initial estimates of preterm birth, which are then post processed and sent to the meta classifier.

[0027] In an embodiment, each first classifier may use different algorithm among above.

[0028] In an embodiment, each first classifier can be trained and the preterm risk estimate accuracy optimized individually and independently of the other first classifiers. Equivalently, all first classifiers can be trained and the preterm risk estimate accuracy optimized at the same time in parallel.Post-Process

[0029] Each output from the first classifiers may be post-processed. The outputs may be optimized by the post-processing. In some embodiments, some outputs may not be postprocessed.

[0030] The purpose of the post-processing from the first classifiers is to keep the data characteristics from the data that contribute to preterm risk accuracy, and remove the characteristics that do not contribute to preterm risk accuracy.Meta Classifier

[0031] The first set of estimates of preterm birth risk from the first classifiers may be post processed, and then fed to the second classifier known as the “Meta” classifier. Therefore, the Meta classifier takes as inputs the post processed preterm birth estimates from one or more of the first classifiers, and then produces as the output the final preterm birth estimate of risk based on the preterm birth risk estimates produced from the first classifiers.

[0032] Outputs from the first classifiers can be identified which first classifier outputs which output. Where it comes.

[0033] More particularly, output 1 from first classifier 1, output 2 from first classifier 2, . . . ., output M from first classifier M can be identified where it comes. For example, inputs to the Meta classifier may be represent as {output 1, output 2, output 3, . . ., output M}. When the postprocessing is performed, the inputs to the Meta classifier may be represented as {post-processed output 1, post-processed output 2, post-processed output 3, . . ., post-processed output M}. Thus, the second classifier may recognize where the outputs come. The Meta classifier’s parameters may be determined during training the second classifier. In an embodiment, the second classifier may include at least one of a decision tree, a random forest algorithm, a support vector machine, a linear regression, a logistic regression algorithm, a clustering algorithm and a neural network algorithm.Result

[0034] The algorithm provides as an output an indication of the likelihood of a preterm birth occurring.Training

[0035] Al model may be trained with supervised learning. The diagnosis engine may be used for Al models for predicting pre-term birth risk. The data for training the Al model may comprise complex impedances, which are measured from cervical tissues of patients. The complex impedances can be represented in rectangular coordinates (e.g., Real +j * Imaginary, where Real and Imaginary are real-valued constants), or polar coordinates (Amplitude * eA(j * phase), where the amplitude is a positive real number and the phase is a real number - positive or negative (the phase is between 0 and 2*PI radians), the “A” symbol means “to the power of,” and j is the complex constant of square root of -1).

[0036] The complex impedances may include impedances at different frequencies. The patients may include pregnant mothers who are in the pre-term birth risk or in normal-term birth. Thus, the complex impedances may match with the patients, which can provide labeled data for the supervised learning.

[0037] In an embodiment, the pregnant mothers may have a preterm birth in the past or not, which can be represented as “a history of preterm birth.” The history of preterm birth may provide a weight to the input data, e.g., yes or no. The history of preterm birth may be one data for training the Al model. The data for training the Al model may further comprise another data, e.g., pH level, metabolites, temperature, pregnancy stage, and human microbiome of the pregnant mothers.

[0038] The applicant can collect the likelihood of preterm birth risk (which may be used as labeled data for a supervised learning) as well as the input data (e.g., input 1, . . ., N) from a plurality of pregnant mothers. Thus, the diagnosis engine can be trained using the input data and the likelihood of preterm birth risk as a learning data or training data. Even the diagnosis engine is trained with unsupervised learning, the result of the unsupervised learning (e.g., clustering) can be compared with the likelihood of preterm birth risk. For example, after clustering, the clustered group are grouped as shown below. Here, cluster 1 corresponds to the preterm birth, cluster 2 corresponds to the normal birth, and cluster 3 needs be re-determined. In order for the cluster 3 to be re-determined, a different Al model may be used. The different Al model may use the optional data for determining the cluster 3 with the preterm birth or normal birth. The Clustered group may be 2 or any different number.before

[0039] In an embodiment, one Al model may include one of a decision tree, a random forest algorithm, a support vector machine, a linear regression, a logistic regression algorithm, a clustering algorithm and a neural network algorithm. Another Al model may include one of a decision tree, a random forest algorithm, a support vector machine, a linear regression, a logistic regression algorithm, a clustering algorithm and a neural network algorithm, but different algorithm from said one Al model. Said Al model and said another Al model may use same input or different inputs during the training.

[0040] In an embodiment, two or more different Al models may include same Al algorithm, e.g., one of a decision tree, a random forest algorithm, a support vector machine, a linear regression, a logistic regression algorithm, a clustering algorithm and a neural network algorithm.

[0041] In an embodiment, two or more different Al models may use same input and the likelihood of preterm birth risk to be trained as learning data (or training data). The two or more different Al models may use different input and likelihood of preterm birth risk associated with said different input.

[0042] More particularly, one Al model may use input data including spectral impedances, pH level, metabolites, temperature, pregnancy stage, preterm birth history, and human microbiome, and likelihood of preterm birth risk associated with said input data. Another Al model may also use input data including spectral impedances, pH level, metabolites, temperature, pregnancy stage, preterm birth history, and human microbiome and likelihood of preterm birth risk associated with said input data as a training data. The spectral impedances said one Al model uses for the training are different from the spectral impedances said another Al model uses. For example, the spectral impedances said one Al model uses may be measured at frequencies different from the spectral impedances said another Al model uses.

[0043] For example, input data every Al model may include is complex impedances measured at a plurality of different frequencies from pregnant mothers.

[0044] For example, said another Al model may exclude at least one of spectral impedances, pH level, metabolites, temperature, pregnancy stage, preterm birth history, and human microbiome from the training data.

[0045] For example, said another Al model may use only spectral impedances and likelihood of preterm birth risk associated with said spectral impedances.

[0046] Referring to above figure again, a supervised learning using the inputs 1 to N, and result from pregnant mothers may be applied thereto. A user of the diagnosis engine may not know parameters inside the first classifiers and the meta classifier due to nature of the machine learning. The user sets the first classifiers and the meta classifier using at least one of the methodologies and let them learn with the inputs 1 to N, and result from pregnant mothers.

[0047] In an embodiment, the second classifier may be trained with outputs from the first classifiers and the and likelihood of preterm birth risk associated with inputs to the first classifiers.

[0048] The various embodiments described above can be combined to provide further embodiments. All of the U.S. patents, U.S. patent application publications, U.S. patent applications, foreign patents, foreign patent applications and non-patent publications referred to in this specification and / or listed in the Application Data Sheetare incorporated herein by reference, in their entirety. Aspects of the embodiments can be modified, if necessary to employ concepts of the various patents, applications and publications to provide yet further embodiments.

[0049] These and other changes can be made to the embodiments in light of the abovedetailed description. In general, in the following claims, the terms used should not be construed to limit the claims to the specific embodiments disclosed in the specification and the claims, but should be construed to include all possible embodiments along with the full scope of equivalents to which such claims are entitled. Accordingly, the claims are not limited by the disclosure.

[0050] This application claims the benefit of priority to U.S. Application No. 63 / 591,368, filed October 18, 2023, which application is hereby incorporated by reference in its entirety.

Claims

CLAIMS1. A diagnosis system for predicting a preterm birth risk, comprising: an artificial intelligence engine comprising: a plurality of first classifiers, each having as first input of complex impedances measured at a plurality of different frequencies from cervical tissues during pregnancy and optionally as a second input at least one of preterm history, pregnancy stage, age, pH level, etc., and a Meta classifier configured to output an indication of the preterm birth risk, with as a third input being the outputs of each of the first classifiers, wherein the first and the second inputs are pre-processed using a plurality of different pre-processing methods, each the same or different for each classifier, and wherein the third inputs are post-processed using a plurality of different postprocessing methods, each the same or different for the outputs of each first classifiers.

2. The diagnosis system of claim 1, wherein the first classifier includes at least one of a decision tree, a random forest algorithm, a support vector machine, a linear regression, a logistic regression algorithm, a clustering algorithm and a neural network algorithm.

3. The diagnosis system of claim 1, wherein the first classifier includes a different Al model, respectively.

4. The diagnosis system of claim 1, wherein the pre-processing method includes at least on of filtering, smoothing, weighting, integration, differentiation, generalization.

5. The diagnosis system of claim 1, wherein the third inputs is represent as {a, b, c, . . ., m}, a to m are outputs originated from the first classifiers.

6. A diagnosis system for predicting a preterm birth risk, comprising: an artificial intelligence engine comprising: a plurality of first classifiers, each having as input of complex impedances measured at a plurality of different frequencies from cervical tissues during pregnancy and as input zero or more of preterm history, pregnancy stage, age, pH level, etc., anda Meta classifier configured to output an indication of the preterm birth risk, with as input being the outputs of each of the first classifiers, wherein each of the first classifiers has inputs that were pre-processed using a plurality of different preprocessing methods, each the same or different for preprocessor for each classifier, and wherein the Meta classifier has inputs that are the post-processed outputs from the first classifiers using a plurality of different post-processing methods, each the same or different for the outputs of each first classifiers.

Images