System, method and apparatus for identifying congenital long qt syndrome in subject

Abstract

To provide an apparatus and method for training an artificial intelligence-supported diagnostic assessment tool.SOLUTION: Described herein are systems, methods and apparatuses for identifying LQTS in a subject. A system may include an electrocardiogram (ECG) sensor configured to detect ECG data. A system may include a computing device configured to: receive electrocardiogram (ECG) data; train an LQTS determination machine learning model on a training dataset including multiple pieces of example ECG data as inputs correlated with multiple pieces of example LQTS data as outputs; and generate LQTS determination data as a function of the ECG data using the trained LQTS determination machine learning model. A system may include a display configured to display LQTS determination data.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates generally to the field of machine learning, and more particularly to a system, method and apparatus for identifying congenital long QT syndrome in a subject. [Background technology]

[0002] ECG-confirmed QT prolongation, particularly when the heart rate-corrected QT interval (QTc) exceeds 500 ms, is associated with an increased risk of cardiac arrhythmias that can lead to syncope, seizures, sudden cardiac arrest (SCA), or ultimately sudden cardiac death (SCD). While QT prolongation is a characteristic feature of congenital long QT syndrome (LQTS), prolonged QTc can also be observed in patients with electrolyte abnormalities (primarily hypokalemia and hypomagnesemia), certain cardiac and noncardiac conditions with disease-related QT prolongation, or in the setting of medications that can prolong the QT interval. Indeed, over 100 FDA-approved drugs have been reported to cause QT prolongation and even drug-induced prolonged QT-mediated SCD, which is often the only or best treatment option for certain patients. More importantly, while LQTS is rare with an estimated prevalence of 1 in 2,000, acquired QT prolongation is fairly common, occurring in approximately 10% of patients undergoing ECG at tertiary centers, and, more importantly, is potentially reversible by addressing its underlying cause. Summary of the Invention

[0003] In one aspect, an apparatus for generating LQTS decision data may include at least a processor and a memory communicatively connected to the at least processor, wherein the memory includes instructions configuring at least the processor to receive electrocardiogram (ECG) data, train an LQTS decision machine learning model on a training dataset including a plurality of exemplary ECG data as inputs correlated to a plurality of exemplary LQTS data as outputs, and generate LQTS decision data as a function of the ECG data using the trained LQTS decision machine learning model, wherein the LQTS decision machine learning model includes a neural network having a plurality of 2D convolutional stacked blocks and a final output layer activated using at least an activation function.

[0004] In another aspect, a method for generating LQTS decision data may include receiving, using at least a processor, electrocardiogram (ECG) data; training, using at least the processor, an LQTS decision machine learning model on a training dataset including a plurality of exemplary ECG data as inputs correlated with a plurality of exemplary LQTS data as outputs; and generating, using at least the processor, LQTS decision data as a function of the ECG data using the trained LQTS decision machine learning model, wherein the LQTS decision machine learning model includes a neural network having a plurality of 2D convolutional stacked blocks and a final output layer activated using at least an activation function.

[0005] In another aspect, a system for generating LQTS decision data includes an electrocardiogram (ECG) sensor configured to detect ECG data and transmit the ECG data to a computing device; a computing device communicatively connected to the ECG sensor, configured to receive the electrocardiogram (ECG) data from the ECG sensor, generate LQTS decision data as a function of the ECG data using a trained LQTS decision machine learning model, and transmit the LQTS decision data to a display, the LQTS decision machine learning model including a neural network having a plurality of 2D convolutional stacked blocks and a final output layer activated using at least an activation function; and a display configured to receive the LQTS decision data from the computing device and display the LQTS decision data.

[0006] These and other aspects and features of non-limiting embodiments of the present invention will become apparent to those skilled in the art upon review of the following description of specific non-limiting embodiments of the present invention in conjunction with the accompanying drawings. For the purpose of illustrating the invention, the drawings show aspects of one or more embodiments of the invention, it being understood, however, that the invention is not limited to the precise arrangements and instrumentalities shown in the drawings. [Brief explanation of the drawings]

[0007] [Figure 1] FIG. 1 illustrates an exemplary system for identifying LQTS in a subject. [Figure 2] Figure 2 is a flow chart of the study cohort. [Figure 3] Figure 3 is a table describing the baseline characteristics of the case (LQTS) and control (acquired long QT) populations (matching ratio 1:5). [Figure 4A] FIG. 4A is a graph of the receiver operating characteristic curves of the first ECG with and without the pacemaker / QRS filter at a matching ratio of 1:5. [Figure 4B]FIG. 4B is a graph of the receiver operating characteristic curves of the first ECG with and without the pacemaker / QRS filter at a matching ratio of 1:10. [Figure 5] FIG. 5 is a table illustrating the main and subgroup analysis examining the performance of the AI-ECG-LQTS detection algorithm. [Figure 6] Figure 6 is a table illustrating the validation of the AI-ECG-LQTS detection model on all ECGs of the cases. [Figure 7] FIG. 7 is a table describing the baseline characteristics of the LQTS patient and control populations (matching ratio=2000). [Figure 8] Figure 8 is a flow chart of the study cohort. [Figure 9] FIG. 9 is a graph of the receiver operating characteristic curves of the first ECG with and without the pacemaker / QRS filter at a matching ratio of 1:2000. [Figure 10] FIG. 10 is a graph of the receiver operating characteristic curves of the first ECG with and without the pacemaker / QRS filter at a matching ratio of 1:200. [Figure 11] FIG. 11 is a box diagram of an exemplary machine learning model. [Figure 12] FIG. 12 is a diagram of an exemplary neural network. [Figure 13] FIG. 13 is a diagram of an exemplary neural network node. [Figure 14] FIG. 14 illustrates an exemplary method for identifying LQTS in a subject. [Figure 15] FIG. 15 is a block diagram of a computing system that may be used to implement any one or more of the methods disclosed herein and any one or more portions thereof. DETAILED DESCRIPTION OF THE INVENTION

[0008] The drawings are not necessarily to scale and may be illustrated by phantom lines, schematic diagrams, and partial views. In some cases, details that are not necessary for understanding the embodiments or that make other details difficult to perceive may be omitted. At a high level, aspects of the present disclosure relate to systems, methods, and devices for identifying congenital long QT syndrome (LQTS) in a subject. In some embodiments, one or more electrocardiogram (ECG) sensors can be used to detect the subject's ECG data, which can be input into a trained neural network to determine whether the subject has LQTS and / or acquired long QT. In some embodiments, the neural network can be trained using ECG data for multiple previous subjects and a determination of whether such previous subjects had LQTS by another means, such as genetic testing.

[0009] Referring now to FIG. 1 , an exemplary embodiment of a system 100 for identifying LQTS in a subject is shown. The system 100 may include a computing device. The system 100 may include a processor. The processor may include, but is not limited to, any processor described herein. The processor may be included in a computing device. The computing device may include any computing device described herein, including, but not limited to, a microcontroller, a microprocessor, a digital signal processor (DSP), and / or a system-on-chip (SoC) described herein. The computing device may be included in, and / or communicate with, a mobile device, including a mobile device such as a mobile phone or smartphone. The computing device may include a single computing device operating independently, or may include two or more computing devices operating in cooperation, parallel, sequentially, etc., where two or more computing devices may be included together in a single computing device or two or more computing devices. The computing device may interface or communicate with one or more additional devices via a network interface device, as described in further detail below. The network interface device may be utilized to connect the computing device to one or more of various networks and one or more devices. Examples of network interface devices include, but are not limited to, a network interface card (eg, a mobile network interface card, a LAN card), a modem, and any combination thereof.Examples of networks include, but are not limited to, wide area networks (e.g., the Internet, enterprise networks), local area networks (e.g., networks associated with an office, building, campus, or other relatively small geographic space), telephone networks, data networks associated with telephone / voice providers (e.g., mobile communications provider data and / or voice networks), direct connections between two computing devices, and any combination thereof. Networks can use wired and / or wireless communication modes. In general, any network topology can be used. Information (e.g., data, software, etc.) can be communicated between computers and / or computing devices.

[0010] 1 , in some embodiments, system 100 may include at least processor 104 and memory 108 communicatively coupled to at least processor 104, where memory 108 includes instructions 112 that configure at least processor 104 to perform one or more processes described herein. Computing device 116 may include processor 104 and / or memory 108. Computing device 116 may be configured to perform one or more processes described herein.

[0011] 1 , computing device 116 may include, for example, but not limited to, a computing device or cluster of computing devices at a first location and a second computing device or cluster of computing devices at a second location. Computing device 116 may include one or more computing devices dedicated to data storage, security, traffic distribution for load balancing, etc. Computing device 116 may distribute one or more computing tasks, as described below, across multiple computing devices that may operate in parallel, serially, redundantly, or any other manner used to distribute tasks or memory among computing devices. Computing device 116 may be implemented using a “shared nothing” architecture, as a non-limiting example.

[0012] Continuing with reference to FIG. 1 , computing device 116 can be designed and / or configured to perform any method, method step, or series of method steps in any embodiment described herein, in any order and with any degree of repetition. For example, computing device 116 may be configured to repeatedly perform a single step or sequence until a desired or directed result is achieved, where the repetition of a step or series of steps may be performed iteratively and / or recursively using the output of a previous iteration as input to a subsequent iteration, where the inputs and / or outputs of the iterations may be aggregated to produce an aggregate result, where one or more variables, such as a global variable, may be decreased or decreased, and / or a larger processing task may be divided into a set of smaller processing tasks addressed iteratively. Computing device 116 may perform any step or series of steps described herein in parallel, such as performing two or more steps simultaneously and / or substantially simultaneously using two or more parallel threads, processor cores, etc., and the division of tasks among parallel threads and / or processes may be performed according to any protocol suitable for dividing tasks among iterations. Those skilled in the art will recognize, upon reviewing this disclosure in its entirety, various ways in which processes, sequences of processes, processing tasks, and / or data can be subdivided, shared, or otherwise processed using iterative, recursive, and / or parallel processing.

[0013] With further reference to FIG. 1 , as used in this disclosure, “communicatively connected” means connected by a connection, attachment, or link between two or more devices that allows for the receipt and / or transmission of information between them. For example, without limitation, this connection may be wired or wireless, direct or indirect, and between two or more components, circuits, devices, systems, etc., allowing for the receipt and / or transmission of data and / or signals between them. The data and / or signals between them may include, but are not limited to, electrical, electromagnetic, magnetic, video, audio, radio, and microwave data and / or signals, combinations thereof, among others. A communicative connection may be achieved, for example, without limitation, via wired or wireless electronic, digital, or analog communication, directly or through one or more intervening devices or components. Furthermore, a communicative connection may include electrically coupling or connecting at least an output of one device, component, or circuit to at least one input of another device, component, or circuit. For example, without limitation, via a bus or other facility for intercommunication between elements of a computing device. A communicative connection may also include an indirect connection, for example, but not limited to, via a wireless connection, a wireless communication, a low-power wide area network, optical communication, magnetic coupling, capacitive coupling, or optical coupling, etc. In some cases, the term "communicatively coupled" may be used instead of communicatively connected in this disclosure.

[0014] 1 , in some embodiments, system 100 can include an electrocardiogram (ECG) sensor 120. ECG sensor 120 can include one or more electrodes. The electrodes can be placed on subject 124, such as on the chest, arms, and legs of subject 124. ECG sensor 120 can include one or more leads. ECG sensor 120 can include a device configured to measure and / or interpret the electrical activity of the subject's heart using the electrodes and / or leads. In some embodiments, ECG sensor 120 can be configured to detect ECG data 128 and / or transmit ECG data 128 to computing device 116. As used herein, "ECG data" refers to data describing the electrical activity of the subject's heart. In some embodiments, ECG data can include rhythm strip ECG data. As used herein, "rhythm strip ECG data" refers to data describing the electrical activity detected using a single electrode. In some embodiments, ECG data can include median beat ECG data. As used herein, "median beat ECG data" is data describing electrical activity detected using multiple leads and / or electrodes. In some embodiments, ECG data 128 may include data collected by two, three, four, five, six, seven, eight, nine, ten, eleven, twelve, or more ECG leads. For example, ECG data 128 may include a median beat collected by twelve ECG leads. In some embodiments, ECG data may be associated with a particular subject 124. In some embodiments, subject 124 may have a ventricular pacemaker.

[0015] With further reference to FIG. 1 , in some embodiments, ECG data 128 may be stored in data store 132 and / or memory 108. Data store 132 may be implemented as, without limitation, a relational database, a key-value lookup database such as a NOSQL database, or any other format or structure for use as a database that one of ordinary skill in the art would recognize as appropriate upon reviewing this disclosure in its entirety. Data store 132 may alternatively or additionally be implemented using a distributed data storage protocol and / or data structure, such as a distributed hash table. Data store 132 may include multiple data entries and / or records, as described above. Data entries in data store 132 may be flagged or linked with one or more additional elements of information that may be reflected in data entry cells and / or linked tables, such as tables related by one or more indexes in a relational database. Those of ordinary skill in the art, upon reviewing this disclosure in its entirety, will recognize various ways in which data entries in a database may store, search, organize, and / or reflect the data and / or records used herein, as well as categories and / or collections of data consistent with this disclosure. The ECG data stored in data store 132 can be used as exemplary ECG data 136 for training machine learning models, as described below. In some embodiments, data store 132 can include an electronic health record database. In some embodiments, the electronic health record database can include health information, such as ECG data and genetic LQTS test data, from multiple subjects. In some embodiments, the health information can be received in an anonymized state and / or can be anonymized by system 100, such as by removing identifying information.

[0016] 1 , in some embodiments, computing device 116 can receive ECG data 128. In some embodiments, computing device 116 can receive ECG data 128 from ECG sensor 120. In some embodiments, computing device 116 can receive ECG data 128 from data store 132. For example, computing device 116 can receive ECG data 128 from data store 132 in situations where a subject has previously received an ECG and it is desired to know whether the subject has LQTS.

[0017] 1 , in some embodiments, data store 132 and / or computing device 116 can receive exemplary LQTS data 140. In some embodiments, subject 124 can undergo a genetic LQTS test 144. In some embodiments, system 100 can include a genetic LQTS test sensor. The genetic LQTS test can include, for example, a test to determine whether a subject has a gene associated with LQTS. In some embodiments, the exemplary LQTS data can be used to train the machine learning models described herein.

[0018] Continuing with reference to FIG. 1 , in some embodiments, system 100 can determine LQTS decision data 148. Here, “LQTS decision data” is data describing whether a subject has LQTS. In a non-limiting example, LQTS decision data 148 can include a categorical answer regarding whether a subject has LQTS or the probability that a subject has LQTS. In some embodiments, LQTS decision data 148 can be determined using an LQTS decision machine learning model 152. LQTS decision machine learning model 152 can be trained using a supervised learning algorithm. LQTS decision machine learning model 152 can include a classifier. LQTS decision machine learning model 152 can include a neural network. LQTS decision machine learning model 152 can include a convolutional neural network. LQTS decision machine learning model 152 can be trained with training data 156 including exemplary ECG data associated with exemplary LQTS decision data. Such a training data set can be obtained, for example, by receiving training data 156 from data store 132. The exemplary ECG data 136 may include ECG data obtained from a previous electrocardiogram performed on the subject. In some embodiments, the exemplary ECG data 136 may be supplemented with ECG data 128 associated with the subject 124. In some embodiments, such data may be used to modify and / or retrain the LQTS determination machine learning model 152. The exemplary LQTS data 140 may be obtained from a previous genetic LQTS test and may be supplemented with data from a genetic LQTS test administered to the subject 124, such as for model retraining purposes. Once the LQTS determination machine learning model 152 is trained, it may be used to determine the LQTS determination data 148. The system 100 may input the ECG data 128 into the LQTS determination machine learning model 152, and the system 100 may receive the LQTS determination data 148 from the model.

[0019] 1 , in some embodiments, the LQTS determination machine learning model 152 can be trained using a reinforcement learning algorithm. For example, the LQTS determination machine learning model 152 can be provided with inputs such as ECG data 128, and the LQTS determination machine learning model 152 can be adjusted based on a cost function, which is based on the model's output. Such a cost function can take into account, for example, whether the obtained LQTS determination data matches the results of a genetic LQTS test.

[0020] 1 , in some embodiments, the LQTS determination machine learning model 152 can be trained using training data 156 including data on subjects who had at least one ECG above the age- and gender-specific 99th percentile for QTc. In some embodiments, the training data 156 can include data on subjects with a QTc > 460 ms for all patients (e.g., not limited to males and females) who are under the age of 13. In some embodiments, the training data 156 can include data on male subjects over the age of 13 with a QTc > 470 ms. In some embodiments, the training data 156 can include data on female subjects over the age of 13 with a QTc > 480 ms. In some embodiments, the training data 156, such as the exemplary LQTS data 140, can include a 1:5 ratio of LQTS output to non-LQTS output. In some embodiments, the training data 156, such as the exemplary LQTS data 140, can include a 1:10 ratio of LQTS output to non-LQTS output. In some embodiments, training data 156, such as exemplary LQTS data 140, may include a 1:200 ratio of LQTS output to non-LQTS output. In some embodiments, training data 156, such as exemplary LQTS data 140, may include a 1:2000 ratio of LQTS output to non-LQTS output. In some embodiments, training data 156, such as exemplary LQTS data 140, may include a 1:X ratio of LQTS output to non-LQTS output, where X is in the range of 5 to 2000. In some embodiments, the control cohort may include male or female patients with acquired QT prolongation <13 years of age, QTc >470 ms for males over 13 years of age, QTc >460 for males over 13 years of age, and QTc >480 ms for females over 13 years of age. In some embodiments, the ECG data may include rhythm strip ECG data. In some embodiments, the exemplary ECG data 136 and / or ECG data 128 may include data in the form of a data structure having dimensions including the number of ECG leads and the number of consecutively obtained readings.In a non-limiting example, data captured from a 12-lead ECG monitored for 10 seconds at 500 Hz can include 5000 data points per lead, generating a 5000 x 12 x 1 input ECG shape. In some embodiments, ECG data having different frequencies, durations, and / or numbers of leads can be used. In some embodiments, the ECGs can be distributed into training, validation, and test data sets. Such distribution can be random. In some embodiments, the training data set can include 70% of the entire data set, the validation data set can include 10% of the entire data set, and the test data set can include 20% of the entire data set. In some embodiments, the training data ECGs can be zero-padded, such as from 5000 x 12 x 1 to 5120 x 12 x 1. In some embodiments, the LQTS decision machine learning model 152 can include multiple stacked blocks of convolutional layers. For example, the LQTS decision machine learning model 152 can include multiple stacked blocks of 2D convolutional layers. For example, the LQTS determination machine learning model 152 may include eight stacked blocks of 2D convolutional layers. In some embodiments, the LQTS determination machine learning model 152 may include one or more of max pooling, batch normalization, dropout, and fully connected layers. For example, the LQTS determination machine learning model 152 may include two fully connected layers. The LQTS determination machine learning model 152 may include a final output layer. Such a final output layer may be activated using a function that assigns a probability to each class of multiple classes. For example, such multiple classes may include a first class associated with subjects with LQTS and a second class associated with subjects without LQTS. In some embodiments, such a final output layer may be activated using a softmax function. In some embodiments, median beat ECG data may be used to train the LQTS determination machine learning model 152. For example, the median beat ECG data may have an input ECG shape of 600 x 12 x 1.

[0021] Continuing with reference to FIG. 1 , in some embodiments, two or more LQTS determination machine learning models 152 are generated. The performance of the LQTS determination machine learning models 152 can be evaluated using the area under the curve (AUC) of a receiver operating characteristic (ROC) curve, positive predictive value (PPV), negative predictive value (NPV), sensitivity, specificity, accuracy, false positive rate (FPR), and / or false negative rate (FNR). In some embodiments, the LQTS determination machine learning model 152 may be selected from multiple models based on one or more such metrics. In some embodiments, the LQTS determination machine learning model 152 may be evaluated using a matching ratio other than 1:5, such as 1:10. In some embodiments, the LQTS determination machine learning model 152 can be evaluated using data on subjects with pacemakers.

[0022] 1 , in some embodiments, system 100 can identify LQTS in a subject. In some embodiments, identification of LQTS in a subject may be made as a function of LQTS determination data 148. In some embodiments, a therapy may be administered to a subject as a function of LQTS determination data 148. For example, a therapy approved for treating LQTS may be administered to a subject having LQTS determination data 148 that indicates the subject is likely to have LQTS. In some embodiments, an appointment with a medical professional may be made based on LQTS determination data 148. For example, an appointment with a medical professional may be made for a subject having LQTS determination data 148 that indicates a high likelihood of having LQTS. Such an appointment may be used, for example, to discuss treatment options.

[0023] 1 , in some embodiments, system 100 can display LQTS determination data 148 to a user, such as subject 124 and / or a medical professional. User device 160 and / or user interface 164 can be used to display LQTS determination data 148. In some embodiments, system 100 can transmit a signal including LQTS determination data 148 to user device 160, which can configure user device 160 to communicate LQTS determination data 148 to the user. User device 160 may communicate LQTS determination data 148 to the user using, for example, a visual format or an audio format. System 100 can communicate visual elements and / or visual element data structures including LQTS determination data 148 to user device 160, which can configure user device 160 to display the visual elements, such as using user interface 164. As used herein, a device “displays” data if the device outputs the data in a format suitable for communication to a user. For example, a device may display data by outputting text or images on a screen, or by outputting sound using a speaker.

[0024] 1 , in some embodiments, the visual element data structure may include a visual element. As used herein, a “visual element” is data that is visually displayed to a user. In some embodiments, the visual element data structure may include rules for displaying the visual element. In some embodiments, the visual element data structure may be determined as a function of the LQTS determination data 148. In some embodiments, the visual element data structure may be determined as a function of an item from a list consisting of an identifier for the subject 124, the ECG data 128, and the LQTS determination data 148. In a non-limiting example, the visual element data structure may be generated such that a visual element describing or highlighting the LQTS determination data 148 is displayed to a user, such as the subject 124.

[0025] 1 , in some embodiments, the visual element may include one or more elements such as text, images, shapes, charts, particle effects, interactable features, etc. For example, the visual element may include a depiction of ECG data 128 along with LQTS determination data 148.

[0026] 1, the visual element data structure may include rules that govern whether or when a visual element is displayed. In a non-limiting example, the visual element data structure may include rules that cause a visual element describing LQTS decision data 148 to be displayed when a user selects LQTS decision data 148 using a graphical user interface (GUI).

[0027] 1, the visual element data structure may include multiple visual elements or rules for presenting multiple visual elements at one time, in one embodiment, approximately 1, 2, 3, 4, 5, 10, 20, or 50 visual elements are displayed simultaneously.

[0028] With further reference to FIG. 1 , a visual element data structure rule can be applied to a single visual element or data, or to two or more visual elements or data. The visual element data structure can classify data into one or more categories and can apply a rule to all data in a category, all data at the intersection of categories, or all data in a subsection of a category (e.g., all data in a first category but not in a second category). For example, multiple ECGs for a subject may be analyzed, and if LQTS determination data 148 based on such multiple ECGs indicates the same result, similar visual elements may be generated. The visual element data structure can rank the data or assign them numerical values. For example, a visual element describing LQTS may be ranked higher for a subject receiving LQTS determination data 148 indicating that the subject has LQTS. The numerical value can, for example, measure the degree to which the first data is associated with a category or second data. The visual element data structure can apply a rule based on the ranking or comparison of the numerical value to a threshold value. For example, a visual element with a higher ranking may be displayed more prominently. Rankings, numerical values, categories, etc. can be used to configure visual element data structure rules.

[0029] 1 , in some embodiments, the visual elements can be interacted with. For example, the visual elements can include an interface such as a button or a menu. In some embodiments, the visual elements can be interacted with using a user device such as a smartphone, tablet, smartwatch, or computer.

[0030] 1 , in some embodiments, computing device 116 can send a visual element data structure to user device 160. In some embodiments, the visual element data structure can configure user device 160 to display the visual element. In some embodiments, the visual element data structure can cause an event handler to be triggered in an application of user device 160, such as a web browser. In some embodiments, the triggering of the event handler can cause a change in the application of user device 160, such as the display of the visual element.

[0031] 1 , in some embodiments, the system 100 can transmit the visual elements to a display. The display can communicate the visual elements to a user, such as the subject 124. The display can include, for example, a smartphone screen, a computer screen, or a tablet screen. The display can be configured to provide a visual interface. The visual interface can include one or more virtual interactive elements, such as, but not limited to, buttons, menus, etc. The display can include one or more physical interactive elements, such as buttons, a computer mouse, or a touch screen, that allow a user, such as the subject 124, to input data into the display. The interactive elements can be configured to enable interaction between the user and the computing device. In some embodiments, the visual element data structure is determined as a function of data input into the display by a user, such as the subject 124.

[0032] With further reference to FIG. 1 , the variables and / or data described herein may be represented as a data structure. In some embodiments, the data structure may include one or more functions and / or variables, such as a class in object-oriented programming. In some embodiments, the data structure may include data in the form of a Boolean, integer, float, string, date, etc. In a non-limiting example, an LQTS determination data structure may include a Boolean value representing whether a subject is predicted to have LQTS. In some embodiments, the data in the data structure may be organized as a linked list, tree, array, matrix, tensor, etc. In a non-limiting example, ECG data 128 may be organized into a matrix. In some embodiments, the data structure may include one or more elements of metadata or be associated with one or more elements of metadata. The data structure may include one or more self-referential data elements that the processor 104 can use in interpreting the data structure. In a non-limiting example, the data structure may include a "*" tag indicating that the content between the tags is a date. <date> " and "< / date> " tag. The data structure can be stored in a database.

[0033] 1, in some embodiments, the data structure may be read and / or manipulated by the processor 104. In a non-limiting example, the ECG data structure may be read and input to the LQTS determination machine learning model 152.

[0034]

[0013] Still referring to Figure 1, in some embodiments, the data structure can be calibrated. In some embodiments, the data structure can be trained using a machine learning algorithm. In a non-limiting example, the data structure can include an array of data representing biases of connections in a neural network. In this example, the neural network can be trained with a set of training data, and a backpropagation algorithm can be used to correct the data in the array. Machine learning models and neural networks are further described herein.

[0035] Objective: To examine whether an artificial intelligence (AI) deep neural network (DNN) derived analysis of 12-lead electrocardiograms (ECGs) can distinguish patients with LQTS from those with acquired long QT.

[0036] Patients and Methods: The study cohort included all patients with genetically confirmed LQTS evaluated at the Windland Smith Rice Genetic Heart Rhythm Clinic and controls from the Mayo Clinic's ECG data vault, which contains over 2.5 million patients. For the AI-DNN model, we included all patients / controls with one or more ECGs with age- and sex-specific QTc values ​​above the 99th percentile [>460 ms <13 years for all patients (male / female), or >470 ms for men, and >480 ms for women over this age]. LQTS patients were age- and sex-matched with controls at a 1:5 ratio. To classify patients, we developed an AI-DNN containing a multilayer convolutional neural network (CNN).

[0037] Results: Of 1,599 patients with genetically confirmed LQTS, 808 had one or more ECGs with QTc above the defined threshold (2,987 ECGs) compared with 361,069 / 2.5M controls (14% of Mayo Clinic patients have an ECG, "presumed negative"; 989,313 ECUs). After age and sex matching and partitioning, 3,309 (training), 411 (validation), and 887 (test) ECGs were used. The model distinguished LQTS from those with acquired QT prolongation with an AUC of 0.896 (accuracy 85%, sensitivity 77%, specificity 87%, PPV 0.58, NPV 0.94). The model remained robust with an AUC approaching or exceeding 0.9, regardless of the matching ratio (range 1:5-1:2,000), the type of ECG data used (rhythm strip or median beat), or after excluding patients with a wide QRS and / or ventricular pacemaker.

[0038] Conclusions: For patients with QTc >99th percentile, this novel AI-DNN serves as an LQTS mutation detector that can identify patients with abnormal QT prolongation secondary to LQTS-causing mutations, rather than acquired QT prolongation, with a positive predictive value of >50% in certain models. This algorithm can facilitate screening for this potentially fatal but highly treatable genetic heart disease.

[0039] Therefore, careful evaluation and timely recognition of QT prolongation and its possible causes are paramount to prevent potentially life-threatening sequelae. The inventors herein set out to examine whether an AI-ECG neural network can distinguish QT prolongation caused by congenital LQTS (i.e., due to the presence of an LQTS-causing mutation) from QT prolongation caused by other exogenous conditions or diagnoses ("acquired" QT prolongation). The development and implementation of such a model can assist providers in early recognition, accurate diagnosis, and appropriate management, including diagnosing LQTS or addressing the underlying causes of acquired QT prolongation.

[0040] method

[0041] Cohort Selection

[0042] All study data were collected from Mayo Clinic's electronic medical records. The Mayo Clinic Institutional Review Board approved this study. We identified all patients with genetically confirmed LQTS seen at the Mayo Clinic Windland Smith Rice Genetic Heart Rhythm Clinic between July 1, 1999, and October 1, 2021. Among these, we used an otherwise healthy population-based standard 99th percentile threshold for QTc, defined as a QTc of >480 ms for women (age 13 or older), >470 ms for men (age 13 or older), and >460 ms for prepubertal males or females (age <13), to identify patients with LQTS who exceeded this threshold on at least one ECG.

[0043] A "control" cohort of patients with acquired QT prolongation was derived from the Mayo Clinic CVMuse ECG database of over 10 million ECGs from over 2.5 million patients (June 1987 to September 2021) using the same 99th percentile QTc cutoff value. All patients with a QTc above the age-sex-specified QT cutoff who i) passed research approval and ii) were not evaluated for QT prolongation in a genetic cardiac rhythm clinic were included in this cohort. Demographics, including age, sex, and BMI at the time of ECG, were extracted for the training, validation, and testing datasets.

[0044] AI model development

[0045] For each patient, a 12-lead ECG monitored at 500 Hz for 10 seconds contained 5,000 data points per lead, providing an input ECG shape of 5,000 × 12 × 1 (60,000 data points per ECG). To train the model, ECGs were randomly shuffled by patient into mutually exclusive training (70%), internal validation (10%), and testing (20%) datasets. We constructed a convolutional neural network (CNN) using the Keras Framework with a Tensorflow backend (Google; Mountain View, CA, USA). The CNN was constructed using a total of eight stacked blocks: a 2D convolutional layer, max pooling, batch normalization, and dropout layers, followed by two fully connected layers, and a final output layer activated using a softmax function to generate two outputs (negative and positive, or LQTS or not). To better accommodate the model's shape manipulation, all ECGs were zero-padded from 5,000 × 12 × 1 to 5,120 × 12 × 1 to maintain uniform dimensions after each pooling layer. Attrition rates of 0.5 and 30 epochs were used to train the network, preserving the model weights from the epoch that achieved the highest area under the curve (AUC ROC) on the internal validation set. The model was trained on a GPU cluster using an Nvidia GeForce Tesla V100 graphics card. The model was tested on the patient's initial ECG and all of the patient's subsequent ECGs obtained at the Mayo Clinic.

[0046] In addition to using 10-second, 12-lead rhythmic ECGs, a separate AI-ECG model using median-beat ECGs was developed for the same cohort to test the model's robustness. For median-beat ECGs, each patient's input was calculated to create a 5000 × 12 × 1 ECG shape instead of a 600 × 12 × 1 shape, representing the patient's median heartbeat across all 12 combined leads. Using similar infrastructure (2D convolution, pooling, batch normalization, and dropout rate), each model was trained independently and its performance compared. Compared to rhythmic ECGs, no zero-padding was applied to the median-beat model.

[0047] Age and gender matching

[0048] For the initial model, performance indicated that the ideal matching ratio was 1:5 (LQTS vs. controls). Performance was monitored and evaluated using the area under the curve (AUC) of the receiver operating characteristic (ROC) curve, as well as the model's positive predictive value (PPV), negative predictive value (NPV), sensitivity, specificity, accuracy, false positive rate (FPR), and false negative rate (FNR). As a secondary analysis, the accuracy of the model in distinguishing LQTS from patients with acquired QT prolongation ("controls") was evaluated at a higher matching ratio of 1:10, independent of the primary analysis. Furthermore, because a wide QRS complex or an active ventricular pacemaker is known to affect the accuracy of QT interval measurements, we further evaluated the performance of the algorithm by performing analyses on the same patients but after excluding patients with a QRS >150 ms (for both 1:5 and 1:10 matching) and / or after excluding patients with pacemakers.

[0049] Prevalence-based matching

[0050] Next, to assess the model's ability to be used as a potential screening tool in clinical practice, we tested its performance at additional matching ratios, both 1:2,000, which reflects the true prevalence of LQTS, and 1:200, which matches the potential prevalence of LQTS and acquired QT prolongation in a tertiary referral center setting. These were performed on both the entire 12-lead ECG and derived median-beat ECGs, including or excluding patients with a wide QRS and / or pacemaker.

[0051] result

[0052] Cohort characteristics

[0053] After IRB-approved certification of study approval, a case cohort of patients with clinically and genetically confirmed LQTS (n = 1,599) was obtained from the Windland Smith Rice Genetic Heart Rhythm Clinic database, and all ECGs for these patients were abstracted from the Mayo Clinic CVMuse ECG database (Figure 2). After filtering for the 99th percentile of QTc, 808 LQTS patients were identified as having one or more ECGs with QTc values ​​above an age- and sex-defined threshold (a total of 2,987 ECGs). For the control cohort of patients with acquired long QT, the Mayo Clinic ECG Data Vault, containing over 10 million ECGs from over 2.5 million patients, was queried using the same 99th percentile threshold, identifying 361,069 patients (14%) with at least one ECG above that specific QTc threshold for a total of 989,313 ECGs (Figure 2). These cohorts were subsequently used to train, validate, and test AI-DNN models at the following case-control ratios: 1) statistically detected (1:5 and 1:10), 2) LQTS disease prevalence / population-based screening (1:2,000), and 3) clinical screening (referral clinic LQTS to QT prolongation ratio; 1:2,000).

[0054] Model development with matching ratios of 1:5 and 1:10

[0055] After filtering the 99th percentile and age-sex matching at 1:5 and 1:10, 3,529 and 5,821 controls with acquired QT prolongation were used, respectively (Figure 2). Figure 3 provides a table showing the basic demographics (sex, age at ECG, BMI) of both cohorts at a 1:5 matching ratio for the training, test, and validation sets, demonstrating that these cohorts were similar across all parameters. Figure 3 also describes the baseline characteristics of the case (LQTS) and control (acquired QT prolongation) populations (matching ratio 1:5).

[0056] After model development, as shown in Figure 4A, AI-ECG demonstrated excellent ability to distinguish LQTS patients from acquired long QT patients with an AUC of 0.896 (accuracy 85%, sensitivity 77%, specificity 87%, PPV 58%, NPV 94%; Figure 5). Similarly, when the model was run with a 1:10 matching ratio, a slight decrease in PPV was observed, but it was equally successful in distinguishing the two groups (AUC 0.896, accuracy 86%, sensitivity 79%, specificity 87%, PPV 47%, NPV 97%; Figure 4B and Figure 5). Figure 5 describes the main and subgroup analyses examining the performance of the AI-ECG-LQTS detection algorithm.

[0057] After excluding patients with wide QRS complexes and / or active ventricular pacemakers, which are known to affect accurate QT assessment, the performance of the AI-ECG algorithm remained excellent, with AUCs of 0.89 and 0.883 at matching ratios of 1:5 and 1:10, respectively (Figures 4A and 4B). Further details of the model's performance at all matching ratios can be found in Figure 5.

[0058] Internal model validation.

[0059] To verify the robustness of the model, we used two separate approaches and separate AI-ECG algorithms using the same dataset. First, rather than using rhythmic ECGs, median-beat ECGs were generated as described above. Similar to rhythmic ECGs, the median-beat model was able to successfully distinguish LQTS patients from patients with acquired QT prolongation at both 1:5 and 1:10 matching ratios (AUC 0.9 and 0.912, respectively) and after excluding wide QRS / pacemaker patients (AUC 0.912 and 0.908, respectively). Further details of the model's performance on median-beat ECGs can be found in Figure 5.

[0060] Furthermore, both models (rhythmic and median-beat ECGs) were tested using all of the patient's ECGs, not just the first ECG with QT prolongation. Again, the algorithm demonstrated robust performance characteristics at both matching ratios (1:5 and 1:10) for both rhythmic and median-beat ECGs, including patients with wide QRS and / or with or without a pacemaker, with AUCs ranging from 0.906 to 0.945. All performance characteristics are shown in Figure 6. Figure 6 illustrates validation of the AI-ECG-LQTS detection model on all ECGs of the cases. Figure 7 provides a table describing the baseline characteristics of the LQTS patient and control populations (matching ratio = 2000).

[0061] Prevalence and screening model scenarios

[0062] Although this AI-ECG QT prolongation algorithm performed well at ideal statistical matching ratios (1:5 and 1:10), these do not reflect the real-world prevalence of LQTS disease (approximately 1 in 2,000 in the general population) or the ratio of LQTS patients compared with patients with acquired QT prolongation that would be seen in a tertiary care center. Therefore, if acquired QT prolongation is present in approximately 10% of patients seen in a tertiary care referral clinic, a screening ratio of 1:200 would reflect the prevalence of LQTS versus acquired QT prolongation. Using the same LQTS cohort and control databases, 49,282 and 120,937 patients with acquired QT prolongation were used for 1:200 and 1:2,000 matching development (Figure 8). Starting with a prevalence of LQTS disease (matching ratio 1:2,000), the AI-ECG model was able to successfully distinguish LQTS from acquired QT prolongation with high AUC (0.933), accuracy (89%), sensitivity (81%), and specificity (90%), although a significant decrease in PPV was observed (5%). Similarly, after excluding patients with a wide QRS and / or pacemaker (AUC 0.948), high performance values ​​were observed for the median beat (AUC 0.952 and 0.951) and all ECG analyses (AUC range 0.930 to 0.952). All performance characteristics are summarized in Figures 3 and 5.

[0063] Finally, to mimic the situation of prolonged QT screening for patients seen at a tertiary referral center, we developed and tested AI-ECG using a matching ratio of 1:200. Here, AI-ECG had an excellent discriminatory AUC of 0.921 in separating LQTS patients from acquired prolonged QT patients (accuracy 88%, sensitivity 78%, specificity 88%), but again, a much lower PPV of 10% was observed (NPV 99.6%). These performance metrics were tracked when including all models, filters, and all ECGs at the 1:200 ratio (Figures 3 and 5).

[0064] Consideration

[0065] Because of the risk of developing potentially fatal ventricular arrhythmias, early and accurate identification of not only a prolonged QTc on a patient's ECG, but also the underlying cause of that QT prolongation, is paramount. While the standard 12-lead is an easy and affordable screening tool to potentially identify individuals with a prolonged QTc, accurate QT measurement, and more importantly, identifying the cause of QT prolongation, remains challenging.

[0066] In clinical practice, QT prolongation is not only associated with LQTS (prevalence of 1 in 2,000 individuals), but is also frequently seen in patients with electrolyte disorders, particularly hypokalemia and hypomagnesemia, various cardiac and noncardiac conditions that may be QT prolonging, or with drugs that prolong the QT interval, either as an intended part of the drug's mechanism of action (i.e., dofetilide) or as an unwanted side effect of noncardiac medications (www.crediblemeds.org). More importantly, QT prolongation unrelated to gene-mediated LQTS is i) much more common (>10% of patients evaluated in tertiary referral clinics) and ii) its cause can often be addressed and corrected. This poses a major dilemma in clinical practice and opens the door to examining the potential of AI-ECG.

[0067] In this study, derived from one of the largest cohorts of LQTS patients evaluated at a single LQTS specialty clinic, combined with our institution's extensive historical ECG database and our growing cardiovascular experience in developing AI-ECG algorithms, we developed a novel AI-ECG algorithm that can successfully distinguish QT prolongation caused by LQTS-causing mutations from QT prolongation caused by other underlying conditions with an AUC > 0.9, regardless of matching ratio, presence or absence of ventricular pacemaker or bundle branch block, or type of ECG data used (rhythm strip vs. median beat).

[0068] More importantly, although we observed an expected decrease in PPV, the accuracy, sensitivity, and specificity remained high after we created a model representing the screening situation, which will be important for the use of the model in clinical practice. For this model, we simulated two possible screening scenarios: matching cases to represent either the prevalence of LQTS in the general population (1:2,000 matching ratio) or the incidence of LQTS to QT prolongation in clinical practice (1:200). By applying these matching ratios, this algorithm can potentially be used in clinical practice to assist providers in identifying QT prolongation, thereby allowing the tool to serve as a de facto LQTS mutation detector when referring to LQTS.

[0069] Importantly, external validation of this AI-derived tool on independent cohorts from different centers and / or populations further supports its validity, while the fact that the model's performance metrics remain similar using various approaches (i.e., different matching ratios, single versus multiple ECGs, or the use of median beat analysis strategies) demonstrates the internal robustness of the developed model, enabling possible implementation in ECG laboratories, electronic medical records, and potentially mobile (monitoring) technologies. When combined with our previously developed algorithms, multiple applications can be envisioned, ranging from simple comments added to ECG reports to automated alerts sent to providers to alert them to the presence of QT prolongation along with accurate AI-derived QTc values ​​and an indication of the underlying cause (LQTS versus acquired QT prolongation).

[0070] conclusion

[0071] For patients with QTc >99th percentile, this novel AI-DNN serves as an LQTS mutation detector that can identify patients with abnormal QT prolongation secondary to LQTS-causing mutations, rather than acquired QT prolongation, with a positive predictive value of up to 50%. This algorithm can facilitate screening for this potentially fatal but highly treatable genetic heart disease and guide appropriate management of the underlying condition.

[0072] Drawing Legend

[0073] Figure 2: Study cohort flowchart. Study cohort flowchart for the 1:5 and 1:10 matching models of the long QT algorithm. Following queries of the Windland Smith Rice Genetic Heart Rhythm Clinic database (left) and the Mayo Clinic ECG Data Vault, patients with >1 ECG with a QTc above the 99th percentile were identified and matched at a 1:5 or 1:10 ratio of LQTS to "acquired long QT" patients. Cohorts were subsequently created by excluding patients with pacemakers and / or wide QRS complexes (QRS >150 ms).

[0074] Figures 4A and 4B: ROC curves of the first ECG with and without pacemaker / QRS filter (matching ratios 1:5 and 1:10). Receiver-operator characteristic curves of the QT algorithm showing excellent performance in distinguishing LQTS from acquired QT prolongation in patients with and without pacemaker and / or wide QRS complexes (BBB, bundle branch block, QRS >150 ms) at a ratio of 1:5 (Figure 4A) or 1:10 (Figure 4B).

[0075] Figure 8. Flowchart of the study cohort (matching ratios 1:200 and 1:2,000). Flowchart of the study cohort for the 1:200 and 1:2,000 matching models of the QT prolongation algorithm. To examine the model's performance at the "screening" or "LQTS prevalence" level, LQTS was matched for patients with acquired QT prolongation. Similar to previous analyses, cohorts were created with and without pacemakers and / or wide QRS complexes.

[0076] Figure 9: ROC of the first ECG with and without pacemaker / QRS filter (matching ratio 1:2,000). Receiver-operator characteristic curves of the QT algorithm using a matching ratio equal to the LQTS disease prevalence in the population (1:2,000) showing the model's performance in distinguishing LQTS from acquired long QT in patients with and without a pacemaker and / or wide QRS complexes (BBB, bundle branch block, QRS >150 ms).

[0077] Figure 10: ROC of the first ECG with and without pacemaker / QRS filter (matching ratio 1:200). Receiver-operator characteristic curve of the QT algorithm using a matching ratio equal to the prevalence of LQTS disease versus the incidence of QT prolongation seen in tertiary referral clinics (10%), resulting in a matching ratio of 1:200. At this ratio, the model showed excellent performance in distinguishing LQTS from acquired QT prolongation, both in patients with and without pacemakers and / or wide QRS complexes (BBB, bundle branch block, QRS >150 ms).

[0078] 11 , an exemplary embodiment of a machine learning module 1100 capable of performing one or more machine learning processes described in this disclosure is shown. The machine learning module can use the machine learning processes to perform the determination, classification, and / or analysis steps, methods, processes, etc., as described in this disclosure. As used in this disclosure, a “machine learning process” is a process that automatically uses training data 1104 to generate an algorithm instantiated in hardware or software logic, data structures, and / or functions executed by a computing device / module to generate output 1108 that generates given data provided as input 1112, as opposed to a non-machine learning software program in which the commands to be executed are predetermined by a user and written in a programming language.

[0079] 11 , as used herein, “training data” refers to data containing correlations that a machine learning process can use to model relationships between two or more categories of data elements. For example, without limitation, training data 1104 can include multiple data entries, also known as “training examples,” each representing a set of data elements recorded, received, and / or generated together, where the data elements may be correlated by shared presence in a given data entry, proximity in a given data entry, etc. The multiple data entries in training data 1104 can estimate one or more trends in correlations between categories of data elements, for example, without limitation, that higher values ​​of a first data element belonging to a first category of data elements tend to correlate with higher values ​​of a second data element belonging to a second category of data elements, indicating a possible proportionality or other mathematical relationship linking values ​​belonging to the two categories. Multiple categories of data elements can be associated with the training data 1104 according to various correlations, which can indicate causal and / or predictive links between the categories of data elements, which can be modeled as relationships, such as mathematical relationships, by a machine learning process, as described in further detail below. The training data 1104 can be formatted and / or organized by categories of data elements, for example, by associating the data elements with one or more descriptors that correspond to the categories of the data element. As a non-limiting example, the training data 1104 can include data entered in a standardized format by a person or process, such that the entry of a given data element in a given field in a form can be mapped to one or more descriptors of the category.Elements in the training data 1104 may be linked to category descriptors by tags, tokens, or other data elements; for example, but not limited to, the training data 1104 may be provided in a fixed-length format, a format that links the location of data to categories such as Comma-Separated Value (CSV) format, and / or a self-describing format such as Extensible Markup Language (XML), JavaScript Object Notation (JSON), etc., that allows a process or device to discover the category of the data.

[0080] Alternatively or additionally, and continuing to refer to FIG. 11 , the training data 1104 can include one or more elements that are uncategorized. That is, the training data 1104 may be unformatted or may not include descriptors for some elements of the data. Machine learning algorithms and / or other processes can sort the training data 1104 according to one or more classifications, for example, using natural language processing algorithms, tokenization, detecting correlation values ​​in the raw data, etc., and categories can be generated using correlation and / or other processing algorithms. As a non-limiting example, in a corpus of text, phrases comprising a number "n" of compounds, such as nouns modified by other nouns, can be identified according to the statistically significant prevalence of n-grams containing such words in a particular order. Such n-grams can be classified as elements of language, such as "words," that are tracked similarly to single words, and new categories can be generated as a result of statistical analysis. Similarly, in a data entry that includes some text data, a person's name may be identified by reference to a list, dictionary, or other compendium of terms, allowing for ad-hoc classification by a machine learning algorithm and / or automated association of data in the data entry with a descriptor or a given format. The ability to automatically classify data entries allows the same training data 1104 to be applied to two or more different machine learning algorithms, as described in further detail below. The training data 1104 used by the machine learning module 1100 can correlate any input data, as described in this disclosure, to any output data, as described in this disclosure. As a non-limiting illustrative example, the input may include ECG data 128 and the output may include LQTS determination data 148.

[0081] 11 , as described in more detail below, one or more supervised and / or unsupervised machine learning processes and / or models can be used to filter, sort, and / or select the training data, and such models can include, but are not limited to, a training data classifier 1116. The training data classifier 1116 can include a “classifier,” as used in this disclosure, which is a machine learning model as defined below, such as a data structure representing and / or using a mathematical model, neural net, or program generated by a machine learning algorithm known as a “classification algorithm,” that classifies inputs into categories or bins of data and outputs the categories or bins of data and / or their associated labels, as described in more detail below. The classifier can be configured to output at least data that labels or identifies data sets that are clustered together, that are found to be close under a distance metric, as described below. The distance metric can include any norm, such as, but not limited to, the Pythagorean norm. The machine learning module 1100 can generate a classifier using a classification algorithm, which is defined as a process by which a computing device and / or any modules and / or components operating thereon derive a classifier from training data 1104. Classification can be performed using, but is not limited to, linear classifiers such as logistic regression and / or naive Bayes classifiers, nearest neighbor classifiers such as k-nearest neighbor classifiers, support vector machines, least squares support vector machines, Fisher's linear discriminant, quadratic classifiers, decision trees, boosted trees, random forest classifiers, learning vector quantization, and / or neural network-based classifiers. As a non-limiting example, the training data classifier 1116 can classify elements of the training data according to subject demographics.

[0082] With further reference to FIG. 11 , training examples for use as training data may be selected from a population of potential examples according to a cohort relevant to the analytical problem or classification task to be solved, etc. Alternatively, or additionally, training data may be selected to span a set of possible situations or inputs for the machine learning model and / or process that will be encountered upon deployment. For example, for each category of input data to a machine learning process or model that may lie within a range of values ​​within a collection of phenomena, such as, but not limited to, images, user data, process data, physical data, etc., the computing device, processor, and / or machine learning model may select training examples that represent each possible value in such range and / or a representative sample of values ​​in such range. Selecting representative samples may include, for example, selecting training examples in a proportion consistent with a statistically determined and / or predicted distribution of such values ​​according to relative frequency, such that more frequently encountered values ​​in the population of data so analyzed are represented by more training examples than less frequently encountered values. Alternatively or additionally, the set of training examples may be compared to a set of representative values ​​in a database and / or presented to a user, so that the process can detect, automatically or via user input, one or more values ​​not included in the set of training examples. A computing device, processor, and / or module can automatically generate missing training examples by receiving and / or retrieving missing input and / or output values ​​and correlating the missing input and / or output values ​​with corresponding output and / or input values ​​co-located in the data record with the retrieved values, provided by a user, other device, etc.

[0083] With further reference to FIG. 11 , the computer, processor, and / or module may be configured to sanitize the training data. As used in this disclosure, "sanitizing" training data is a process in which training examples that prevent a machine learning model from converging and / or processing to a useful result are removed. For example, without limitation, the training examples may include input and / or output values ​​that deviate from commonly encountered values ​​so that the machine learning algorithm using the training examples is adapted to less likely quantities as inputs and / or outputs. For example, values ​​that are more than a threshold number of standard deviations from the mean, average, or expected value may be eliminated. Alternatively or additionally, one or more training examples may be identified as having low-quality data, where "low quality" is defined as having a signal-to-noise ratio below a threshold.

[0084] As a non-limiting example, and with further reference to FIG. 11 , images used to train an image classifier or other machine learning model, and / or processes that take images as input or generate images as output, may be rejected if the image quality is below a threshold. For example, but not by way of limitation, a computing device, processor, and / or module may perform blur detection, where rejecting one or more blur detections may be performed by, by way of non-limiting example, performing a Fourier transform or approximation, such as a fast Fourier transform (FFT), of the image and analyzing the distribution of low and high frequencies in the resulting frequency-domain representation of the image, where the number of high-frequency values ​​below a threshold level may indicate blur. As a further non-limiting example, blur detection may be performed by convolving the image or a channel of the image, etc., with a Laplacian kernel. This may generate a numerical score that reflects some rapid changes in intensity present in the image, such that a high score indicates clarity and a low score indicates blur. Blur detection can be performed using gradient-based operators, which measure the gradient or first derivative of an image based on the hypothesis that abrupt changes indicate sharp edges in the image and therefore lower blur levels. Blur detection can be performed using wavelet-based operators that utilize the ability of discrete wavelet transform coefficients to describe the frequency and spatial content of an image. Blur detection may also be performed using statistics-based operators that utilize some image statistics as texture descriptors to calculate the focus level. Blur detection can be performed by using discrete cosine transform (DCT) coefficients to calculate the focus level of an image from its frequency content.

[0085] With continued reference to FIG. 11 , a computing device, processor, and / or module can be configured to be preconditioned on one or more training examples. For example, but not by way of limitation, if a machine learning model and / or process has one or more inputs and / or outputs that require, transmit, or receive a particular number of bits, samples, or other data units, the elements of one or more training examples used as or compared to the inputs and / or outputs can be modified to have such a number of data units. For example, a computing device, processor, and / or module can convert a smaller number of units, such as in a low pixel count image, to a desired number of units, e.g., by upsampling and interpolation. As a non-limiting example, a low pixel count image may have 100 pixels, but the desired number of pixels may be 128. A processor can interpolate the low pixel count image to convert the 100 pixels to 128 pixels. It should also be noted that, upon reading this disclosure, one of ordinary skill in the art will know various methods for interpolating a smaller number of data units, such as samples, pixels, bits, etc., to a desired number of such units. In some cases, the set of interpolation rules may be trained with highly detailed inputs and / or outputs and corresponding sets of inputs and / or outputs, downsampled to a smaller number of units, and a neural network or other machine learning model trained to predict interpolated pixel values ​​using the training data. As a non-limiting example, sample inputs and / or outputs, such as a sample picture having sample augmented data units (e.g., pixels added between original pixels), may be input to a neural network or machine learning model, which outputs a pseudo-replica sample picture having dummy values ​​assigned to pixels between the original pixels based on the set of interpolation rules.As a non-limiting example, in the context of an image classifier, a machine learning model may have a set of interpolation rules trained with highly detailed images and sets of images downsampled to a smaller number of pixels, and a neural network or other machine learning model trained using those examples to predict interpolated pixel values ​​in a facial image context. As a result, inputs with sample-expanded data units (those added between original data units, dummy values) may be run through the trained neural network and / or model, which may fill in values ​​to replace the dummy values. Alternatively or additionally, the processor, computing device, and / or module may utilize a sample expander method, a low-pass filter, or both. As used in this disclosure, a "low-pass filter" is a filter that passes signals with frequencies below a selected cutoff frequency and attenuates signals with frequencies above the cutoff frequency. The exact frequency response of the filter depends on the filter design. The computing device, processor, and / or module may fill in data units between original data units using averaging, such as luma or chroma averaging within the image.

[0086] In some embodiments, with continued reference to FIG. 11 , a computing device, processor, and / or module can downsample elements of a training example to a desired smaller number of data elements. As a non-limiting example, a high pixel count image may have 256 pixels, but the desired number of pixels may be 128. The processor can downsample the high pixel count image to convert the 256 pixels to 128 pixels. In some embodiments, the processor can be configured to perform downsampling on the data. Downsampling, also known as decimation, can involve removing every Nth entry, all but the Nth entry, etc., in a sequence of samples, a process known as “compression,” which can be performed, for example, by an N-sample compressor implemented using hardware or software. Anti-aliasing and / or anti-imaging filters, and / or low-pass filters can be used to remove compression artifacts.

[0087] 11 , the machine learning module 1100 may be configured to execute a lazy learning process 1120 and / or protocol, which may alternatively be referred to as a “lazy loading” or “call on demand” process and / or protocol, in which machine learning is performed upon receiving inputs that are converted into outputs by combining the inputs and training set and deriving an algorithm used to generate outputs on demand. For example, an initial set of simulations may be run to cover initial heuristics and / or “first guesses” at outputs and / or relationships. As a non-limiting example, the initial heuristic may include ranking associations between inputs and elements of the training data 1104. The heuristic may include selecting several highest-ranking associations and / or training data 1104 elements. The lazy learning may implement any suitable lazy learning algorithm, including, but not limited to, a K-nearest neighbor algorithm, a lazy naive Bayes algorithm, etc. Upon reviewing this disclosure in its entirety, those skilled in the art will recognize a variety of lazy learning algorithms that may be applied to generate outputs as described in this disclosure, including, but not limited to, lazy learning applications of machine learning algorithms as described in more detail below.

[0088] Alternatively or additionally, and continuing to refer to FIG. 11 , the machine learning model 1124 can be generated using a machine learning process such as those described herein. As used herein, a “machine learning model” is a data structure that represents and / or instantiates a mathematical and / or algorithmic representation of a relationship between inputs and outputs, generated using any machine learning process, including but not limited to any of those processes described above, and stored in memory; once the inputs are submitted to the machine learning model 1124, the machine learning model generates an output based on the derived relationship. For example, without limitation, a linear regression model generated using a linear regression algorithm may calculate a linear combination of the input data using coefficients derived during the machine learning process to calculate the output data. As a further non-limiting example, the machine learning model 1124 may be generated by creating an artificial neural network, such as a convolutional neural network, that includes an input layer of nodes, one or more hidden layers, and an output layer of nodes. Connections between nodes can be created through a process of "training" the network, where elements from a training data 1104 set are applied to the input nodes, and then an appropriate training algorithm (e.g., Levenberg-Marquardt, conjugate gradient, simulated annealing, or other algorithm) is used to adjust the connections and weights between nodes in adjacent layers of the neural network to produce desired values ​​at the output nodes. This process is sometimes referred to as deep learning.

[0089] 11 , the machine learning algorithm can include at least a supervised machine learning process 1128. As defined herein, at least a supervised machine learning process 1128 includes an algorithm that receives a training set relating some inputs to some outputs and attempts to generate one or more data structures that represent and / or instantiate one or more mathematical relationships relating the inputs to the outputs, each of the one or more mathematical relationships being optimal according to some criteria specified to the algorithm using some scoring function. For example, the supervised learning algorithm can include ECG data as described above as input, LQTS data as output, and a scoring function that represents a desired type of relationship to be found between the input and the output, where the scoring function can, for example, maximize the probability that a given input and / or combination of component inputs is associated with a given output and minimize the probability that a given input is not associated with a given output. The scoring function can be expressed as a risk function that represents the "expected loss" of the algorithm with respect to the input to the output, where the loss is calculated as an error function that represents the degree to which the prediction produced by the relationship is inaccurate when compared to a given input-output pair provided in the training data 1104. Those skilled in the art will recognize, upon reviewing this disclosure in its entirety, various possible variations of at least a supervised machine learning process 1128 that can be used to determine the relationship between inputs and outputs. A supervised machine learning process can include a classification algorithm as defined above.

[0090] With further reference to FIG. 11 , training a supervised machine learning process can include iteratively updating coefficients, biases, and weights based on, but not limited to, an error function, an expected loss, and / or a risk function. For example, outputs generated by a supervised machine learning model using example inputs in the training examples may be compared with example outputs from the training examples. An error function can be generated based on the comparison, which can include any error function suitable for use in any machine learning algorithm described in this disclosure, including, for example, the squared difference between one or more sets of comparison values. Such an error function can be used to update one or more weights, biases, coefficients, or other parameters of the machine learning model via any suitable process, including, but not limited to, a gradient descent process, a least-squares process, and / or other processes described in this disclosure. This may be done iteratively and / or recursively to gradually adjust such weights, biases, coefficients, or other parameters. The updating can be performed using one or more backpropagation algorithms in a neural network. The iterative and / or recursive updating of weights, biases, coefficients, or other parameters as described above can be performed until the currently available training data is exhausted and / or a convergence check is passed, where a "convergence check" is a test of a condition selected as indicating that the model and / or its weights, biases, coefficients, or other parameters have reached a certain degree of accuracy. The convergence check can, for example, compare the difference between two or more consecutive error or error function values, with a difference below a threshold amount being taken to indicate convergence. Alternatively or additionally, one or more error and / or error function values ​​evaluated in a training iteration can be compared to a threshold.

[0091] With further reference to FIG. 11 , a computing device, processor, and / or module may be configured to perform the methods, method steps, series of method steps, and / or algorithms described with reference to this figure in any order and at any iteration. For example, a computing device, processor, and / or module may be configured to repeatedly execute a single step, sequence, and / or algorithm until a desired or directed result is achieved. The step or series of step repetitions may be performed iteratively and / or recursively using the output of a previous iteration as input to a subsequent iteration, aggregating the inputs and / or outputs of the iterations to generate an aggregate result, decreasing or decrementing one or more variables, such as global variables, and / or dividing a larger processing task into a set of iteratively addressed smaller processing tasks. The computing device, processor, and / or module may execute any step, series of steps, or algorithm in parallel, such as performing two or more steps simultaneously and / or substantially simultaneously using two or more parallel threads, processor cores, etc. The division of tasks among parallel threads and / or processes may be performed according to any protocol suitable for dividing tasks among iterations. Those skilled in the art will recognize, upon reviewing this disclosure in its entirety, various ways in which processes, sequences of processes, processing tasks, and / or data can be subdivided, shared, or otherwise processed using iterative, recursive, and / or parallel processing.

[0092] 11 , the machine learning process can include at least an unsupervised machine learning process 1132. As used herein, an unsupervised machine learning process is a process that draws inferences within a dataset without regard to labels. As a result, the unsupervised machine learning process is free to discover any structure, relationships, and / or correlations provided in the data. The unsupervised process 1132 may not require a response variable and can be used to find patterns of interest and / or inferences between variables, determine the degree of correlation between two or more variables, etc.

[0093] Further referring to FIG. 11 , the machine learning module 1100 can be designed and configured to create the machine learning model 1124 using techniques for developing linear regression models. The linear regression model can include ordinary least squares regression, which aims to minimize the square of the difference between predicted and actual results according to an appropriate norm for measuring such difference (e.g., a vector space distance norm). To improve the minimization, the coefficients of the resulting linear equation can be modified. The linear regression model can include a ridge regression method, in which the function to be minimized includes a least-squares function and a term that multiplies the square of each coefficient by a scalar to penalize large coefficients. The linear regression model can include a least absolute shrinkage and selection operator (LASSO) model, in which ridge regression is combined with multiplying the least-squares term by a coefficient equal to 1 divided by twice the number of samples. The linear regression model can include a multitask lasso model, in which the norm applied to the least-squares term of the lasso model is the Frobenius norm, which is equivalent to the square root of the sum of the squares of all terms. The linear regression model may include an elastic net model, a multitask elastic net model, a minimum angle regression model, a LARS lasso model, an orthogonal matching pursuit model, a Bayesian regression model, a logistic regression model, a stochastic gradient descent model, a perceptron model, a passive aggression algorithm, a robustness regression model, a Huber regression model, or any other suitable model that may occur to one of ordinary skill in the art upon reviewing this disclosure in its entirety. In one embodiment, the linear regression model may be generalized to a polynomial regression model, whereby a polynomial (e.g., quadratic, cubic, or higher order equation) that provides the best predicted output / actual output fit is found. As will be apparent to one of ordinary skill in the art upon reviewing this disclosure in its entirety, methods similar to those described above can be applied to minimize the error function.

[0094] Continuing with reference to FIG. 11 , the machine learning algorithm may include, but is not limited to, linear discriminant analysis. The machine learning algorithm may include quadratic discriminant analysis. The machine learning algorithm may include kernel ridge regression. The machine learning algorithm may include support vector machines (including, but not limited to, support vector classification-based regression processes). The machine learning algorithm may include stochastic gradient descent algorithms, including classification and regression algorithms based on stochastic gradient descent. The machine learning algorithm may include nearest neighbor algorithms. The machine learning algorithm may include various forms of latent space regularization, such as variational regularization. The machine learning algorithm may include Gaussian processes, such as Gaussian process regression. The machine learning algorithm may include cross-decomposition algorithms, including partial least squares and / or canonical correlation analysis. The machine learning algorithm may include naive Bayes methods. The machine learning algorithm may include decision tree-based algorithms, such as decision tree classification or regression algorithms. The machine learning algorithm may include ensemble methods, such as bagging meta-estimators, forests of randomized trees, AdaBoost, gradient tree boosting, and / or voting classifier methods. The machine learning algorithms may include neural net algorithms, including convolutional neural net processes.

[0095] 11 , the machine learning model and / or process can be deployed or instantiated by incorporating it into a program, device, system, and / or module. For example, without limitation, the machine learning model, neural network, and / or some or all of their parameters can be stored and / or deployed in any memory or circuit. Parameters such as coefficients, weights, and / or biases can be stored as circuit-based constants, such as arrays of wires and / or binary inputs and / or outputs set to logic “1” and “0” voltage levels in a logic circuit to represent numbers according to any suitable encoding system, including two's complement, or can be stored in any volatile and / or non-volatile memory. Similarly, mathematical operations and data inputs and / or outputs to or from the model, neural network layer, etc. can be instantiated in hardware circuits and / or in the form of instructions in firmware, machine code such as binary opcode instructions, assembly language, or any higher-level programming language. Any technique for hardware and / or software instantiation of memory, instructions, data structures, and / or algorithms may be used to instantiate the machine learning process and / or model, including, but not limited to, the manufacture and / or configuration of non-reconfigurable hardware elements, circuits, and / or modules such as ASICs; the manufacture and / or configuration of reconfigurable hardware elements, circuits, and / or modules such as, but not limited to, FPGAs; non-reconfigurable and / or configurable non-rewriteable memory elements, circuits, and / or modules such as, but not limited to, rewriteable ROMs; reconfigurable and / or rewriteable memory elements, circuits, and / or modules such as, but not limited to, other memory technologies described in this disclosure; and / or any combination of the manufacture and / or configuration of any computing device and / or components thereof described in this disclosure.Such deployed and / or instantiated machine learning models and / or algorithms may receive inputs from, and generate outputs to, any other processes, modules, and / or components described in this disclosure.

[0096] 11 , any process of training, retraining, deployment, and / or instantiation of any machine learning model and / or algorithm can be performed and / or repeated after initial deployment and / or instantiation to modify, refine, and / or improve the machine learning model and / or algorithm. Such retraining, deployment, and / or instantiation can be performed as a periodic or periodic process, such as retraining, deployment, and / or instantiation at regular elapsed periods, after some measure of quantity such as the number of bytes or other measure of data processed, the number of uses or runs of the processes described in this disclosure, and / or according to a software, firmware, or other update schedule. Alternatively or additionally, retraining, deployment, and / or instantiation can be event-based, triggered by, without limitation, user input indicating suboptimal or other problematic performance, and / or by an automated on-site inspection and / or audit process, and the output of the machine learning model and / or algorithm, and / or its error and / or error function, can be compared to any threshold, convergence check, etc., and / or the output of the processes described herein can be compared to similar thresholds, convergence checks, etc. Event-based retraining, deployment, and / or instantiation may alternatively or additionally be triggered by the receipt and / or generation of one or more new training examples, which may be compared to a preconfigured threshold, and exceeding the preconfigured threshold may trigger retraining, deployment, and / or instantiation.

[0097] 11 , retraining and / or additional training can be performed using any of the processes for training described above, using any current or previously deployed version of the machine learning model and / or algorithm as a starting point. Training data for retraining may be collected, preprocessed, screened, classified, sanitized, or otherwise processed according to any of the processes described in this disclosure. Training data can include, but is not limited to, training examples including inputs and correlation outputs used, received, and / or generated from any version of any system, module, machine learning model or algorithm, device, and / or method described in this disclosure. Such examples can be modified and / or labeled according to user feedback or other processes to indicate desired results and / or can have actual or measured results from the process being modeled and / or predicted by the system, module, machine learning model or algorithm, device, and / or method as “desired” results to be compared with the output of the training process as described above.

[0098] The rearrangement may be performed using any reconfiguration and / or rewriting of reconfigurable and / or rewritable circuit and / or memory elements, or may be performed by the generation of new hardware and / or software components, circuits, instructions, etc., which may be added to and / or replace existing hardware and / or software components, circuits, instructions, etc.

[0099] 11 , one or more processes or algorithms described above may be performed by at least a dedicated hardware unit 1136. For purposes of this figure, a “dedicated hardware unit” is a hardware component, circuitry, etc., other than a main control circuit and / or processor, that performs the method steps described in the present disclosure, that is specifically designated or selected to perform one or more particular tasks and / or processes described with reference to this figure, such as, but not limited to, preconditioning and / or sanitizing training data and / or training a machine learning algorithm and / or model. The dedicated hardware unit 1136 may include, but is not limited to, hardware units that can efficiently use pipelining, parallel processing, etc. to perform iterative or massive calculations, such as matrix-based calculations for updating or adjusting parameters, weights, coefficients, and / or biases of a machine learning model and / or neural network; such hardware units may be optimized for such processes by including dedicated circuitry for matrix and / or signal processing operations, for example, including multiple arithmetic and / or logic circuit units, such as multipliers and / or adders, that can operate simultaneously and / or in parallel. Such special-purpose hardware units 1136 may include, but are not limited to, graphical processing units (GPUs), special-purpose signal processing modules, FPGAs, or other reconfigurable hardware configured to instantiate parallel processing units for one or more specific tasks, and the computing device, processor, apparatus, or module may be configured to instruct one or more special-purpose hardware units 1136 to perform one or more operations described herein, such as evaluation of model and / or algorithm outputs, one-time or iterative updates to parameters, coefficients, weights, and / or biases, and / or any other operations, such as vector and / or matrix operations, described in this disclosure.

[0100] Referring now to FIG. 12, an exemplary embodiment of a neural network 1200 is shown. A neural network 1200, also known as an artificial neural network, is a network of "nodes," or data structures having one or more inputs, one or more outputs, and a function that determines the output based on the inputs. Such nodes can be organized into networks such as, but not limited to, a convolutional neural network, including an input layer of nodes 1204, one or more hidden layers 1208, and an output layer of nodes 1212. Connections between nodes can be created through a process of "training" the network, in which elements from a training data set are applied to the input nodes, and then an appropriate training algorithm (e.g., Levenberg-Marquardt, conjugate gradient, simulated annealing, or other algorithm) is used to adjust the connections and weights between nodes in adjacent layers of the neural network to produce desired values ​​for the output nodes. This process is sometimes referred to as deep learning. Connections may run only from input nodes to output nodes in a "feedforward" network, or may feed the output of one layer back into the input of the same or a different layer in a "recurrent network." As a further non-limiting example, a neural network may include a convolutional neural network that includes an input layer of nodes, one or more hidden layers, and an output layer of nodes. As used in this disclosure, a "convolutional neural network" is a neural network in which at least one hidden layer is a convolutional layer that convolves the input to that layer with a subset of the input, known as a "kernel," along with one or more additional layers, such as a pooling layer, a fully connected layer, etc.

[0101] 13, an exemplary embodiment of a neural network node 1300 is shown. The node may include multiple inputs xi, which may receive values ​​from other nodes and / or inputs to the neural network that includes the node. The node may perform one or more activation functions to generate its output given one or more inputs, including, but not limited to, computing a binary step function that compares the input to a threshold and outputs either a logic 1 or logic 0 output or the like; a linear activation function, where the output is directly proportional to the input; and / or a non-linear activation function, where the output is not proportional to the input. A non-linear activation function may be, but is not limited to, an activation function of the form x given an input x: Sigmoid function of JPEG2025135599000002.jpg1540, format JPEG2025135599000003.jpg1419 tanh (hyperbolic tangent) function, f(x)=tanh 2 tanh derivatives such as f(x), normalized linear unit functions such as f(x)=max(0,x), "leaky" and / or "parametric" normalized linear unit functions such as f(x)=max(ax,x), for some value of α (this function may be replaced and / or weighted by its own derivative in some embodiments). JPEG2025135599000004.jpg1890 and other exponential linear unit functions, when the input to the instantaneous layer is xi f(x) = x*sigmoid(x) for some values ​​of a, b, and r, which can include softmax functions such as JPEG2025135599000005.jpg1738, and swish functions such as f(x) = x*sigmoid(x). a Gaussian error linear unit function, such as JPEG2025135599000006.jpg16102, and / or JPEG2025135599000007.jpg2396. In principle, there are no restrictions on the nature of the functions of the inputs x i that can be used as activation functions. As a non-limiting illustrative example, a node may have a function for each input xi Weight w multiplied by i A weighted sum of the inputs can be performed using a weighted sum of the inputs. Additionally or alternatively, a bias b may be added to the weighted sum of the inputs such that an offset is added to each unit in the neural network layer that is independent of the input to the layer. The weighted sum can then be input to a function φ, which can generate one or more outputs y. The weights w applied to the inputs xi i can indicate whether an input is "excitatory," e.g., by having a large corresponding weight, indicating that the input has a strong influence on one or more outputs y, and / or "inhibitory," e.g., by having a small corresponding weight, indicating that the input has a weak influence on another input y. i The value of may be determined by training the neural network using training data, which may be performed using any suitable process, such as those described above.

[0102] 13, a "convolutional neural network," as used in this disclosure, is a neural network in which at least one hidden layer is a convolutional layer that convolves the input to that layer with a subset of the input known as a "kernel," along with one or more additional layers, such as a pooling layer, a fully connected layer, etc. CNN may include, but is not limited to, deep neural network (DNN) extensions, where a DNN is defined as a neural network with two or more hidden layers.

[0103] With further reference to FIG. 13 , in some embodiments, a convolutional neural network can learn from images. In a non-limiting example, the convolutional neural network can perform tasks such as image classification, detecting objects depicted in an image, image segmentation, and / or image processing. In some embodiments, the convolutional neural network can operate such that each node in the input layer is connected only to regions of nodes in the hidden layer. In some embodiments, the regions can collectively create a feature map from the input layer to the hidden layer. In some embodiments, the convolutional neural network can include layers in which all nodes have the same weights and biases. In some embodiments, this allows the convolutional neural network to detect features such as edges across different locations in an image.

[0104] 14, an exemplary embodiment of a method 1400 for identifying LQTS in a subject is shown. As described with reference to other figures, method 1400 can be implemented, including, but not limited to, one or more steps. One or more steps of method 1400 can be implemented using, but are not limited to, at least a processor.

[0105] 14 , in some embodiments, method 1400 can include receiving 1405 electrocardiogram (ECG) data. In some embodiments, the ECG data includes rhythm strip ECG data. In some embodiments, the ECG data includes median beat ECG data. In some embodiments, the ECG data includes data collected using 12 ECG leads. In some embodiments, the ECG data is associated with a subject, and the subject has a ventricular pacemaker.

[0106] 14 , in some embodiments, method 1400 can include training 1410 an LQTS determination machine learning model on a training dataset including a plurality of exemplary ECG data as inputs correlated with a plurality of exemplary LQTS data as outputs. In some embodiments, the LQTS determination machine learning model can include a neural network. In some embodiments, the LQTS determination machine learning model can include a convolutional neural network. In some embodiments, the exemplary LQTS data can include a 1:5 ratio of LQTS output to non-LQTS output. In some embodiments, the exemplary LQTS data can include a 1:200 ratio of LQTS output to non-LQTS output.

[0107] Still referring to FIG. 14, in some embodiments, the method 1400 may include a step 1415 of generating LQTS decision data as a function of the ECG data using a trained LQTS decision machine learning model.

[0108] 14 , in some embodiments, the ECG data is associated with the subject and the method further includes identifying an LQTS in the subject as a function of the LQTS determination data. In some embodiments, the ECG data is associated with the subject and the method further includes administering a therapy to the subject in response to the LQTS determination data.

[0109] 14 , in some embodiments, the LQTS deterministic machine learning model includes a neural network having a plurality of 2D convolutional stacked blocks and a final output layer that is activated using a function that assigns a probability to each class of a plurality of classes. In some embodiments, method 1400 may further include zero-padding the ECG data. In some embodiments, method 1400 may further include correcting data of ECG data recorded using the first set of leads to simulate data of the second set of leads. In some embodiments, the ECG data is received from an electronic health record database.

[0110] The system, device, method, or feature may match any system, device, method, or feature disclosed in U.S. patent application Ser. No. 18 / 592,680, filed March 1, 2024, and entitled "APPARATUS AND METHOD FOR TRAINING AN ARTIFICIAL INTELLIGENCE-SUPPORTED DIAGNOSTIC ASSESSMENT TOOL," each of which is incorporated by reference herein in its entirety. For example, system 100 may modify ECG data 128 recorded using a first leadset to simulate data from a second leadset.

[0111] It should be noted that any one or more of the aspects and embodiments described herein may be conveniently implemented using one or more machines (e.g., one or more computing devices utilized as user computing devices for electronic documents, one or more server devices such as document servers, etc.) programmed in accordance with the teachings herein, as would be apparent to one skilled in the computer arts. As would be apparent to one skilled in the art, appropriate software coding could be readily produced by skilled programmers based on the teachings of the present disclosure. The above-described aspects and implementations using software and / or software modules may also include appropriate hardware to assist in implementing the machine-executable instructions of the software and / or software modules.

[0112] Such software may be a computer program product using a machine-readable storage medium. A machine-readable storage medium may be any medium that can store and / or encode a set of instructions for execution by a machine (e.g., a computing device) and cause the machine to perform any one of the methodologies and / or embodiments described herein. Examples of machine-readable storage media include, but are not limited to, magnetic disks, optical disks (e.g., CDs, CD-Rs, DVDs, DVD-Rs, etc.), magneto-optical disks, read-only memory "ROM" devices, random-access memory "RAM" devices, magnetic cards, optical cards, solid-state memory devices, EPROMs, EEPROMs, and any combination thereof. As used herein, machine-readable medium is intended to include a single medium as well as a collection of physically separate media, such as a compact disc in combination with computer memory or a collection of one or more hard disk drives. As used herein, machine-readable storage medium does not include a transitory form of signal transmission.

[0113] Such software may also include information (e.g., data) carried in a data signal on a data carrier, such as a carrier wave. For example, machine-executable information may be contained in a data carrier signal embodied in a data carrier, the signal encoding a sequence of instructions, or portions thereof, for execution by a machine (e.g., a computing device), and any associated information (e.g., data structures and data) that cause the machine to perform any one of the methodologies and / or embodiments described herein.

[0114] Examples of computing devices include, but are not limited to, e-book reading devices, computer workstations, terminal computers, server computers, handheld devices (e.g., tablet computers, smartphones, etc.), web appliances, network routers, network switches, network bridges, any machine capable of executing a set of instructions that specify actions that the machine should take, and any combination thereof. In one example, a computing device may include and / or be included in a kiosk.

[0115] 15 illustrates a schematic diagram of one embodiment of a computing device in the exemplary form of a computer system 1500 upon which a set of instructions for causing a control system to perform any one or more of the aspects and / or methodologies of the present disclosure may be executed. It is also contemplated that multiple computing devices may be utilized to implement a specially configured set of instructions for causing one or more of the devices to perform any one or more of the aspects and / or methodologies of the present disclosure. The computer system 1500 includes a processor 1504 and a memory 1508, which communicate with each other and with other components via a bus 1512. The bus 1512 may include any of several types of bus structures, including, but not limited to, a memory bus, a memory controller, a peripheral bus, a local bus, and any combination thereof, using any of a variety of bus architectures.

[0116] The processor 1504 may include any suitable processor, such as, but not limited to, a processor incorporating logic circuitry for performing arithmetic and logical operations, such as an arithmetic logic unit (ALU), which may be coordinated with a state machine and directed by operational input from memory and / or sensors; the processor 1504 may be organized according to the Von Neumann and / or Harvard architectures, as non-limiting examples. The processor 1504 may include, incorporate, and / or be integrated into, but is not limited to, a microcontroller, a microprocessor, a digital signal processor (DSP), a field programmable gate array (FPGA), a complex programmable logic device (CPLD), a graphical processing unit (GPU), a general-purpose GPU, a tensor processing unit (TPU), an analog or mixed-signal processor, a trusted platform module (TPM), a floating-point unit (FPU), and / or a system-on-chip (SoC).

[0117] Memory 1508 may include a variety of components (e.g., machine-readable media), including, but not limited to, random-access memory components, read-only components, and any combination thereof. In one example, a basic input / output system 1516 (BIOS), containing the basic routines that help to transfer information between elements within computer system 1500, such as during start-up, may be stored in memory 1508. Memory 1508 may also include instructions (e.g., software) 1520 (e.g., stored on one or more machine-readable media) that embody any one or more of the aspects and / or methodologies of the present disclosure. In another example, memory 1508 may further include any number of program modules, including, but not limited to, an operating system, one or more application programs, other program modules, program data, and any combination thereof.

[0118] Computer system 1500 may also include storage 1524. Examples of storage (e.g., storage 1524) include, but are not limited to, hard disk drives, magnetic disk drives, optical disk drives combined with optical media, solid-state memory devices, and any combination thereof. Storage 1524 may be connected to bus 1512 by an appropriate interface (not shown). Exemplary interfaces include, but are not limited to, SCSI, Advanced Technology Attachment (ATA), Serial ATA, Universal Serial Bus (USB), IEEE 1394 (FIREWIRE®), and any combination thereof. In one example, storage 1524 (or one or more components thereof) may be removably interfaced with computer system 1500 (e.g., via an external port connector (not shown)). In particular, storage 1524 and associated machine-readable media 1528 may provide nonvolatile and / or volatile storage of machine-readable instructions, data structures, program modules, and / or other data for computer system 1500. In one example, the software 1520 may reside, completely or partially, within the machine-readable medium 1528. In another example, the software 1520 may reside, completely or partially, within the processor 1504.

[0119] Computer system 1500 may also include input devices 1532. In one example, a user of computer system 1500 may input commands and / or other information into computer system 1500 via input devices 1532. Examples of input devices 1532 include, but are not limited to, an alphanumeric input device (e.g., a keyboard), a pointing device, a joystick, a gamepad, an audio input device (e.g., a microphone, a voice response system, etc.), a cursor control device (e.g., a mouse), a touchpad, an optical scanner, a video capture device (e.g., a still camera, a video camera), a touch screen, and any combination thereof. Input devices 1532 may interface with bus 1512 via any of a variety of interfaces (not shown), including, but not limited to, a serial interface, a parallel interface, a game port, a USB interface, a FIREWIRE® interface, a direct interface to bus 1512, and any combination thereof. Input devices 1532 may include a touch screen interface, which may be part of or separate from display 1536, as described further below. The input device 1532 may be utilized as a user selection device for selecting one or more graphical representations within the graphical interface, as described above.

[0120] A user may also input commands and / or other information into computer system 1500 via storage device 1524 (e.g., a removable disk drive, a flash drive, etc.) and / or network interface device 1540. A network interface device such as network interface device 1540 may be utilized to connect computer system 1500 to one or more of various networks, such as network 1544, and one or more remote devices 1548 connected thereto. Examples of network interface devices include, but are not limited to, a network interface card (e.g., a mobile network interface card, a LAN card), a modem, and any combination thereof. Examples of networks include, but are not limited to, a wide area network (e.g., the Internet, an enterprise network), a local area network (e.g., a network associated with an office, building, campus, or other relatively small geographic space), a telephone network, a data network associated with a telephone / voice provider (e.g., a mobile communications provider data and / or voice network), a direct connection between two computing devices, and any combination thereof. A network such as network 1544 may use wired and / or wireless communication modes. In general, any network topology may be used. Information (eg, data, software 1520 , etc.) can be communicated to and from computer system 1500 via network interface device(s) 1540 .

[0121] Computer system 1500 may further include a video display adapter 1552 for communicating displayable images to a display device, such as display device 1536. Examples of display devices include, but are not limited to, a liquid crystal display (LCD), a cathode ray tube (CRT), a plasma display, a light emitting diode (LED) display, and any combination thereof. Display adapter 1552 and display device 1536 may be utilized in combination with processor 1504 to provide graphical representations of aspects of the present disclosure. In addition to a display device, computer system 1500 may include one or more other peripheral output devices, including, but not limited to, audio speakers, a printer, and any combination thereof. Such peripheral output devices may be connected to bus 1512 via peripheral interface 1556. Examples of peripheral interfaces include, but are not limited to, a serial port, a USB connection, a FIREWIRE® connection, a parallel connection, and any combination thereof.

[0122] The foregoing has been a detailed description of exemplary embodiments of the present invention. Various modifications and additions may be made without departing from the spirit and scope of the present invention. Features of each of the various embodiments described above may be combined with features of other described embodiments as needed to provide multiple feature combinations in related new embodiments. Moreover, while the above describes several separate embodiments, what has been described herein is merely illustrative of the application of the principles of the present invention. Furthermore, while certain methods herein may be illustrated and / or described as being performed in a particular order, the order may be highly variable within the skill of those skilled in the art to achieve methods, systems, apparatus, and software according to the present disclosure. Accordingly, this description is intended to be taken by way of example only, and not to limit the scope of the present invention.

[0123] Exemplary embodiments are disclosed above and shown in the accompanying drawings. Those skilled in the art will understand that various modifications, omissions, and additions can be made to what is specifically disclosed herein without departing from the spirit and scope of the present invention.

Claims

1. 1. An apparatus for generating LQTS decision data, comprising: At least a processor and a memory communicatively connected to the at least processor, the memory comprising: receiving electrocardiogram (ECG) data; training an LQTS determination machine learning model on a training dataset including a plurality of exemplary ECG data as inputs correlated with a plurality of exemplary LQTS data as outputs; instructions to configure the at least processor to use the trained LQTS determination machine learning model to generate LQTS determination data as a function of the ECG data; The apparatus, wherein the LQTS decision machine learning model includes a neural network having a plurality of 2D convolution stacked blocks and a final output layer activated using at least an activation function.

2. The apparatus of claim 1 , wherein the memory includes instructions that configure the at least one processor to zero-pad the ECG data.

3. 10. The apparatus of claim 1, wherein receiving the ECG data includes modifying data of the ECG data recorded using a first leadset to simulate data of a second leadset.

4. The device of claim 1 , wherein the ECG data includes rhythm strip ECG data.

5. The apparatus of claim 1 , wherein the ECG data comprises median beat ECG data.

6. The device of claim 1 , wherein the ECG data comprises data collected using 12 ECG leads.

7. The device of claim 1 , wherein the memory includes instructions that configure the at least one processor to receive the ECG data from an electronic health record database.

8. The apparatus of claim 1 , wherein the exemplary LQTS data includes a ratio of LQTS output to non-LQTS output of 1:

5.

9. The apparatus of claim 1 , wherein the exemplary LQTS data comprises a ratio of LQTS output to non-LQTS output of 1:

200.

10. 1. A method for generating LQTS decision data, comprising: receiving electrocardiogram (ECG) data using at least a processor; using the at least one processor to train an LQTS determination machine learning model on a training dataset including a plurality of exemplary ECG data as inputs correlated with a plurality of exemplary LQTS data as outputs; and using the at least a processor to generate LQTS decision data as a function of the ECG data using the trained LQTS decision machine learning model; The method, wherein the LQTS decision machine learning model includes a neural network having a plurality of 2D convolution stacked blocks and a final output layer activated using at least an activation function.

11. The method of claim 10 further comprising zero-padding the ECG data.

12. 11. The method of claim 10, wherein receiving the ECG data includes modifying data of the ECG data recorded using a first leadset to simulate data of a second leadset.

13. The method of claim 10 , wherein the ECG data comprises rhythm strip ECG data.

14. The method of claim 10 , wherein the ECG data comprises median beat ECG data.

15. The method of claim 10 , wherein the ECG data comprises data collected using 12 ECG leads.

16. The method of claim 10 , wherein the ECG data is received from an electronic health record database.

17. The method of claim 10 , wherein the exemplary LQTS data comprises a ratio of LQTS output to non-LQTS output of 1:

5.

18. The method of claim 10 , wherein the exemplary LQTS data comprises a ratio of LQTS output to non-LQTS output of 1:

200.

19. the ECG data is associated with a subject; 11. The method of claim 10, wherein the method further comprises administering a treatment to the subject as a function of the LQTS determination data.

20. 1. A system for generating LQTS decision data, comprising: an electrocardiogram (ECG) sensor configured to detect ECG data; a computing device communicatively connected to the ECG sensor to receive electrocardiogram (ECG) data from the ECG sensor; generating LQTS decision data as a function of the ECG data using a trained LQTS decision machine learning model; transmitting said LQTS decision data to a display; A computing device configured such that the LQTS decision machine learning model includes a neural network having a plurality of 2D convolution stacked blocks and a final output layer activated using at least an activation function; A display, receiving LQTS decision data from the computing device; a display configured to display the LQTS determination data.

Images