2-read QT interval prediction

Abstract

[0006] An embodiment of the present disclosure provides a mobile electrocardiogram (ECG) sensor including an electrode assembly including electrodes, the electrode assembly sensing a cardiac-related signal when in contact with a user's body and generating an electrical signal representing the sensed cardiac-related signal. The ECG sensor further includes a processing device operably coupled to the electrode assembly, the processing device providing the sensed cardiac-related signal to a machine learning module trained to predict a 12-lead QT interval (QTc) value from a mobile ECG sensor with fewer than 12 leads. The ECG sensor also includes a housing containing the electrode assembly and the processing device.

Description

Technical Field

[0001] Cross - reference to Related Applications This application claims the benefit of U.S. Provisional Application No. 63 / 044,882, filed June 26, 2020, entitled TWO - LEAD QT INTERVAL PREDICTION, the entire content of which is incorporated herein by reference.

Background Art

[0002] It is estimated that by 2030, more than 23 million people will die each year from cardiovascular diseases. Cardiovascular diseases are prevalent not only in third - world countries but also in the entire population of first - world countries and regardless of socioeconomic status. Monitoring cardiovascular function can be useful for the treatment and prevention of cardiovascular diseases. For example, patients with A - fib (or other types of arrhythmias) can be monitored for long periods using a Holter monitor or other portable electrocardiogram (ECG) devices to manage the disease. Such devices can continuously monitor the electrical activity of the cardiovascular system, for example, for at least 24 hours. Such monitoring can be important, among other things, for detecting conditions such as acute coronary syndrome (ACS).

[0003] The mammalian heart generates and conducts an electric current that signals and initiates the coordinated contractions of the heart. In humans, the electrical signal is generated by a part of the heart known as the SA node. After being generated by the SA node, the current spreads throughout the myocardium in a predictable way in a healthy heart.

[0004] Generally, an electrocardiogram (ECG) is a graphical representation of the electrical conduction of the heart over time projected onto the surface of the body. An ECG is typically displayed on a graph with x - and y - axes. Typically, the x - axis of an ECG represents time, and the y - axis of an ECG represents the potential (in millivolts) of the current that conducts through the heart during normal heart function.

Summary of the Invention

Means for Solving the Problems

[0005] Novel features of the present invention are specifically described in the appended claims. A better understanding of the features and advantages of the present invention will be obtained by referring to the following detailed description, which describes exemplary embodiments in which the principles of the present invention are utilized, and to the accompanying drawings. [Brief explanation of the drawing]

[0006] [Figure 1] Image representations of a prior art electrocardiograph having 10 electrodes placed on a patient's body to acquire a prior art 12-lead electrocardiogram, according to some embodiments of the present disclosure. [Figure 2] This is a chest image representation showing examples of electrode placements on the chest for acquiring a prior art 12-lead electrocardiogram according to some embodiments of the present disclosure. [Figure 3] This figure shows an exemplary lead I annotated to show a PQRST wave generated by a 12-lead electrocardiograph according to some embodiments of the present disclosure. [Figure 4] This figure shows an exemplary 12-lead electrocardiogram in a conventional format according to some embodiments of the present disclosure. [Figure 5A] This figure shows an exemplary ECG sensing device according to some embodiments of the present disclosure. [Figure 5B] Figure 5A is a hardware block diagram of an ECG sensing device according to some embodiments of the present disclosure. [Figure 5C] This figure shows the use of the ECG sensing device shown in Figure 5A for measuring a user's limb leads, according to some embodiments of the present disclosure. [Figure 5D] Figure 5A shows the use of the ECG sensing device for measuring a user's chest leads according to some embodiments of the present disclosure. [Figure 5E] Figure 5A shows the use of the ECG sensing device for measuring a user's chest leads according to some embodiments of the present disclosure. [Figure 6]This is a flowchart of a method for predicting the QT interval according to some embodiments of the present disclosure. [Figure 7] This is a block diagram of an exemplary computing device capable of performing one or more of the operations described herein, according to some embodiments of the present disclosure. [Modes for carrying out the invention]

[0007] It should be understood that, in its application, this disclosure is not limited to the details of the structures, experiments, illustrative data, and / or component arrangements described below. Embodiments of this disclosure are possible in other forms or can be implemented or carried out in various ways. It should also be understood that the terms used herein are for illustrative purposes only and should not be considered limiting.

[0008] An electrocardiogram (ECG) provides several ECG waveforms that represent the electrical activity of a person's heart. ECG monitoring devices may have a set of electrodes for recording (referred herein to as “acquiring an ECG”) these ECG waveforms of a patient’s heart. The set of electrodes may be placed at multiple locations on the patient’s skin, and the electrical signals (ECG waveforms) recorded between each pair of electrodes in the set of electrodes may be called leads. Various numbers of leads may be used to acquire an ECG, and different numbers of electrodes and combinations of electrodes may be used to form various leads. Exemplary numbers of electrodes used to acquire an ECG are 3, 5, and 12 leads.

[0009] Figure 1 is an image representation of 10 electrodes of a conventional ECG sensing device placed on a patient to acquire a standard 12-lead ECG. The electrode placed on the right arm is commonly referred to as the RA. The electrode placed on the left arm is referred to as the LA. The RA and LA electrodes are placed in the same locations on the left and right arms, and preferably, but not necessarily, near the wrist. The leg electrodes may be referred to as the RL for the right leg and the LL for the left leg. The RL and LL electrodes are placed in the same locations on the left and right legs, and preferably, but not necessarily, near the ankle.

[0010] Figure 2 shows the placement of six electrodes on the chest, labeled V1, V2, V3, V4, V5, and V6. V1 is positioned, for example, in the fourth intercostal space between rib 4 and rib 5, just to the right of the sternum. V2 is positioned, for example, in the fourth intercostal space between rib 4 and rib 5, just to the left of the sternum. V3 is positioned in the fifth intercostal space, midway between electrodes V2 and V4. V4 is positioned in the fifth intercostal space between ribs 5 and 6 on the left midclavicular line. V5 is positioned horizontally on the left anterior axillary line, similar to V4. V6 is positioned horizontally on the left midaxillary line, similar to V4 and V5.

[0011] Next, the electrocardiograph calculates and outputs three limb lead waveforms. Limb leads I, II, and III are bipolar leads with one positive and one negative electrode. Lead I is the voltage between the left arm (LA) and the right arm (RA), for example, I = LA - RA. Lead II is the voltage between the left bundle branch (LL) and the right arm (RA), for example, II = LL - RA. Lead III is the voltage between the left bundle branch (LL) and the left arm (LA), for example, III = LL - LA. Leads I, II, and III are commonly referred to as "limb leads".

[0012] A unipolar pole also has two poles, but the negative pole is a composite pole composed of signals from multiple other electrodes. In a conventional electrocardiograph for acquiring a 12-lead ECG, all leads except the limb leads are unipolar (aVR, aVL, aVF, V1, V2, V3, V4, V5, and V6). The augmented limb leads (aVR, aVL, and aVF) view the heart from different angles (or vectors) and are determined by the potential difference between one of the RA, LA, and LL and a composite consisting of two of the RA, LA, and LL. Therefore, three electrodes placed at the RA, LA, and LL simultaneously sense aVR, aVL, and aVF based on the above relationship. In other words, leads I, II, and III each require input from only two electrodes, while aVR, aVL, and aVF may require input from three electrodes placed at the RA, LA, and LL.

[0013] For example, the right augmentation vector (aVR) is a combination of left arm and left leg electrodes, with the positive electrode placed on the right arm and the negative electrode "enhancing" the signal intensity of the positive electrode on the right arm. Therefore, the right augmentation vector (aVR) is equal to RA - (LA + LL) / 2 or -(I + II) / 2. The left augmentation vector (aVL) is equal to LA - (RA + LL) / 2 or (I - II) / 2. The foot augmentation vector (aVF) is equal to LL - (RA + LA) / 2 or (II - I) / 2.

[0014] In one embodiment, the six electrodes on the patient's chest are close enough to the heart that augmentation is not required. Wilson's central terminal (often used in CT) W , V W A composite electrode called Wilson's center terminal (or symbolized as WCT) is used as the negative terminal. Wilson's center terminal is generated by connecting electrodes RA, LA, and LL together through a simple resistance network to give the average potential of the entire body that approximates the potential (i.e., zero) at infinite distance. Wilson's center terminal WCT is calculated as (RA + LA + LL) / 3.

[0015] ECG waveforms recorded by an ECG monitoring device (each corresponding to an ECG lead) may contain data corresponding to the electrical activity of a person's heart. A typical heartbeat may contain several fluctuations of potential that can be classified into waves and complexes, including P waves, QRS complexes, T waves, and U waves, as is known in the art. In other words, each ECG waveform may contain, as is known in the art, P waves, QRS complexes, T waves, and U waves. The shape and duration of these waves may relate to various characteristics of a person's heart, such as the size of the person's atria (e.g., showing atrial dilation), and may be the first source of person-specific heartbeat characteristics. Each wave or complex of multiple waveforms (i.e., QRS complexes) is associated with different phases of cardiac depolarization and repolarization. ECG waveforms can be analyzed (typically after standard filtering and "cleaning" of the signal) for various indicators useful for detecting cardiac events or conditions, such as detecting and characterizing arrhythmias. Such indicators may include, for example, the amplitude and morphology of the ECG waveform (e.g., the amplitude and morphology of the QRS complex), R-wave-ST segment and T-wave amplitude analysis, and heart rate variability (HRV).

[0016] Figure 3 shows an exemplary lead I, annotated to show the P wave, QRS wave, and T wave / complex generated by a 12-lead electrocardiograph. Typically, an ECG of a normally beating heart has predictable waveforms in each of the 12 ECG leads. ECG segments between two waves are called segments, and ECG segments between three or more waves are called intervals. For example, the ECG segment between the end of the S wave (part of the QRS complex) and the beginning of the T wave is called an ST segment, and the ECG segment between the beginning of the Q wave (part of the QRS complex) and the end of the T wave is called a QT interval.

[0017] Figure 4 shows an exemplary 12-lead electrocardiogram in a conventional format. As shown in Figure 4, for a standard ECG waveform trace, 12 ECG leads are individually displayed on the X-axis and Y-axis, where the Y-axis represents time and the X-axis represents voltage. In these traces, all 12 ECG waveforms are aligned with respect to their X-axis. That is, the P-wave, QRS-wave, and T-wave of all leads all occur simultaneously along each X-axis of their respective traces. For example, in a conventional ECG waveform trace, if a QRS complex occurs at 1 second on the X-axis in the lead I waveform trace, the QRS complex occurs at 1 second in each of the other 11 ECG waveforms (i.e., leads II, III, aVR, aVL, aVF, V1, V2, V3, V4, V5, and V6).

[0018] The standard time-aligned format enables a healthcare provider to more easily obtain information from the 12 sensed ECG waveforms. In a conventional ECG trace, time alignment is facilitated by the waveforms being sensed simultaneously by the ten electrodes of a conventional ECG that are all placed on the skin of the individual at the same time where the ECG is sensed. That is, since all 12 ECG leads of a conventional ECG are sensed simultaneously, time alignment is achieved simply by displaying all of the waveforms together on the same axis.

[0019] It should be noted that a set of three or more leads can be analyzed to derive information for generating a complete 12-lead ECG. Such a conversion can be performed using a machine learning model (e.g., neural network, deep learning techniques, etc.). The machine learning model can be trained using 12-lead ECG data corresponding to a population of individuals. The data can be preprocessed to filter the data in a manner suitable for the application before being input into the machine learning model. For example, the data can be classified according to height, gender, weight, nationality, etc. before being used to train one or more machine learning models such that the resulting one or models are fine-tuned to a particular type of individual. In a further embodiment, the machine learning model can be further trained based on the user's own ECG data to fine-tune and personalize the model to further reduce any residual composite error.

[0020] Conventional 12-lead electrocardiograms provide very useful information about an individual's heart health and condition, but conventional electrocardiograph equipment is expensive and the procedure is not normally available in areas other than hospitals and examination rooms. Thus, monitoring is not frequently performed even in first world countries and in the poorer regions of the world, electrocardiographs may not even be available.

[0021] FIG. 5A shows top and bottom views of an exemplary ECG sensing device 500 comprising a set of electrodes 502 (also referred to as an electrode assembly) according to some embodiments of the present disclosure. In some embodiments, for example, one or more capacitive electrodes are used in the ECG sensing device 500 such that the capacitive electrodes sense the potential through clothing worn on the user's body. Similarly, a conductive spray or gel can be placed on the user's body such that typical electrodes sense the potential through clothing worn on the user's body.

[0022] In one embodiment, the ECG sensing device 500 is constructed, in whole or in part, from stainless steel or some other suitable material. In one embodiment, the ECG device 500 includes an external coating such as titanium nitride or another suitable coating. Advantageously, such materials can enhance biocompatibility and optimize electrode properties.

[0023] In one embodiment, device 500, referred herein as a mobile computing device, includes all the components necessary to sense, record, and display ECG signals and analysis. In another embodiment, device 500 connects via wire or wirelessly to a separate mobile computing device (e.g., computing device 550). In such a case, device 500 may sense ECG signals and transmit the unmodified or modified signals to the mobile computing device for further analysis and / or display. In yet another embodiment, any combination of the two examples given above is possible. For example, the ECG sensing device 500 may be considered a self-contained mobile computing device capable of performing all the operations described herein, but the ECG sensing device 500 may still connect to and interact with a second mobile computing device for any appropriate purpose (offloading processing / analysis, display, etc.).

[0024] The ECG sensing device 500 may include one or more control devices and / or indicators. For example, the device 500 may include buttons, dials, etc., for selecting functions (e.g., turning ECG reading on / off, starting to transmit ECG information, etc.). The ECG sensing device 500 may further include a display for displaying the recorded ECG.

[0025] As shown in Figure 5A, the ECG sensing device 500 may include a housing 520, with two electrodes 502A and 502B positioned on the top surface of the housing 520 and a third electrode 502C positioned on the bottom surface of the housing 520. The electrodes 502 may be insulated from each other via an insulator 504 or other suitable material so that they can sense and record separate signals. In some embodiments, the electrodes 502 may consist of silver-silver chloride (or some other suitable material) electrodes. In some embodiments, the ECG sensing device 500 may include electrode connectors (not shown), such as female sockets on one end or on the side, which allow one or more ECG electrodes to be connected to the ECG sensing device 500 for use on the skin with or without adhesive (e.g., conductive gel and electrodes 502).

[0026] Figure 5B shows a hardware block diagram of an ECG sensing device 500, which may include hardware such as a processing device 505 (e.g., a processor, central processing unit (CPU)), memory 510 (e.g., random access memory (RAM), storage devices (e.g., hard disk drives (HDDs), solid-state drives (SSDs), etc.), and other hardware devices (e.g., analog-to-digital converters (ADCs), etc.). The storage device may comprise persistent storage capable of storing data. Persistent storage may be a local storage unit or a remote storage unit. Persistent storage may be a magnetic storage unit, an optical storage unit, a solid-state storage unit, an electronic storage unit (main memory), or a similar storage unit. Persistent storage may also be a monolithic / single device or a distributed set of devices. In some embodiments, the processing device 505 may comprise a dedicated ECG waveform processing and analysis chip that provides built-in read-off detection. The ECG sensing device 500 may include an ADC (not shown) (e.g., a 24-bit ADC operating at 500 Hz or higher) having a sampling frequency high enough to accurately convert the ECG waveform measured by the electrode set 502 into a digital signal for processing by the processing device 505.

[0027] The memory 510 may include a read synthesis software module 510A (hereinafter referred to as module 510A) and a QT prediction software module 510B (hereinafter referred to as module 510B). The processing device 205 may run module 207A to synthesize ECG waveforms corresponding to reads not measured by the electrodes of the ECG sensing device 500, as will be discussed in more detail herein. The processing device 505 may run module 510B to accurately predict the user's QT interval, as will be discussed in more detail herein.

[0028] The ECG sensing device 500 may further comprise a transceiver 508 that can implement any suitable protocol for wirelessly transmitting ECG data to one or more local and / or remote computing devices (e.g., computing device 550). For example, the transceiver 508 may comprise a Bluetooth chip for transmitting ECG data to a local computing device (e.g., a user's laptop or smartphone) via Bluetooth®. In other embodiments, the transceiver 508 may include (or be coupled to) a network interface device configured to communicate using a cellular data network (e.g., using GSM, GSM plus EDGE, CDMA, quad-band, or other cellular protocols) or a WiFi (e.g., 802.11 protocol) network for transmitting ECG data to a remote computing device (e.g., a physician's or healthcare provider's computing device) and / or a local computing device.

[0029] As will be discussed in more detail herein, the computing device 550 may be used to provide instructions for operating the ECG sensing device 500, or, for example, to correspond to a healthcare provider system to which the ECG data measured by the ECG sensing device 500 should be transmitted.

[0030] As shown in Figure 5C, in one practical example, the user holds the device with one or both hands so that each hand touches electrodes 502A and 502B on the ECG sensing device 500, while the left leg is in contact with electrode 502C. The ECG sensing device 500 (optionally having a separate mobile computing device) can then be used to record read I, read II, and read III, from which at least three additional reads can be determined (e.g., by running module 510A), as will be described in more detail herein. Specifically, the enhanced reads aVR, aVL, and aVF can be determined using reads I, II, and III. The user may be sitting, standing, or in any comfortable position.

[0031] Figures 5D and 5E illustrate an embodiment in which a user may also record anterior chest leads V1, V2, V3, V4, V5, and V6 using the ECG sensing device 500, as described herein. The user may hold the ECG sensing device 500 so that each of the user's hands is in contact with electrodes 502A and 502B, while the device is held against the chest so that a third electrode (e.g., 502C) is in contact with one of six anterior chest positions represented as “CP1”, “CP2”, “CP3”, “CP4”, “CP5”, and “CP6”. For example, the user may start with the ECG sensing device 500 positioned so that electrode 502C is in contact with CP1, and from there, the user may move the ECG sensing device 500 to sequentially contact each of the six electrode positions corresponding to leads V2, V3, V4, V5, and V6. In some embodiments, while a user touches electrodes 502A and 502B of the ECG sensing device 500 with their right and left hands, respectively, and simultaneously holds a third electrode (e.g., 502C) of the device 500 relative to positions on the user's chest corresponding to V1, V2, V3, V4, V5, and V6, each of the potentials sensed at the chest positions corresponding to V1, V2, V3, V4, V5, and V6 is sensed simultaneously with the potentials sensed at LA and RA. Lead I is equivalent to the potential difference between LA and RA. Therefore, in some embodiments, measuring the potential at any of the chest positions corresponding to V1, V2, V3, V4, V5, and V6 together with the potentials at the LA and RA positions is equivalent to the potential difference between the chest position and Lead I. In other words, for example, if all three electrodes of device 500 are used as described, V1 (potential at the V1 chest position) = ("CP1") - WCT (WCT = (RA + LA + LL) / 3 or (Lead I + Lead II) / 3).

[0032] The six anterior chest locations can be expressed as a composite value known as ("CP1", "CP2", "CP3", "CP4", "CP5", and "CP6") and Wilson's central terminal ("WCT"). "CP(x)" corresponds to one of the six potentials sensed at the anatomical anterior chest lead location (where "x" is the location number 1-6). For example, CP1 is the ECG measurement sensed at the location where the electrode is placed to measure V1, which is approximately in the second intercostal space just to the right of the sternum. Thus, lead V1 = CP1 - WCT.

[0033] The WCT is equal to one-third of the sum of the potentials sensed in the right upper limb, left upper limb, and left lower limb, i.e., 1 / 3(RA+LA+LL). In a standard ECG using 10 simultaneously placed electrodes, RA, LA, and LL, which determine the WCT, are sensed simultaneously with CP1, CP2, CP3, CP4, CP5, and CP6, so the WCT value is generated as soon as the anterior chest lead is sensed.

[0034] In these embodiments, the electrodes 502 are positioned and configured to simultaneously sense / calculate six limb leads, leads I, II, III, aVR, aVL, and aVF, when the user places the first electrode 502A in contact with the right upper limb, the second electrode 502B in contact with the left upper limb, and the third electrode 502C in contact with the left lower limb.

[0035] As also described herein, the ECG sensing device 500 is configured to sequentially sense six leads V1, V2, V3, V4, V5, and V6 when the user, for example, places the first electrode 502A in contact with the right upper limb, the second electrode 502B in contact with the left upper limb, and the third electrode 502C in contact with the area of ​​the user's chest corresponding to the anterior chest lead position.

[0036] In some embodiments of the ECG sensing device 500 having three electrodes as described herein, RA, LA, and LL, which determine WCT, are not sensed simultaneously with one or more anterior chest leads. That is, if one of the three electrodes of the ECG sensing device 500 is held against the user's chest wall, only two electrodes remain free and cannot simultaneously determine the conventional WCT. In some of these embodiments, RA is set to 0. When RA=0, we provide WCT=(0+LA+LL) / 3 or ((LA-0)+(LL-0)) / 3, which can further be expressed as WCT=(Lead I+Lead II) / 3.

[0037] Similarly, in these embodiments where RA is set to 0, the average WCT = (average read I + average read II) / 3. In some embodiments, the average WCT is generated using average read I and average read II generated using an ensemble averaging method for read I and read II waveforms sensed by the ECG sensing device described herein. Generating an average WCT is beneficial, for example, for signal filtering and also simplifies the alignment of values ​​for subtraction purposes. That is, in some embodiments, CP1, CP2, CP3, CP4, CP5, and CP6 are each averaged, and the average WCT is subtracted from each of them to generate V1, V2, V3, V4, V5, and V6, respectively.

[0038] Several machine learning (ML) methods may also be used to synthesize a complete set of 12 reads from a set of reads measured by the ECG sensing device 500. ML is well-suited to continuous monitoring of one or more criteria for identifying anomalies or trends in input data, regardless of scale, compared to training examples used to train a model. The ML models described herein may be trained on user data from a population of users and / or on other training examples to suit design needs for the model. ML models that may be used in the embodiments described herein include, but are not limited to, Bayesian, Markov, Gaussian processes, clustering algorithms, generative models, kernels, and neural network algorithms. Some embodiments utilize machine learning models based on trained neural networks (e.g., trained recurrent neural networks (RNNs) or trained convolutional neural networks (CNNs)).

[0039] For example, ML models can utilize artificial neural networks (ANNs) for supervised classification, and the model's results represent the probability that an input sample belongs to a particular class of data or exhibit some specific characteristics. In another example, data-driven methods based on convolutional neural networks (CNNs) are used. By using convolutional operations, ML models can consider temporally closed correlations between input samples to infer a single output data point. More specifically, a single output sample (each chest read) at a typical time t is influenced by all input samples (all limb reads) from t-τ to t+τ. The value of τ, which represents the receptive field of the network, is highly dependent on the model architecture and typically increases with its depth, i.e., the number of consecutive layers. The ability to generalize to unknown data and avoid overfitting problems is paramount for all data-driven methods. Complex models, with small datasets, can lead to excellent performance on training sets but may perform poorly on unknown data. To optimize the model, any appropriate normalization method can be used, such as inter-layer and intra-layer normalization (e.g., batch normalization and layer normalization) and data augmentation techniques. Finally, to improve the effectiveness and efficiency of the model, the use of residual connections, i.e., identity mapping which allows gradients to pass through layers during backpropagation of gradient-based optimization algorithms, may be utilized.

[0040] The use of AI / deep learning in multi-read ECG sensing devices could enable patients to monitor their own cardiac electrical activity (in the hospital or at home) without needing to visit a hospital or use bulky hardware.

[0041] In some embodiments, the memory 510 of the ECG sensing device 500 or other mobile computing device (e.g., computing device 550) may include an instruction software module (not shown) that displays or otherwise communicates instructions to the individual, instructing the user on how to position the ECG sensing device 500 to perform an ECG (e.g., on a standard anterior chest lead chest position), and instructing the user on where the user should position themselves to perform an ECG. For example, the display may show an image of a location on the user's chest where the user is instructed to hold a third electrode while holding electrodes 1 and 2 in the user's left and right hands, respectively.

[0042] In some embodiments, software on the ECG sensing device 500 or computing device 550 is configured to recognize whether the first electrode is in contact with the left hand and the second electrode with the right hand, or whether the first electrode is in contact with the left hand and the second electrode with the right hand. For example, in some embodiments, the third electrode is positioned on a different surface of the ECG sensing device 500 from the first and second electrodes, so that the user is likely to need to change the position of their hand to make contact with the third electrode at an anterior chest lead position on the user's chest after making contact with the user's left leg. In some embodiments, software on the ECG sensing device 500 or other mobile computing device receives information from sensors coupled to or integrated with the ECG sensing device 500, which provide information about the device's position in space. Examples of classes of sensors that sense such information include, but are not limited to, accelerometers, inclinometers, and gyrometers.

[0043] In some embodiments, the ECG sensing device 500 is configured to sense an ECG when one or more of the electrodes 502 are not engaged by the user. For example, in some embodiments, the ECG sensing device 500 has three electrodes and is configured to sense an ECG when all three electrodes are engaged by the user, or when any two of the three electrodes are engaged by the user. That is, in this embodiment, for example, if the user touches the first electrode to the skin surface on the user's right upper limb and the second electrode to the skin surface on the user's left upper limb, but does not touch the third electrode, the ECG sensing device senses an ECG. In this example, lead I is sensed when two of the three electrodes are touched by the right upper limb and the left upper limb, respectively. Similarly, lead II is sensed when two of the three electrodes are touched by the right upper limb and the left lower limb, respectively. Similarly, lead III is sensed when two of the three electrodes are touched by the left upper limb and the left lower limb, respectively. In this embodiment, the ECG sensing device 500 recognizes, for example, that two or more electrodes are in contact with the user, while one or more of the electrodes are not, by sensing the electrode potential from two or more electrodes that are in contact with the user, but not sensing the electrode potential from electrodes that are not in contact with the user.

[0044] In some embodiments of the ECG sensing device described herein, exemplary embodiments are shown in Figures 5A to 5E, a mobile computing device (e.g., computing device 550) is configured to run the software applications described herein. In further embodiments, the mobile computing device includes one or more hardware central processing units (CPUs) or general-purpose graphics processing units (GPGPUs) that perform the functions of the device. Still in further embodiments, the mobile computing device further comprises an operating system configured to execute executable instructions. In some embodiments, the mobile computing device is optionally connected to a computer network. In further embodiments, the mobile computing device is optionally connected to the Internet to access the World Wide Web. Still in further embodiments, the mobile computing device is optionally connected to a cloud computing infrastructure. In other embodiments, the mobile computing device is optionally connected to the Internet. In other embodiments, the mobile computing device is optionally connected to a data storage device.

[0045] As described herein, suitable mobile computing devices include, in non-limiting examples, server computers, desktop computers, laptop computers, notebook computers, subnotebook computers, netbook computers, netpad computers, handheld computers, smartphones, smartwatches, digital wearable devices, and tablet computers.

[0046] In some embodiments, a mobile computing device includes an operating system configured to execute executable instructions. The operating system is software, including programs and data, that manage the device's hardware and provide services for running applications. Non-limiting examples of suitable operating systems include FreeBSD, OpenBSD, NetBSD®, Linux®, Apple® Mac OS X Server®, Oracle® Solaris®, Windows Server®, and Novell® NetWare®. Those skilled in the art will recognize that suitable personal computer operating systems include, non-limiting examples, UNIX-like operating systems such as Microsoft® Windows®, Apple® Mac OS X®, UNIX®, and GNU / Linux®. In some embodiments, the operating system is provided by cloud computing.

[0047] In some embodiments, a mobile computing device includes a storage and / or memory device. The storage and / or memory device is one or more physical devices used to temporarily or permanently store data or programs. In some embodiments, the device is volatile memory and requires power to maintain the stored information. In some embodiments, the device is non-volatile memory and retains the stored information when the mobile computing device is not powered. In further embodiments, the non-volatile memory comprises flash memory. In some embodiments, the non-volatile memory comprises dynamic random access memory (DRAM). In some embodiments, the non-volatile memory comprises ferroelectric random access memory (FRAM®). In some embodiments, the non-volatile memory comprises phase-change random access memory (PRAM). In other embodiments, the device is a storage device that, as a non-limiting example, includes CD-ROMs, DVDs, flash memory devices, magnetic disk drives, magnetic tapes, optical disk drives, and cloud computing-based storage. In further embodiments, the storage and / or memory device is a combination of devices such as those disclosed herein.

[0048] In some embodiments, the mobile computing device includes a display for sending visual information to the user. In some embodiments, the mobile computing device includes an input device for receiving information from the user. In some embodiments, the input device is a keyboard. In some embodiments, the input device is a pointing device, including, in non-limiting examples, a mouse, trackball, trackpad, joystick, game controller, or stylus. In some embodiments, the input device is a touchscreen or multitouchscreen. In other embodiments, the input device is a microphone for capturing voice or other sound input. In other embodiments, the input device is a video camera or other sensor for capturing motion or visual input. Still in further embodiments, the input device is a combination of devices such as those disclosed herein.

[0049] In various embodiments, the platforms, systems, media, and methods described herein include a cloud computing environment. In some embodiments, the cloud computing environment comprises multiple computing processors.

[0050] Figures 5A to 5E illustrate exemplary embodiments of the user aspects described herein, but it should be understood that, in general, numerous electrode positions, shapes, and sizes may be used in the devices described herein to ensure that an individual can comfortably and naturally contact the electrodes. For example, all three electrodes may be positioned entirely on the side of a computing device or device cover.

[0051] In any of the embodiments shown in Figures 5A to 5E, one or more electrodes may be configured to be removable from the ECG sensing device 500. In these embodiments, the ECG sensing device has either a male or female connector configured to snap-fit ​​to, for example, the corresponding male or female connector on the removable electrode.

[0052] The embodiments shown in Figures 5A to 5E illustrate an ECG sensing device comprising three electrodes, but it should be understood that other numbers of ECG electrodes may be incorporated into the ECG sensing devices described herein.

[0053] In general, any of the techniques, components, and / or subsystems described above may be used in or combined with any of the other examples. For example, any of the ECG devices described herein may include any of the features described above.

[0054] QTc prolongation, whether secondary to genetic and / or acquired risk factors, represents an independent risk factor for SCD and a predictor of both all-cause and cardiovascular mortality in various clinical settings. Importantly, studies have demonstrated that approximately 1% of all individuals undergoing 12-read ECGs in hospitals or outpatients have a QTc ≥ 500 milliseconds. It should be noted that meeting or exceeding this threshold increases the risk of death by approximately two to four times and serves as a stronger predictor of all-cause mortality than conventional comorbidity indices, at least over a short period (e.g., 30 days). While this specification discusses the 500-millisecond threshold, any appropriate threshold (e.g., 450 milliseconds, 475 milliseconds) may be used to indicate QTc prolongation representing an increased risk of cardiovascular mortality. In addition, a specific amount of QTc increase over a particular period may also indicate QTc prolongation representing an increased risk of cardiovascular mortality. For example, if a user's QTc is measured at 400 milliseconds and then again at 450 milliseconds (within the threshold period), this may indicate a QTc extension, even if 450 is below the QTc threshold for the individual measurement (e.g., 500 milliseconds).

[0055] Furthermore, in many situations, the occurrence of QTc ≥ 500 milliseconds is caused, at least partially, by the presence of a potentially lethal but highly treatable genetic condition (i.e., congenital LQTS) and / or modifiable risk factors (e.g., electrolyte abnormalities, use of one or more QTc prolongers, or potential QT excitation disorders). Therefore, in many situations, the identification of substantial QTc prolongation provides a crucial opportunity to i) identify vulnerable and at-risk hosts and ii) make necessary changes to potential life-saving measures (i.e., initiation of beta-blockers, discontinuation of QTc prolongers, or correction of hypokalemia and hypomagnesemia) to mitigate the risk of TdP and SCD.

[0056] However, because it relies on bulky 12-lead ECG systems and trained ECG technicians, the clinical settings in which QTc can be monitored are largely limited to “snapshot” assessments in hospitals and outpatient clinics.

[0057] Therefore, in some embodiments, the ECG sensing device 500 may be used to predict the user's QT interval. As discussed herein, machine learning (ML) is well suited to continuous monitoring of one or more criteria for identifying anomalies or trends in input data, regardless of scale, compared to training examples used to train a model. ML models that may be used in the embodiments described herein include, but are not limited to, Bayesian, Markov, Gaussian processes, clustering algorithms, generative models, kernels, and neural network algorithms. Some embodiments utilize machine learning models based on trained neural networks (e.g., trained recurrent neural networks (RNNs) or trained convolutional neural networks (CNNs)).

[0058] The ECG sensing device 500 may run module 510B to predict the QT interval, as discussed herein. As described herein, module 510B may comprise a deep neural network (DNN) for predicting the QT interval, but any suitable ML model may be used. The user may deploy the ECG sensing device 500 in any suitable manner described herein (for example, in response to the receipt of a command, as discussed in further detail herein) and perform an ECG using the ECG sensing device 500. The processing device 505 may process the recorded ECG signal using an averaged beat algorithm, the input to the DNN may be the averaged beat in the form of a 2 × 450 length signal in millivolts. The average pulse signal may first be clamped via a hyperbolic tangent function before passing through an initial convolutional layer (e.g., 2D convolution, leakage rectifier linear unit, maximum pooling), before being processed by three separate pass-through stages (e.g., two separate 2D convolutions, leakage rectifier linear unit, maximum pooling), and before a pass-through input (via trainable parameters) is added, with dropout applied before downsampling, pooling, and convolution. After the final pass-through, the signal is sent to two separate feedforward network layers that output a probability vector about the output class of a given QT value, the expected value of which is considered the predicted QT interval. The loss function for training is the sum of the cross-entropy term of the probability vector and the target QT interval and the root mean square error of the average QT interval. It should be noted that the ML model may utilize (and be trained on) any appropriate number of reads or combinations of reads. In some embodiments, the ML model may be trained to predict a QT interval based on a combination of reads I, II, or III and any appropriate v read. For example, an ML model can be trained on reads I and v5, reads II and v5, or reads III and v5 (or any other suitable v-reads).

[0059] The DNN can be trained using annotated standard 12-read ECG data from various patients obtained at the same location where the user was when the ECG was acquired. In some embodiments, the ECG signals can be preprocessed before training. For example, a bandpass filter may be applied to simulate the ambulatory bandwidth (0.1–40 Hz) of the ECG sensing device 500, and any suitable filtering algorithm (e.g., Alivecor®'s FDA-approved extended filtering algorithm) may be applied before processing the resulting 12-read ECGs using an average beating algorithm for leads I and II. The source ECGs containing the training data may be bucketed based on QT intervals, and ECGs are randomly selected within randomly chosen buckets. Since the ECG signals received from the user may include leads generated using a separate lead conversion ML algorithm (e.g., via module 510A), in some embodiments, the DNN can be further fitted using additional training data including synthesized leads.

[0060] All devices described herein are suitable for use in a variety of systems, which may include one or more servers, one or more sensors, an electronic data communication network, and other ECG sensing devices. In some embodiments, multiple ECG sensing devices described herein transmit ECG data to one or more remote servers via an electronic data communication network. In some embodiments, the ECG data is analyzed using one or more remote servers. In some embodiments, arrhythmia detection is performed using remote servers that analyze the received ECG data.

[0061] All devices and systems described herein may also include one or more software modules. In some embodiments, the software includes an application configured to run on a mobile computing device such as a smartphone, smartwatch, or tablet computer. The software receives and processes ECG data received from an ECG sensing device. The software identifies distinct leads in the transmitted data, for example, based on which electrode the ECG data originated from. For example, the software may identify lead I based on signals from two electrodes measuring the potential difference between the right upper limb and the left upper limb. Once the ECG is identified, the software may be further configured to display a single-lead or multi-lead ECG on the display screen of the mobile computing device. The software may be configured to display six leads I, II, III, aVR, aVL, and aVF simultaneously on the display screen. The software may be configured to display one or more of the six leads I, II, III, aVR, aVL, and aVF on the display screen at one time, and the user can manually switch screens to view different leads on different switched screens.

[0062] The software modules described herein include computer-readable and executable code. In various embodiments, a software module includes files, sections of code, programming objects, programming structures, or a combination thereof. In further various embodiments, a software module includes multiple files, multiple sections of code, multiple programming objects, multiple programming structures, or a combination thereof. In various embodiments, one or more software modules include, as non-limiting examples, web applications, mobile applications, and standalone applications. In some embodiments, a software module resides within one computer program or application. In other embodiments, a software module resides within two or more computer programs or applications. In some embodiments, a software module is hosted on one machine. In other embodiments, a software module is hosted on two or more machines. In further embodiments, a software module is hosted on a cloud computing platform. In some embodiments, a software module is hosted on one or more machines in one location. In other embodiments, a software module is hosted on one or more machines in two or more locations.

[0063] In some embodiments, two or more sensed leads that are not sensed simultaneously are time-aligned to generate a time-aligned ECG trace that displays two or more leads in a time-aligned format, such as in a conventional standard 12-lead ECG trace. In some embodiments of the ECG sensing devices described herein, one or more ECG sensing electrodes are not simultaneously placed on the skin of the individual whose ECG is sensed (i.e., several leads may be sensed sequentially). For example, limb leads (I, II, III, aVR, aVL, and aVF) are sensed simultaneously, but one or more of the chest leads are sensed separately from the limb leads. Therefore, in these embodiments, the six limb leads are not automatically time-aligned with the individually and separately sensed chest leads, and further processing is performed by a software application to time-align one or more of the limb leads with one or more of the chest leads. In some embodiments, one or more of the six chest leads are sensed individually so that the individually sensed chest leads are time-aligned by a software application with the six limb leads and other chest leads. In some embodiments, the software application described herein aligns two or more sensed chest leads with each other and six sensed limb leads separately so that two sets of six leads (i.e., six time-aligned chest leads and six separately time-aligned limb leads) are time-aligned. In some embodiments, the software described herein aligns two or more sensed chest leads with each other and with the sensed limb leads so that all twelve sensed leads are time-aligned.

[0064] In some embodiments, one or more average or median waveforms are generated for the first and second leads so that the waveforms corresponding to different heartbeats are time-aligned. That is, in some embodiments where one or more leads are not sensed simultaneously, an average or median waveform is generated for one or more of these leads, and the average or median waveforms are time-aligned so that the P, QRS, and T waveforms / compounds are aligned vertically along the X-axis.

[0065] Memory 510 may include a time-aligning software module (not shown) capable of performing time-alignment of the P, QRS, and T waveforms / compounds of each read sensed by the ECG sensing device 500 so that the sensed ECG reads are aligned when displayed as waveforms in conventional ECG tracing. In some embodiments of the ECG sensing device, the ECG sensing device includes a software application configured to time-align two or more sensed ECG reads. In some embodiments of the ECG sensing device, the software application configured to time-align two or more sensed ECG reads is a component of a system that receives data from the ECG sensing device.

[0066] When the first and second electrodes of the ECG sensing device described herein are in contact with the user's right upper limb and left upper limb, simultaneously with the third electrode of the device being in contact with any one of the six anterior chest lead positions, lead I is sensed along with the sensed anterior chest leads. That is, since lead I is equal to the voltage sensed at the left upper limb minus the voltage sensed at the right upper limb, when the left upper limb, right upper limb, and chest are all in contact with the electrodes of the ECG sensing device described herein, lead I is sensed in addition to the anterior chest leads. Thus, when all six anterior chest leads are sensed sequentially, six corresponding "anterior chest lead I records," namely V1-lead I, V2-lead I, V3-lead I, V4-lead I, V5-lead I, and V6-lead I, are also generated. Each of these six anterior chest lead I records is used to time-align each anterior chest lead with the limb leads, and thus time-align the anterior chest leads.

[0067] In some embodiments, the time-aligning software module aligns the chest leads V1, V2, V3, V4, V5, and V6 by utilizing the existence of chest lead I recordings sensed simultaneously with each of the waveforms V1, V2, V3, V4, V5, and V6. That is, the chest lead I recordings V1-Lead I, V2-Lead I, V3-Lead I, V4-Lead I, V5-Lead I, and V6-Lead I are each time-aligned with the chest lead recording sensed simultaneously with them. Each chest lead I recording is time-aligned with the lead I sensed together with the limb leads by, for example, moving the chest lead I recording a certain distance along the Y-axis, and each chest lead I recording is time-aligned with the chest leads, so each of the chest leads V1, V2, V3, V4, V5, and V6 is time-aligned when moved along the Y-axis by the same distance as the simultaneously sensed chest lead I recording. For example, “V1-Lead I” is a Lead I recording time-aligned with V1. “V1-Lead I” is not the same as “Lead I,” which is a recorded Lead I sensed simultaneously with the other five limb leads using the ECG sensing device described herein. Since these two different Lead I recordings are not typically sensed simultaneously using the ECG sensing device described herein, “V1-Lead I” is also not necessarily time-aligned with “Lead I.” However, since both “V1-Lead I” and “Lead I” are Lead I recordings, they can be time-aligned in a fairly simple way, as they are both expected to have very similar (if not identical) waveform morphologies and timings when averaged. For example, if the R-wave peak in the averaged “Lead I” occurs at 1 second and the R-wave peak in the averaged “V1-Lead I” occurs at 1.5 seconds, the averaged “V1-Lead I” can be repositioned or shifted by 0.5 seconds along the Y-axis so that its R-wave peak occurs at 1 second, as in the averaged “Lead I.” Since V1 is time-aligned to V1-Lead I, V1 must also be shifted 0.5 seconds along the Y-axis to time-align with the averaged "Lead I".When V1 is time-aligned to "Lead I," V1 is also time-aligned with the other five limb leads that are already time-aligned to "Lead I." Similar alignment occurs for V2, V3, V4, V5, and V6 by aligning V2-Lead I, V3-Lead I, V4-Lead I, V5-Lead I, and V6-Lead I to "Lead I," respectively.

[0068] Figure 6 is a flowchart of a method 600 for predicting the QT interval according to some embodiments of the present disclosure. The method 600 may be implemented by processing logic which may comprise hardware (e.g., circuits, dedicated logic, programmable logic, processors, processing devices, central processing units (CPUs), system-on-a-chip (SoCs), etc.), software (e.g., instructions that operate / execute on the processing device), firmware (e.g., microcode), or a combination thereof. In some embodiments, the method 600 may be implemented by an ECG sensing device 500 (via a processing device 505) as shown in Figure 5A.

[0069] In block 605, the ECG sensing device 500 may provide the user with commands regarding the position where the user's body should be located (in some embodiments, the computing device 550 may provide these commands) and commands for positioning the ECG sensing device 500 relative to the user's body to record an ECG. In block 610, the processing device 505 may perform an ECG and acquire the user's ECG signal. In block 615, the processing device 505 may process the recorded ECG signal using an average heartbeat algorithm, and in block 620, the processing device 505 may execute module 510B to predict the user's QT interval. As described herein, module 510B may comprise a deep neural network (DNN) for predicting the QT interval, but any suitable ML model may be used. The input to the DNN may be, for example, the average heartbeat in the form of a 2 × 450 length signal in millivolts. The mean pulsating signal may first be clamped via a hyperbolic tangent function before passing through an initial convolutional layer (e.g., 2D convolution, leak rectifier linear unit, max pooling) before being processed by three separate pass-through stages (two separate 2D convolutions, leak rectifier linear unit, max pooling, etc.), and before a pass-through input is added (via trainable parameters), with dropout applied before downsampling, pooling, and convolution. After the final pass-through, the signal may be sent to two separate feedforward network layers that output a probability vector about the output class of a given QT value, the expected value of which is taken as the predicted QT interval. The loss function for training is the sum of the cross-entropy term of the probability vector and the target QT interval and the root mean square error of the average QT interval.

[0070] In some embodiments, one or more electrodes may be located outside the mobile computing device. In such embodiments, one or more external electrodes may be coupled to the mobile computing device wirelessly or by wiring. Non-limiting examples of wireless connections may include, for example, a WiFi connection between one or more external electrodes and the device, a Bluetooth® connection between one or more external electrodes and the device, a low-power Bluetooth connection between one or more external electrodes and the device, an NFC (Near Field Communication) connection between one or more external electrodes and the device, or a near-field ultrasonic communication connection between one or more external electrodes and the device. It should be understood by those skilled in the art that other means of wirelessly communicating with the device are suitable for use in the systems, devices, and methods described herein.

[0071] In addition, software combined with any of the systems, devices, or methods described herein may be configured to analyze ECG data received from an ECG sensing device. The analysis may include generating QRS axis and T axis values ​​using six leads I, II, III, aVR, aVL, and aVF, as described herein. In addition, software combined with any of the systems, devices, or methods described herein may determine the QRST angle by calculating the difference between the QRS axis and the T axis, as described herein. The analysis may further include rhythm analysis, which may include determining heart rate variability, QT interval, or corrected QT interval.

[0072] In addition, software combined with any of the systems, devices, or methods described herein may be used to determine diagnoses or abnormalities related to the ECG. For example, as described, axis deviation may be associated with abnormalities of right or left ventricular hypertrophy. For example, heart rate variability may be associated with the diagnosis of atrial fibrillation. For example, changes in the QT interval may indicate certain arrhythmias.

[0073] Any of the systems, devices, and methods described herein may also be combined with sensors that measure physiological parameters. For example, any of the systems, devices, or methods described herein may be combined with a blood pressure sensor. For example, any of the systems, devices, or methods described herein may be combined with a photoplethysmogram (PPG) sensor. For example, any of the systems, devices, or methods described herein may be combined with a temperature sensor. For example, any of the systems, devices, or methods described herein may be combined with a pulse oximetry sensor. For example, any of the systems, devices, or methods described herein may be combined with an accelerometer. Those skilled in the art will understand that other sensors for monitoring or detecting physiological parameters are suitable for use with the systems, devices, and methods described herein.

[0074] In some embodiments, the sensed physiological data is transmitted to a processor in any of the systems, devices, and methods described herein. Software combined with any of the systems, devices, and methods described herein may use the sensed physiological data in combination with the sensed ECG to perform an analysis. For example, blood pressure data may be combined with ECG data by the software to provide an analysis that determines the presence of ventricular tachycardia, a condition that is immediately life-threatening.

[0075] The systems, devices, and methods described herein may include either or both transmitters and receivers for transmitting and receiving wireless signals.

[0076] In some embodiments, the software described herein also triggers the transmission of a signal to a server when an abnormal analysis result is determined. For example, an abnormal analysis result includes an abnormal ECG. For example, an abnormal analysis result includes an abnormal QRS axis. For example, an abnormal analysis result includes an abnormal QRST angle. In some embodiments, an abnormal analysis result includes an abnormal ECG. For example, an abnormal analysis result includes abnormal heart rate variability values. For example, an abnormal analysis result includes abnormal physiological parameter values. The transmitted signal may include a signal to an emergency medical provider. For example, if a life-threatening condition such as VT is determined, the software described herein may transmit an emergency signal to an emergency operator, emergency medical provider (e.g., paramedics), or other third-party monitor.

[0077] A 6-read ECG may be displayed on the display screen, and the 6-read ECG includes the read I, the read II, the read III, the read aVR, the read aVL, and the read aVF.

[0078] Preferred embodiments of the systems, devices, and methods described herein are shown and described herein, but it will be apparent to those skilled in the art that such embodiments are provided merely as examples. Now, numerous modifications, changes, and substitutions will be made to those skilled in the art without departing from the subject matter described herein. It should be understood that various alternatives to the embodiments of the systems, devices, and methods described herein may be used when carrying out the systems, devices, and methods described herein. The following claims define the scope of the invention, and methods and structures within the scope of these claims, as well as their equivalents, are intended to be covered thereby.

[0079] In one embodiment, the methods and systems described herein may be combined with methods and systems for guiding and / or modifying electrode placement on a patient. These methods may be particularly useful for guiding electrode placement for ECG measurement. Generally, the systems, devices, and methods contemplated herein process a patient photograph to output electrode positions on the patient. Typically, this may involve presenting an image of the patient (e.g., a modified version of the patient photograph) indicating the locations for electrodes relative to an actual patient photograph.

[0080] For example, a system or device for guiding electrode placement as described herein may include control logic for controlling a processor (e.g., a microprocessor in a computing device such as a handheld computing device) to receive a photograph of the patient, analyze the patient to determine the correct placement of electrodes, and output an image of the patient with the correct predetermined electrode positions marked. Generally, the control logic may be configured as software, hardware, or firmware and may control a general-purpose computing device (e.g., a computer, tablet, etc.) or a mobile communication device (e.g., a smartphone such as an iPhone® or Android®) to receive or acquire a photograph and output an image of the patient. The processing steps may be performed remotely or locally. Generally, the processing steps may include comparing the photograph of the patient to a database (e.g., an electrode placement database) of corresponding predetermined, conventional, or standard positions of electrodes associated with various body types and each body type. The photograph of the patient may also be normalized before being compared to the patient database by adjusting the size of the photograph, and / or, optionally, the aspect ratio, brightness, contrast, or other image features, in order to enable direct comparison with the database. Normalization may be performed using markers included as part of the photograph. For example, a patient's photograph may be taken with a marker of a known or recognizable size placed on the patient, and the marker may be used as a normalization marker to normalize the photograph before comparison with a database. Normalization may also be performed to equalize the brightness, contrast, sharpness, or other image quality of the photograph. The marker may be placed or applied directly to the patient (e.g., the patient's torso) by means of adhesive, for example.

[0081] Also intended and described are methods performed by devices and systems for guiding electrode placement, such as methods for guiding electrode placement on a patient.

[0082] For example, the method intended herein for guiding the proper placement of electrodes on a patient includes the steps of: comparing a photograph of a patient with an electrode placement database in order to determine the positioning of electrodes on the patient, wherein the electrode placement database includes representations of a plurality of body types and predetermined electrode placement positions corresponding to each body type; and presenting an image of the patient showing the positions of electrodes on the patient's image.

[0083] These methods can be particularly adapted to guide the placement of ECG electrodes in patients in a standard or conventional configuration. Therefore, the database may be configured to include multiple body types, each with a corresponding conventional / standard electrode placement position.

[0084] In some variations, electrode positions may be determined and shown for all electrodes (e.g., all 10 electrode positions used in a standard 12-electrode lead). However, in some variations, a subset of electrode positions may be determined and / or shown. For example, a method for guiding the positioning of a standard / conventional 12-lead electrode arrangement may determine and show only six electrode positions on the patient's chest. In some variations, where other electrode positions may be determined for one or more important electrode positions, only the positions of the important electrodes may be shown.

[0085] In general, any suitable photograph of the subject may be used. In some variations, the system, device, or method may include taking or acquiring a photograph. In some variations, the photograph may be taken by a system or device performing the method (e.g., a smartphone or other handheld computer device). The systems, devices, and methods described herein may instruct the user on how to take a photograph of the patient, including positioning the patient (e.g., facing forward, standing, sitting, lying down), approximately how far away from the patient to take the photograph, and placing normalization markers on or near the patient. The photograph may be received as a digital image. The photograph may include an image of the patient, in particular an image of the area of ​​the patient's body to which electrodes should be applied. For example, when applying ECG electrodes, the photograph may include the patient's torso or chest. Additional areas of the patient's body may include the patient's head, legs, etc. The patient may be standing, sitting, or lying down. The area of ​​the patient to which electrodes should be applied is typically nude (e.g., topless, or at least partially topless, so that the skin can be visualized). As mentioned above, in some variations, a normalization marker may be included as part of the photograph. For example, a reference marker may be placed on the patient, and the reference / normalization marker typically has a known or standard size, such as a coin (e.g., a US 25-cent coin, penny, etc.). In some variations, the reference marker may be provided and may have a distinct shape or color. In some variations, the marker is automatically recognized by the device. For example, the marker may include a readable code (e.g., a barcode, alphanumeric code, QR code®, etc.), or alternatively, the device may identify the marker by color, shape, etc.

[0086] In variations where a method, system, or device guides a user through taking or acquiring a photograph, the photograph may be certified by the system or device. Certifying a photograph may include checking the photograph to ensure that it is appropriate and can be analyzed (e.g., compared) against a database.

[0087] As used herein, the terms “user” and “patient” are broadly intended to include any subject on which the methods, devices, and systems may be used to assist in positioning electrodes thereon. Patients may include animals (in systems and devices specifically configured for use in that type of animal) or humans, and may include healthy or unhealthy individuals. “User” as used herein may be any person using the systems, methods, and devices described herein. In some modifications, the systems, devices, and methods described herein may be used by a patient to guide the placement of electrodes onto the patient themselves, so the user is the same as the patient.

[0088] In some variations, comparing photographs with an electrode placement database may involve determining a standard electrode placement for a 12-lead ECG on a patient.

[0089] Generally, comparing a patient's photograph to an electrode placement database may involve determining the match (e.g., the closest match) between the photograph and one or more typical body types in the patient database. Once one or more closely matching typical body types are identified, the electrode placement corresponding to the typical body types of those matches can be mapped to the patient's photograph. If two or more matches are identified, the electrode placement can be determined from standard electrode placements corresponding to multiple typical body types by weighting, averaging, or finding a consensus criterion among the closest matches and mapping this standard electrode placement to the patient's photograph using other appropriate statistical methods.

[0090] As will be explained in more detail below, electrode placement databases typically contain multiple representations (e.g., >10, >100, >1,000, >10,000, etc.) of standard / conventional electrode placements for different bodies. Body type representations may include information extracted from images of the body (e.g., photographs, parts of photographs, etc.) or from images of the body that include electrode placements specific to that body, and the electrode placements are confirmed or verified to be within standard / conventional boundaries. Various body types may include body types of different shapes and sizes (height, weight, form), sex (male / female), age (infant, child, adult, elderly), body shape (shoulder width, chest size, waist size, etc.), etc. Each body type representation may be unique, but may include similar body types, forming clusters of body types around more common body types. All body types within the database may be pre-normalized to allow comparison between different representations. Multiple different electrode placement databases may be used. For example, separate databases may be used for different patient positions (lying down, sitting, standing, etc.) or for different patients' sex, age, etc. Furthermore, different electrode placement databases may be used for different standard / conventional electrode placements.

[0091] Therefore, in addition to normalizing photographs before comparison with an electrode placement database, photographs may be processed to prepare them for comparison with the database. In variations where the comparison is performed by extracting features from the photograph and comparing these extracted features with representations of body shape in the database, feature extraction may be performed on the photograph before (or as part of) the comparison. For example, if comparing a patient's photograph with an electrode placement database involves determining anatomical landmarks from the photograph and comparing the anatomical landmarks with the electrode placement database, the anatomical landmarks may be extracted from the photograph first. Photographs may also be processed to remove any patient-identifying features (e.g., all or part of the patient's face) that may be relevant to protect patient privacy.

[0092] As mentioned above, comparing a photograph with a database may involve interpolation between the closest match to the photograph and two or more typical body types within the patient database.

[0093] In some variations, comparing a patient's photograph to a database involves using pattern recognition to determine the closest match between the photograph and a typical body type in the database. In some variations, comparing a patient's photograph to an electrode placement database involves comparing a normalized photograph of the patient to an electrode placement database.

[0094] The methods, devices, and systems contemplated herein may also include presenting patient images showing the positions of electrodes on the patient's image. Any suitable patient image may be presented, including a modified version of the patient's photograph showing the electrode positions determined by comparison with a database. In some variations, the patient image is displayed digitally (e.g., on a handheld computing device). It may also be magnified (zoomed in / zoomed out) or manipulated so that the user can see where to place the electrodes. In some variations, the image may include additional guidelines, including measurements (ruler, distance in inches, mm, etc.) relating to the patient and / or other electrodes, including patient landmarks such as anatomical landmarks.

[0095] The presentation of patient images showing the conventional / standard positions of electrodes may show all electrodes or some of the electrodes. In some variations, the presentation of images may include a series of images showing patients separately with different electrode positions to better enable the user to proceed with the process of applying or repositioning the electrodes. Generally, the presentation of patient images may be visual (showing images) and may also include text (written / oral) instructions for applying the electrodes. For example, in variations of the systems and methods described herein intended for use with a handheld computer device such as a smartphone, the device may be controlled to allow the user to proceed with both taking a photograph of the patient and positioning (or repositioning) the electrodes by viewing the smartphone screen.

[0096] In some modifications, the methods, devices, and systems described herein may be used to correct and / or verify the placement of electrodes already present on a patient. For example, a user may take or receive a photograph of a patient who already has ECG electrodes on their chest. Comparing the patient's photograph to an electrode placement database may also allow for a comparison of the existing electrode placements on the patient with determined standard / conventional placements. Thus, comparing a patient's photograph to an electrode placement database may include comparing a photograph of a patient with one or more electrodes already placed on the patient's chest to the electrode placement database. The placements of one or more electrodes already placed on the patient's chest may then be verified automatically (as indicated when one or more occur) or passively by overlaying the correct placements onto the patient's photograph (in some specific way, e.g., indicated by color) to form a presented image. In some modifications, the presented image includes a patient image showing the correct placement of electrodes on the patient's image.

[0097] Furthermore, the method for guiding the placement of an ECG electrode comprises the steps of: receiving a photograph of the patient including the patient's chest; comparing the photograph of the patient with an electrode placement database in order to determine the positioning of an electrode on the patient, wherein the electrode placement database includes representations of a plurality of body types and predetermined conventional ECG electrode placement positions corresponding to each body type; and presenting an image of the patient showing the positions of conventional ECG electrode positions on the image of the patient. The method of claim 17, wherein comparing the photograph of the patient with an electrode placement database includes determining the closest match between the photograph and a typical body type in the electrode placement database.

[0098] As described above, comparing a patient's photograph to an electrode placement database involves determining anatomical landmarks from the photograph and comparing those anatomical landmarks to the electrode placement database. In some variations, comparing a patient's photograph to the database involves using pattern recognition to determine the closest match between the photograph and a typical body shape in the electrode placement database.

[0099] In any of the modifications described herein, comparing a patient's photograph with an electrode placement database may be performed independently of other steps. For example, a smartphone may be used (e.g., using an application downloaded to the smartphone) to acquire a patient's photograph and present an image of the patient showing the conventional electrode placement, and the comparison of the photograph with the database may be performed remotely using a remote server. Thus, the database may be maintained separately from the application on the smartphone (or other device). This may allow for modification, updating, or otherwise changing the database and / or the mechanism for comparing patient photographs with the database. The generated image may then be presented on the handheld computer device after the handheld computer device has received the information (or generated image) from the remote database. Alternatively, in some modifications, all steps are performed at a local level (e.g., using a handheld computing device such as a smartphone or tablet computer).

[0100] As described above, the patient's photograph may include a normalization marker. Therefore, the step of receiving the patient's photograph may include receiving the patient's photograph which includes the normalization marker. In some variations, the received patient's photograph may include electrodes on the patient's chest, and the method, device, or system may verify the placement of electrodes already on the chest against conventional ECG electrode placement positions.

[0101] Also described herein is a method for determining the placement of an ECG electrode, comprising the steps of: receiving a photograph of a patient including a normalization marker; normalizing the photograph using the normalization marker; comparing the normalized photograph with an electrode placement database including representations of multiple body types and predetermined ECG electrode placement positions for each body type, in order to determine the positioning of the electrode relative to the patient; and presenting an image of the patient showing the position of the ECG electrode on the patient's image.

[0102] The system or device may be configured to receive a photograph of a patient including the area of ​​the patient to which electrodes should be applied, analyze the photograph, and perform any or all of the steps described above to provide a patient image (or any other patient-specific map) showing the location of one or more electrodes on the patient based on predetermined conventional and / or standard electrode positions.

[0103] Many of the examples described herein are specific to systems, devices, and methods for positioning ECG electrodes (e.g., device 500) according to standard or conventional 12-lead ECG electrode configurations, but these systems, devices, and methods may be used (or adapted for use) with any given conventional and / or standard electrode positioning system, including electrodes for electroencephalography (EEG), electromyography (EMG), galvanic skin reflex (GSR), electrooculography (EOG), bioelectrical impedance (BI), etc. For example, an electrode placement database may include various body types and corresponding predetermined conventional and / or standard electrode positions for each body type for EEG, EMG, GSR, EOG, BI, etc. In some modifications, different electrode placement schemes (different conventional and / or standard electrode placements) may be linked to each body type in the database, and the user may select which placement scheme to display. In other modifications, two or more placement schemes may be shown sequentially or simultaneously for the same patient. For example, regarding ECG electrode placement, the electrode configuration can accommodate 3-lead ECG, 5-lead ECG, and 12-lead ECG.

[0104] A system for guiding electrode placement may generally include control logic that can run as software, hardware, or firmware (or a combination thereof) that receives a photograph of a patient, uses an electrode placement database to determine conventional and / or standard electrode placements for that patient, and outputs a map or image of the patient showing where the electrodes should be placed. The system may also be configured to guide or direct the user through the process of taking a photograph of the patient and / or positioning the electrodes relative to the patient. In some variations, the system may be configured to guide the user by audible, written, and / or visual instructions. The system may be configured to operate (e.g., control) from a handheld computing device, including a smartphone (e.g., iPhone, Android, etc.), to receive (and possibly take) a photograph of the patient and output an image of the patient with the determined electrode positions marked. For example, the system may be configured as a smartphone application downloadable to a smartphone.

[0105] Any of the systems described herein may be dedicated systems pre-configured to receive a photograph of a patient and output an image of the patient indicating electrode placement locations, without requiring the download of an application (e.g., software) to a separate device. For example, the system may include a camera for taking a photograph of the patient and control logic for receiving the image, controlling the analysis of the photograph to determine electrode placement using an electrode placement database, and outputting a map or image of the patient indicating the location of one or more electrodes according to conventional and / or standard electrode positioning schemes. The system may include all or part of the electrode placement database, or the system may communicate with a remote electrode placement database. Furthermore, the system may include a comparison unit which may include comparison logic for comparing a photograph of the patient with a body type in the electrode placement database to find one or more close matches between the patient and the body type in the database, from which predetermined conventional and / or standard electrode positions can be extrapolated into the photograph of the patient.

[0106] The system may also be configured to use a normalization marker contained within a patient's photograph (or may be included as part of the system). A normalization marker is typically a distinct marker that the system / device described herein can identify within a photograph and can be used to provide a scale and / or orientation for reference within the photograph. For example, a normalization marker may be a sticker that can be attached to the patient's skin, the sticker may be brightly colored and have a known size (e.g., an orange circle with a diameter of 1 inch). Thus, the system / device can identify this sizing marker in the photograph and normalize the photograph using the normalization marker. In some variations, the normalization marker may also provide a reference position that the system can use when providing instructions regarding electrode placement. In some variations, two or more sizing markers may be used. A normalization marker may be a common object of known dimensions, such as a coin. The user may indicate which normalization marker is being used in the system / device (e.g., from a menu of possible normalization markers).

[0107] As described above, patient images indicating electrode positioning can be presented to the user on a handheld computer device. For example, the handheld computer device may be a network-connected mobile phone, smartphone, tablet computer, or camera.

[0108] Figure 7 is a block diagram of an exemplary computing device 700 capable of performing one or more of the operations described herein, according to several embodiments. In some embodiments, computing device 700 may represent the internal hardware of ECG device 500. Computing device 700 may be connected to other computing devices in a LAN, intranet, extranet, and / or the Internet. Computing devices may operate in the capacity of a server machine in a client-server network environment, or in the capacity of a client in a peer-to-peer network environment. Computing devices may be provided by a personal computer (PC), a set-top box (STB), a server, a network router, a switch or bridge, or any machine capable of executing a set of instructions (sequential or otherwise) that specify the actions to be taken by the machine. Furthermore, although only a single computing device is shown, the term “computing device” shall also be interpreted to include any set of computing devices that individually or collectively execute a set of instructions (or sets of instructions) to perform the methods discussed herein.

[0109] An exemplary computing device 700 may include a processing device (e.g., a general-purpose processor, PLD, etc.) 702, main memory 704 (e.g., synchronous dynamic random access memory (DRAM), read-only memory (ROM)), and static memory 706 (e.g., flash memory and data storage device 718), which can communicate with each other via a bus 730.

[0110] The processing device 702 may be provided by one or more general-purpose processing devices, such as a microprocessor or a central processing unit. In exemplary examples, the processing device 702 may comprise a composite instruction set computing (CISC) microprocessor, a reduced instruction set computing (RISC) microprocessor, a very long instruction word (VLIW) microprocessor, or a processor implementing a processor or combination of instruction sets. The processing device 702 may also comprise one or more dedicated processing devices, such as an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), a digital signal processor (DSP), or a network processor. The processing device 702 may be configured to perform the operations and steps discussed herein, in accordance with one or more aspects of this disclosure.

[0111] The computing device 700 may further include a network interface device 708 capable of communicating with the network 720. The computing device 700 may also include a video display unit 710 (e.g., a liquid crystal display (LCD) or a cathode ray tube (CRT)), an alphanumeric input device 712 (e.g., a keyboard), a cursor control device 714 (e.g., a mouse), and an acoustic signal generation device 716 (e.g., a speaker). In one embodiment, the video display unit 710, the alphanumeric input device 712, and the cursor control device 714 may be combined into a single component or device (e.g., an LCD touchscreen).

[0112] The data storage device 718 may include a computer-readable storage medium 728 in which ECG instructions 726, for example, one or more sets of instructions for performing the operations described herein in one or more aspects of the present disclosure, can be stored. The ECG instructions 726 may also reside entirely or at least partially in the main memory 704 and / or the processing device 702 while they are being executed by the computing device 700, the main memory 704 and the processing device 702 also constitute the computer-readable medium. The instructions 726 may be further transmitted or received over the network 720 via the network interface device 708.

[0113] Although the computer-readable storage medium 728 is shown as a single medium in the illustrative examples, the term “computer-readable storage medium” should be interpreted to include a single or multiple mediums that store one or more sets of instructions (e.g., a centralized or distributed database and / or associated caches and servers). The term “computer-readable storage medium” should also be interpreted to include any medium that can store, encode, or carry a set of instructions for execution by a machine, causing the machine to execute in the manner described herein. Accordingly, the term “computer-readable storage medium” should be interpreted to include, but not limited to, solid memory, optical media, and magnetic media.

[0114] The methods and illustrative examples described herein are not inherently related to any particular computer or other device. Various general-purpose systems may be used in accordance with the teachings described herein, or it may be more convenient to construct more specialized equipment to perform the required method steps. The structures required for these various systems will be as described above.

[0115] The above description is illustrative and not limiting. While this disclosure has been described with reference to specific illustrative examples, it should be recognized that this disclosure is not limited to the examples described. The scope of this disclosure should be determined by reference to the following claims, along with the entire scope of equivalents to which the claims are entitled.

[0116] As used herein, the singular forms “a,” “an,” and “the” are intended to include the plural forms as well, unless the context explicitly indicates otherwise. It will be further understood that, as used herein, the terms “equipped,” “equipped,” “contain,” and / or “contain” specify the presence of a described feature, integer, step, action, element, and / or component, but do not exclude the presence or addition of one or more other features, integers, steps, actions, elements, components, and / or groups thereof. Therefore, the terms used herein are intended solely to describe a particular embodiment and are not intended to be limiting.

[0117] Furthermore, it should be noted that in some alternative implementations, the functions / operations shown may occur in a different order than that shown in the diagrams. For example, two consecutive diagrams may actually be executed substantially sequentially, or sometimes in reverse order, depending on the functions / operations involved.

[0118] Although the operation of the method has been described in a specific order, it should be understood that other operations may be performed between the described operations, the described operations may be timed to occur at slightly different intervals, or the described operations may be distributed within a system that allows for the occurrence of processing operations at various intervals related to the processing.

[0119] Various units, circuits, or other components may be described or claimed as “configured” or “configurable” to perform a task. In such contexts, the terms “configured” or “configurable” are used to mean a structure, indicating that the unit / circuit / component includes a structure (e.g., a circuit) that performs a task while in operation. Thus, a unit / circuit / component may be said to be configured to perform a task, or configurable to perform a task, even when the specified unit / circuit / component is not currently operating (e.g., not turned on). Units / circuits / components used with the words “configured” or “configurable” include hardware, such as circuits, and memory that stores executable program instructions to perform operations. Referring to a unit / circuit / component as “configured” or “configurable” to perform one or more tasks is explicitly intended not to invoke 35 U.S. SC 112, paragraph 6, with respect to that unit / circuit / component. In addition, “configured” or “configurable” may include a general-purpose structure (e.g., a general-purpose circuit) operated by software and / or firmware (e.g., an FPGA or general-purpose processor running the software) to operate in a manner that can perform the task in question. “Configured” may also include adapting a manufacturing process (e.g., semiconductor manufacturing equipment) to produce a device (e.g., an integrated circuit) adapted to implement or perform one or more tasks. “Configured” is expressly intended not to apply to blank media, an unprogrammed processor or an unprogrammed general-purpose computer, or an unprogrammed programmable logic device, programmable gate array or other unprogrammed device unless accompanied by a programmed medium that grants the unprogrammed device the ability to be configured to perform the disclosed function.

[0120] The above description has been made with reference to specific embodiments for illustrative purposes. However, the above exemplary discussion is not intended to be exhaustive or to limit the invention to the exact form disclosed. Many modifications and variations are possible in light of the above teachings. The embodiments have been selected and described to best illustrate the principles of the embodiments and their practical applications, thereby enabling those skilled in the art to best utilize the embodiments and various modifications to suit specific uses to be considered. Accordingly, these embodiments should be considered illustrative rather than restrictive, and the invention should not be limited to the details given herein, but may be modified within the scope of the appended claims and equivalents. [Explanation of Symbols]

[0121] 207A module 500 ECG sensing devices, devices 502 Electrode set, electrodes 502A electrode 502B Electrode 502C electrode 504 Insulator 505 Processing Devices 508 Transceiver 510 memory 510A Lead Synthesis Software Module, Module 510B QT Prediction Software Module, Module 520 Housing 550 computing devices 700 Computing Devices 702 Processing Device 704 Main Memory 706 Static Memory 708 Network Interface Device 710 Video Display Unit 712 Alphanumeric input devices 714 Cursor control device 716 Acoustic signal generation device 718 Data Storage Devices 720 Network 726 ECG instruction, instruction 728 Computer-readable storage media 730 Bus

Claims

1. A mobile electrocardiogram (ECG) sensor, An electrode assembly comprising electrodes, which senses cardiac-related signals when in contact with the user's body and generates an electrical signal representing the sensed cardiac-related signals, A processing device operably coupled to the electrode assembly, A training dataset is provided that includes multiple electrocardiogram (ECG) measurements and corresponding QT interval labels for each of the multiple ECG measurements. To generate a probability vector corresponding to each of the aforementioned multiple ECG measurements, a machine learning (ML) model is used to analyze each of the aforementioned multiple ECG measurements. For each of the multiple probability vectors, The aforementioned probability vector is compared with the corresponding QT interval label. The ML model is updated based on a loss function which is the sum of the cross-entropy of the probability vector and the corresponding QT interval label and the root mean square error of the average QT interval. The ML model is trained by performing the following: To predict a 12-read QT interval (QTc) value based on a sensed cardiac-related signal with fewer than 12 leads, the machine learning module is provided with the sensed cardiac-related signal. Analyze the predicted QTc values ​​to determine whether a health abnormality is present. Processing device and, A housing including the electrode assembly and the processing device A mobile electrocardiogram (ECG) sensor equipped with [features / equipment].

2. The mobile ECG sensor according to claim 1, wherein the plurality of ECG measurements in the training dataset include ECG measurements from a plurality of users, and the ML model is trained to predict the 12-read QTc value for a single user.

3. The mobile ECG sensor according to claim 1, wherein the plurality of ECG measurements in the training dataset include ECG measurements from a single user, and the ML model is trained to predict the 12-read QTc value for a single user.

4. The mobile ECG sensor according to claim 1, wherein the ML model is a deep neural network.

5. The mobile ECG sensor according to claim 1, wherein the sensed cardiac-related signals include a lead I signal and a lead II signal.

6. The mobile ECG sensor according to claim 1, wherein when QTc prolongation is detected, it is determined that the health abnormality is present.

7. The mobile ECG sensor according to claim 1, wherein the processing device further transmits a notification to a client device in response to the presence of the health abnormality.

8. A mobile electrocardiogram (ECG) system, An electrode assembly comprising electrodes, which senses cardiac-related signals when in contact with the user's body and generates an electrical signal representing the sensed cardiac-related signals, A processing device operably coupled to the electrode assembly, A training dataset is provided that includes multiple electrocardiogram (ECG) measurements and corresponding QT interval labels for each of the multiple ECG measurements. To generate a probability vector corresponding to each of the aforementioned multiple ECG measurements, a machine learning (ML) model is used to analyze each of the aforementioned multiple ECG measurements. For each of the multiple probability vectors, The aforementioned probability vector is compared with the corresponding QT interval label. The ML model is updated based on a loss function which is the sum of the cross-entropy of the probability vector and the corresponding QT interval label and the root mean square error of the average QT interval. The ML model is trained by performing the following: To predict a 12-read QT interval (QTc) value based on a sensed cardiac-related signal with fewer than 12 leads, the machine learning module is provided with the sensed cardiac-related signal. Analyze the predicted QTc values ​​to determine whether a health abnormality is present. Processing device and, A display operably connected to the aforementioned processing device, The processing device has a memory containing instructions for processing the detected heart-related signals and displaying the heart-related signals on the display. A mobile electrocardiogram (ECG) system equipped with [features / equipment].

9. The mobile ECG system according to claim 8, wherein the plurality of ECG measurements in the training dataset include ECG measurements from a plurality of users, and the ML model is trained to predict the 12-read QTc value for a single user.

10. The mobile ECG system according to claim 8, wherein the plurality of ECG measurements in the training dataset include ECG measurements from a plurality of users, and the ML model is trained to predict the 12-read QTc value for a single user.

11. The mobile ECG system according to claim 8, wherein the ML model is a deep neural network.

12. The mobile ECG system according to claim 8, wherein the sensed cardiac-related signals include a lead I signal and a lead II signal.

13. The mobile ECG system according to claim 8, wherein the presence of the health abnormality is determined when QTc prolongation is detected.

14. The mobile ECG system according to claim 8, wherein the processing device further transmits a notification to a client device in response to the presence of the health abnormality.

15. A method performed by a computer, The steps include receiving cardiac-related signals sensed by the electrode assembly from the user's body, from an electrode assembly equipped with electrodes, The steps include generating an electrical signal representing the sensed cardiac-related signal, Depending on the processing device, The steps include providing a training dataset that includes multiple electrocardiogram (ECG) measurements and corresponding QT interval labels for each of the multiple ECG measurements, To generate a probability vector corresponding to each of the aforementioned multiple ECG measurements, a machine learning (ML) model is used to analyze each of the aforementioned multiple ECG measurements. For each of the multiple probability vectors, The aforementioned probability vector is compared with the corresponding QT interval label. The ML model is updated based on a loss function which is the sum of the cross-entropy of the probability vector and the corresponding QT interval label and the root mean square error of the average QT interval. The steps include training the ML model by performing the following, The steps include providing the cardiac-related signals to a machine learning module in order to predict a 12-read QT interval (QTc) value based on the sensed cardiac-related signals, which have fewer than 12 leads, The processing device performs the steps of analyzing the predicted QTc value to determine whether a health abnormality is present. A method that includes this.

16. The method according to claim 15, wherein the plurality of ECG measurements in the training dataset include ECG measurements from a plurality of users, and the ML model is trained to predict the 12-read QTc value for a single user.

17. The method according to claim 15, further comprising the step of determining that the health abnormality is present when QTc prolongation is detected.

18. The method according to claim 15, further comprising the step of sending a notification to a client device in response to the presence of the aforementioned health abnormality.

19. The mobile ECG sensor according to claim 6, wherein the QTc extension corresponds to an increase in QTc that exceeds a threshold amount and is within a threshold time.

20. The mobile ECG system according to claim 13, wherein the QTc extension corresponds to an increase in QTc that exceeds a threshold amount and is within a threshold time.

21. The method according to claim 17, wherein the QTc extension corresponds to an increase in QTc that exceeds a threshold amount and is within a threshold time.

Images