Patent Grant
10729351
The teachings herein relate to an automated electrocardiography (ECG) analysis system. More particularly, the teachings herein relate to systems and methods displaying quantitative indicators for an ST-T interval of an ECG waveform during measurement of the ECG waveform. These systems and methods use the harmonic signals of a conventional ECG waveform and previously recorded data from normal and abnormal patients to calculate indicators for an ST-T interval of an ECG waveform.
The systems and methods herein can be performed in conjunction with a processor, controller, or computer system, such as the computer system of
Non-ST Segment Elevation Myocardial Infarction Problem
Clinically, the current measuring method of myocardial infarction still requires elevation or depression ST segment in traditional ECG as the diagnostic criteria, and it has been used for more than half a century. Among all heart-related diseases, CAD, AMI, and ACS attack rates amount to 60% to 70%. Noninvasive ECG examination is the only most common, convenient, and wide-spread method to detect arrhythmia, coronary vascular heart disease and all other cardiac disease; however, low detection rate of CAD and AMI by such examination is the major problem to be dealt with in the world at present.
According to many official sources, the waveforms of ST segment in ECG fall an additional 50% following 70% changes with myocardial ischemia. (See
ECG itself is a “qualitative” diagnosis method based on changes of waveform morphology patterns. However, many characteristics of morphological waveform stay unchanged or have no patterns. In the case that the waveform stays unchanged at ST segment, there is a need to establish a standard for Myocardial Infarction without changes of ST segment.
ECG Quantitative Problem
Since the development of ECG, the electrophysiological signal separation of specific time periods of cardiac self-conduction system within the P wave, the QRS complex and the T wave has never been achieved. There is standardized data available in cardiac science, but such quantitative data has never been used in ECG for many reasons. Conventionally, only the character of an ECG waveform is read. As a result, only qualitative information is provided in clinical applications.
ECG is a noninvasive electrophysiological technology. It is able to scan and record cardiac electrical conduction signals. It is the only tracing image of cardiac “bioelectric conduction”, which is the character marker of the life of a heart. However, in the medical field, ECG is only a morphological technique, and the target of its reading, analyzing and determining is waveform character. It is a qualitative technology that does not have quantitative data, which is due to the ECG waveform being associated with deformation and instability. The standard measuring points frequently disappear, hide, overlap and shift. Consequently, they are frequently not shown in ECG, and thus cannot be used for making a determination.
Moreover, conventional ECG cannot measure all digital parameters. The established standards cannot be applied clinically, and cannot be measured. As a result, ECG is unable to achieve a data-based quantitative application. Hence, to date, ECG is still a qualitative application. The foregoing is the reason for which ECG is deemed as a technology that needs experience. However, it is noted that such experience needs to be accumulated from a great number of cases with incorrect diagnosis.
The heart is the most important organ in human body. In addition to heart diseases, a variety of other diseases may also cause abnormalities in heart. Hence, not only cardiologists, but also all doctors in other departments need to read ECG. However, the morphological character changes in ECG cause difficulties for doctors to read ECG. Hence, the issue of how to use the ECG data has confused clinical practices for many years. In traditional ECG, diagnosis of diseases is still made according to morphological character changes in waveforms.
As a result, systems and methods are needed to add quantitative scientific indicators to ECG. Such systems and methods are needed to allow doctors to understand and utilize the knowledge saved in traditional ECG, as well as to reduce learning difficulties, reduce guesses in case diagnosis, reduce misjudgment rate, and improve reliability, diagnosis rate and accuracy, whenever waveform character changes.
ECG Accuracy Problem
Since the first ECG instrument was invented in 1903, its accuracy rate for diagnosis has always been a problem in clinical applications. For people with abnormal conditions, ECG waveform variations are not the same for the same person and are not completely identical even for the same disease. They are at most self-similar. Self-similarly, for example, refers to an object having a shape that is similar to the shape of one of its parts. As a result, ECG science is one of the most complicated disciplines in medicine.
It can be seen from numerous signal processing methods that, during a lifetime, each beat of a person's heart has different specific signal variation, and the difference is significant. However, generally one is unable to observe this from a conventional linear ECG waveform with the naked eye.
Since computers started to be widely used in ECG analysis the 1970s, people have been consistently exploring, searching, and studying how to automate ECG analysis and diagnosis. In the past half a century, thousands of scholars have made efforts in studying algorithms, exploring pattern recognition, and applying those in ECG mapping and automatic diagnosis.
However, wide clinical use of such systems has yet to be achieved. There are at least three technical reasons for this.
1. The ECG waveform is morphological, and generally no consistent mapping points can be found. In other words, the information in the ECG waveform is conveyed through its structure or form. Also, the waveform is abstractly self-similar. In particular, there is no rule for abnormal variations, the time axis signals interfere with each other on left and right sides of as well as above the x-axis, non-linear variations are invisible, and the same disease may have hundreds of millions of variations, but they are not clearly displayed on the ECG waveform. As a result, all ECG parameters are, in general, not accurate, and it is almost impossible to measure these parameters after the waveform changes. Therefore, the highest accuracy of automatic diagnosis by existing ECG software reaches around 38%. Also, this accuracy is only achieved for simple ECG waveform variations and not for many complex waveforms. This is because no mapping point can be found due to the loss or disappearance or deformation of the P-QRS-T waveform.
2. The second reason systems for automated ECG analysis and diagnosis have not been adopted clinically is related to how a conventional ECG waveform has been measured. As described in the '204 Patent and below, the conventional ECG waveform is a single time domain waveform that represents a combination of many different frequency domain signals from different parts of the heart muscle. As a result, information specific to these different parts of the heart muscle are generally lost. In addition, the conventional ECG waveform is a linear waveform, while the heart is a nonlinear system, and the vast majority of variations as a result of abnormality are nonlinear.
3. The third reason systems for automated ECG analysis and diagnosis have not been adopted clinically is related to the high number of false positives found in normal and abnormal populations. For example, in many cases, conventional ECG waveforms show abnormal results in tests of normal people and also show normal results in tests of abnormal people, which makes it extremely difficult for clinical reading and understanding and makes it impossible to determine whether a result is normal or abnormal.
However, the heart is an electrified organ, and there is no doubt that the electrophysiological responses of a heart organ are the fastest and most sensitive measurements to diagnose heart problems. ECG remains one of the most extensively used clinical tools used at present along with blood tests and imaging, despite the lack of accurate systems for automated ECG analysis and diagnosis. As a result, there is a significant need for such systems.
Recent advancements have addressed one of the three technical problems. This is the conventional ECG waveform measurement problem. As described in the '204 Patent and below, an ECG device has been developed that uses signal processing to detect one or more subwaveforms within the P, Q, R, S, T, U, and J waveforms of a conventional ECG waveform and/or within the intervals between the P, Q, R, S, T, U, and J waveforms of a conventional ECG waveform. In other words, the device of the '204 Patent can provide information (subwaveforms) about different frequency domain signals from different parts of the heart muscle. A waveform displaying these subwaveforms is referred to as a saah ECG waveform, for example. In
As described in the '930 Patent and below, one way the different frequency domain signals from different parts of the heart muscle can be measured is through multi-domain ECG. In multi-domain ECG heart signals are measured using different frequency bands. These multi-domain ECG heart signals can be displayed in one diagram as an electrophysiocardiogram (EPCG) waveform.
As a result of the systems of the '204 Patent and the '930 Patent, the technical problem of measuring the different frequency domain signals from different parts of the heart muscle has been addressed.
ECG Parameter Measurement Problem
The heart beats day and night from the first day of a human life to the last day of life. In a whole life, the heart beats about 2.5 billion to 3 billion times. In this regard, it could be calculated that the heart each time pumps 80 ml blood. Accordingly, based on the fact that the heart beats about 70 times per minute on average, the heart hence pumps 8,000 liters of blood per day, which is equivalent to the volume of 40 barrels of gasoline, and the total weight would be 8 tons. Therefore, the heart pumps 3,000 tons of blood per year. If a person lives for 80 years, the number will reach 240,000 tons. It is noted that after the age of 60 years old, a person has a 45% chance of having the condition of arrhythmia, and about half of those cases become life threatening. According to a variety of different scientific predictions, some scientists believe that it is reasonable to predict that the average lifespan of a person is about 80 years old.
The heart is a charged elastic mucus organ, so an ECG instrument is the only instrument that is able to scan and record the physiological and pathological signs of the cardiac electrophysiology (heart ultrasound provides hemodynamic data, CT & MIR provide histological imaging data). ECG provides electrophysiological signals, in particular noninvasive electrophysiological data. It is able to scan and record cardiac electrical conduction signals. By far, it is the only tool that can record the scanning image of “bioelectric conduction,” the identifier of a living heart.
On the other hand, however, in the medical field, ECG is also only a morphological signal. It needs to be read and analyzed to determine its various waveforms. In this regard, it is an area that highly relies on a practitioner's experience. In the history of ECG, there are many data-based parameter gold standards, such as P-R interval, Q-T interval, ST segment, QRS complex, P-J interval, J-T interval, VAT.
However, due to the fact that ECG waveforms are prone to certain issues such as deformation, instability, standard point loss, and so on, conventional ECG instruments are unable to accurately measure these ECG parameters. As a result, many established standards cannot be applied in clinical practice.
Furthermore, as for the data measured manually, a very small variation can result in a difference of tens of milliseconds. Clinically, only simple standards can be used at present, such as: HR, RR interval, PP interval. As a result, only a few very simple standards can be used in current clinical practice, including HR, RR interval, PP interval, etc.
As described above, the ECG systems of the '204 Patent and the '930 Patent have addressed the technical problem of accurately measuring the different frequency domain signals from different parts of the heart muscle.
Additional systems and methods, however, are needed to accurately measure ECG parameters such as the P-R interval, Q-T interval, ST segment, QRS complex, P-J interval, J-T interval, and VAT so that these standards can be used in clinical practice.
ECG History
Electrical signals produced by a human heart were observed through electrodes attached to a patient's skin as early as 1879. Between 1897 and 1911 various methods were used to detect these electrical signals and record a heartbeat in real-time. In 1924, Willem Einthoven was awarded the Nobel Prize in medicine for identifying the various waveforms of a heartbeat and assigning the letters P, Q, R, S, T, U, and J to these waveforms. Since the early 1900s, the equipment used for electrocardiography (ECG or EKG) has changed. However, the basic waveforms detected and analyzed have not changed.
An ECG device detects electrical impulses or changes in the electrical potential between two electrodes attached to the skin of a patient as the heart muscle contracts or beats. Electrically, the contraction of the heart is caused by depolarization and repolarization of various parts of the heart muscle. Initially, or at rest, the muscle cells of the heart have a negative charge. In order to cause them to contract, they receive an influx of positive ions Na+ and Ca++. This influx of positive ions is called depolarization. The return of negative ions to bring the heart back to a resting state is called repolarization. Depolarization and repolarization of the heart affect different parts of the heart over time giving rise to the P, Q, R, S, T, U, and J waveforms.
PR segment 220 of
Waveforms Q 230, R 240, and S 250 of
ST segment 260 of
The point in
T waveform 270 of
Not shown in
As shown in
Artificial Intelligence
Artificial Intelligence (AI) generally refers to languages, algorithms, and operating systems that relate to how a computer system can carry out tasks that were previously only completed by relying on human intelligence. It is a general term and often does not include implementation or application. The definition of AI has evolved over time, however, and this phenomenon is referred to as the “AI effect.” The AI effect can be summarized as the prescription that “AI intends to complete a collection of all tasks that cannot be implemented without relying on human intelligence at the present.” In the 1940s and 1950s, a group of scientists from different fields (mathematics, psychology, engineering, economics and politics) began to explore the possibility of manufacturing an artificial brain. In 1956, AI was established as a discipline. The organizers of the 1956 Dartmouth Artificial Intelligence Conference were Marvin Minsky, John McCarthy, and two other senior scientists, Claude Shannon and Nathan Rochester, with the latter coming from IBM. At the 1956 Dartmouth Artificial Intelligence Conference, the name and tasks of AI were determined, and at the same time, initial achievements and the earliest group of researchers appeared. As a result, this event has been extensively acknowledged as a sign of the birth of AI. It is clear that AI is now a technological field, a second revolution since the invention of the computer, and a certain trend in the future. It is being applied in all industries, exists everywhere, and is used on almost everything on the earth. In the medical field, AI is now used in the following: medical imaging, sensor-based data analysis, conversion of bioinformatics, and development of public health policies. AI is also used in the clinical applications. These applications include cancer treatment: recognition of mitosis of cancerous tumor cells, identification of disease types and degrees of aggravation, shortening chemotherapy time, and mitigating damage caused by chemotherapy for cancer patients. These applications also include ophthalmological diagnosis: recognition of early signs of eye disease, such as senile macular degeneration, and diabetic retinopathy and surgical treatment: AI surgical robots, etc. Google has also formed a team called DeepMind Health, which cooperated with the Imperial College London and the Royal Free Hospital in London, UK. They released a mobile application called Streams, and medical professionals can use Streams to observe treatment results in a faster manner. Overall, in the medical field, the AI system can be used on any job that previously required human thinking.
Before one or more embodiments of the invention are described in detail, one skilled in the art will appreciate that the invention is not limited in its application to the details of construction, the arrangements of components, and the arrangement of steps set forth in the following detailed description. The invention is capable of other embodiments and of being practiced or being carried out in various ways. Also, it is to be understood that the phraseology and terminology used herein is for the purpose of description and should not be regarded as limiting.
Computer-Implemented System
Computer system 100 may be coupled via bus 102 to a display 112, such as a cathode ray tube (CRT) or liquid crystal display (LCD), for displaying information to a computer user. An input device 114, including alphanumeric and other keys, is coupled to bus 102 for communicating information and command selections to processor 104. Another type of user input device is cursor control 116, such as a mouse, a trackball or cursor direction keys for communicating direction information and command selections to processor 104 and for controlling cursor movement on display 112. This input device typically has two degrees of freedom in two axes, a first axis (i.e., x) and a second axis (i.e., y), that allows the device to specify positions in a plane.
A computer system 100 can perform the present teachings. Consistent with certain implementations of the present teachings, results are provided by computer system 100 in response to processor 104 executing one or more sequences of one or more instructions contained in memory 106. Such instructions may be read into memory 106 from another computer-readable medium, such as storage device 110. Execution of the sequences of instructions contained in memory 106 causes processor 104 to perform the process described herein. Alternatively, hard-wired circuitry may be used in place of or in combination with software instructions to implement the present teachings. Thus implementations of the present teachings are not limited to any specific combination of hardware circuitry and software.
In various embodiments, computer system 100 can be connected to one or more other computer systems, like computer system 100, across a network to form a networked system. The network can include a private network or a public network such as the Internet. In the networked system, one or more computer systems can store and serve the data to other computer systems. The one or more computer systems that store and serve the data can be referred to as servers or the cloud, in a cloud computing scenario. The other computer systems that send and receive data to and from the servers or the cloud can be referred to as client or cloud devices, for example.
The term “computer-readable medium” as used herein refers to any media that participates in providing instructions to processor 104 for execution. Such a medium may take many forms, including but not limited to, non-volatile media, volatile media, and transmission media. Non-volatile media includes, for example, optical or magnetic disks, such as storage device 110. Volatile media includes dynamic memory, such as memory 106. Transmission media includes coaxial cables, copper wire, and fiber optics, including the wires that comprise bus 102.
Common forms of computer-readable media or computer program products include, for example, a floppy disk, a flexible disk, hard disk, magnetic tape, or any other magnetic medium, a CD-ROM, digital video disc (DVD), a Blu-ray Disc, any other optical medium, a thumb drive, a memory card, a RAM, PROM, and EPROM, a FLASH-EPROM, any other memory chip or cartridge, or any other tangible medium from which a computer can read.
Various forms of computer-readable media may be involved in carrying one or more sequences of one or more instructions to processor 104 for execution. For example, the instructions may initially be carried on the magnetic disk of a remote computer. The remote computer can load the instructions into its dynamic memory and send the instructions over a telephone line using a modem. A modem local to computer system 100 can receive the data on the telephone line and use an infra-red transmitter to convert the data to an infra-red signal. An infra-red detector coupled to bus 102 can receive the data carried in the infra-red signal and place the data on bus 102. Bus 102 carries the data to memory 106, from which processor 104 retrieves and executes the instructions. The instructions received by memory 106 may optionally be stored on storage device 110 either before or after execution by processor 104.
In accordance with various embodiments, instructions configured to be executed by a processor to perform a method are stored on a computer-readable medium. The computer-readable medium can be a device that stores digital information. For example, a computer-readable medium includes a compact disc read-only memory (CD-ROM) as is known in the art for storing software. The computer-readable medium is accessed by a processor suitable for executing instructions configured to be executed.
The following descriptions of various implementations of the present teachings have been presented for purposes of illustration and description. It is not exhaustive and does not limit the present teachings to the precise form disclosed. Modifications and variations are possible in light of the above teachings or may be acquired from practicing of the present teachings. Additionally, the described implementation includes software but the present teachings may be implemented as a combination of hardware and software or in hardware alone. The present teachings may be implemented with both object-oriented and non-object-oriented programming systems.
Subwaveform Detection of the P, Q, R, S, T, U, and J Waveforms
As described above, electrical signals flow through all the different muscle tissues of the heart. For the last 100 years, conventional ECG devices have been able to detect some of these signals in the form of the P, Q, R, S, T, U, and J waveforms. These waveforms have aided in the diagnosis and treatment of many heart problems.
Unfortunately, however, the P, Q, R, S, T, U, and J waveforms do not provide a complete picture of the operation of all the different muscle tissues of the heart. As a result, improved systems and methods are needed to detect and analyze more information from the electrical signals that flow through all the different muscle tissues of the heart as it is beating. This additional information can be used to diagnose and treat many more heart problems.
In various embodiments, additional information is obtained from the electrical signals produced by a heart through signal processing. More specifically, signal processing is added to an ECG device in order to detect more information from the electrical signals that flow through all the different muscle tissues of the heart as it is beating.
A voltage signal is detected between two electrodes 710 by detector 720. Detector 720 also typically amplifies the voltage signal. Detector 720 can also convert the voltage signal to a digital voltage signal using an analog to digital converter (A/D).
Detector 720 provides the detected and amplified voltage signal from each pair of electrodes 710 to display 730. Display 730 can be an electronic display device including, but not limited to, a cathode ray tube (CRT) device, light emitting diode (LED) device, or Liquid crystal display (LCD) device. Display 730 can also be a printer device. Additionally, display 730 can include a memory device to record detected signals. The memory device can be, but is not limited to, a volatile electronic memory, such as random access memory (RAM), a non-volatile electronic memory, such as electrically erasable programmable read-only memory (EEPROM or Flash memory), or a magnetic hard drive.
Display 730 displays a continuous loop of the detected P, Q, R, S, T, U, and J waveforms as shown in
A voltage signal is detected between two electrodes 810 by detector 820. Detector 820 also amplifies the voltage signal. Detector 820 also converts the voltage signal to a digital voltage signal using an analog to digital converter (A/D).
Detector 820 provides the detected and amplified voltage signal from each pair of electrodes 810 to signal processor 830. Detector 820 can also provide the detected and amplified voltage signal from each pair of electrodes 810 directly to display device 840 to display the conventional P, Q, R, S, T, U, and J waveforms.
Signal processor 830 detects or calculates one or more subwaveforms within and/or in the interval between the P, Q, R, S, T, U, and J waveforms of each detected and amplified voltage signal. A waveform is a shape or form of a signal. A subwaveform is shape or form of a signal that is within or part of another signal.
Signal processor 830 can be a separate electronic device that can include, but is not limited to, an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), or a general purpose processor. Signal processor 830 can be software implemented on another processor of the ECG device, such as a processor of display device 840. Signal processor 830 can also be a remote server that receives the detected and amplified voltage signal from detector 820, detects or calculates one or more subwaveforms within and/or in the interval between the P, Q, R, S, T, U, and J waveforms, and sends the detected and amplified voltage signal and the one or more subwaveforms to display device 840.
Signal processor 830 sends one or more subwaveforms of each detected and amplified voltage signal to display device 840. Signal processor 830 can also calculate and send to the display device 840 the time periods of the one or more subwaveforms, the times between the one or more subwaveforms, and the times of the one or more subwaveforms in relation to the P, Q, R, S, T, U, and J waveforms and or the intervals between the P, Q, R, S, T, U, and J waveforms. Signal processor 830 can also compare this timing information to stored normal timing information. Based on the comparison, signal processor 830 can determine differences from the normal data and send this difference information and any of the timing information to display device 840.
Display device 840 displays a continuous loop of the one or more subwaveforms for each pair of electrodes 810. Display device 840 can also display part or all of the conventional P, Q, R, S, T, U, and J waveforms for comparison with the one or more subwaveforms. Like display 730 of
In various embodiments, an ECG device using signal processing to detect one or more subwaveforms within the P, Q, R, S, T, U, and J waveforms and/or within the intervals between the P, Q, R, S, T, U, and J waveforms is herein referred to as a saah ECG device. The voltage difference signals produced by a saah ECG device are referred to as saah ECG waveforms. The term “saah” is an acronym for some of the anatomically distinct portions of muscle tissue that produce subwaveforms. Specifically, saah stands for sinoatrial node (SAN), atria (right atrium (RA) and left atrium (LA)), atrioventricular node (AVN), and bundle of His (HIS). However, a saah ECG waveform is not limited to including subwaveforms representing the SAN, the atria, the AVN, and the HIS. A saah ECG waveform can include any subwaveform the P, Q, R, S, T, U, and J waveforms and/or within the intervals between the P, Q, R, S, T, U, and J waveforms.
In various embodiments, the subwaveforms of a saah ECG waveform are detected using signal processing. Electrodes 810 of the saah ECG of
The signal processing can be performed directly on the time domain signal received from a detector or the time domain signal received from a detector can be converted to the frequency domain for algorithmic processing. In
As described above, through animal and/or human experimentation, the frequency bands associated with depolarization of the SAN, atria, AVN, HIS, and BB of the heart are determined. Based on these frequency bands, band pass filters are created. Blocks 1041-1045 represent the band pass filters created to filter the PR interval frequency domain signal for frequency bands of the SAN, atria, AVN, HIS, and BB of the heart, respectively.
In block 1050, the frequency domain subwaveforms detected from the band pass filtering the frequency bands of the SAN, atria, AVN, HIS, and BB of the heart are summed. In block 1060, the filtered and summed frequency domain signal of the PR interval is converted back to a time domain signal. The frequency domain signal is converted into a time domain signal using a Fourier transform, for example.
The PR interval filtered and summed time domain signal 1070 includes five time domain subwaveforms 910-950. Subwaveforms 910-950 represent depolarization of the SAN, atria, AVN, HIS, and BB of the heart, respectively. Time domain signal 1070 can be used to replace PR interval 1020 in ECG waveform 200, for example. As a result, a saah ECG waveform is produced.
In block 1130, the detected and amplified heart signals are processed using a signal processor. The signal processor detects the conventional P, Q, R, S, T, U, and J waveforms and sends them to the display of block 1160. The signal processor also detects or calculates subwaveforms within the conventional P, Q, R, S, T, U, and J waveforms and/or within intervals between the conventional P, Q, R, S, T, U, and J waveforms. The signal processor sends the subwaveforms to block 1140 for further processing. The processor of block 1140 produces the saah ECG waveform that includes the subwaveforms and sends the saah ECG waveform to the display of block 1160. The processor of block 1140 calculates additional information or new data from the saah ECG waveform. This new data can include, but is not limited to, timing information about the subwaveforms, timing information about the intervals between the subwaveforms, and timing information about the subwaveforms and their relation to the conventional P, Q, R, S, T, U, and J waveforms. In block 1150, this new data is sent to the display of block 1160.
The display of block 1160 displays a continuous loop of the conventional ECG waveform, the saah ECG waveform, and the new data from the subwaveforms. The display of block 1160 can display this information on an electronic display or print it on paper. The display of block 1160 can also record this information. The display of block 1160 can record this information on any type of memory device.
Plot 1200 also shows new data or timing information about the subwaveforms and their relation to the conventional P, Q, R, S, T, U, and J waveforms. For example, the time interval between line 1231 and line 1232 relates subwaveform A of saah ECG waveform 1220 to P waveform 1240 of conventional ECG waveform 1210. The time interval between line 1232 and line 1233 relates subwaveforms B and C of saah ECG waveform 1220 to P waveform 1240 conventional ECG waveform 1210. The time interval between line 1233 and line 1234 relates subwaveforms D and E of saah ECG waveform 1220 to PR segment 1250 conventional ECG waveform 1210.
In block 1330, the sampled heart signals are processed using a signal processor. The signal processor produces four different types of information from the sampled heart signals. As shown in block 1340, the signal processor produces conventional ECG waveforms including the conventional P, Q, R, S, T, U, and J waveforms and sends them to display 1380. As shown in block 1350, the signal processor produces saah ECG waveforms. These saah ECG waveforms include subwaveforms of the conventional P, Q, R, S, T, U, and J waveforms and the intervals between them. Note that the arrow between blocks 1330 and 1350 show information following in both directions. This shows that information from the saah ECG waveforms is further analyzed by the signal processor.
As shown in block 1360, the signal processor further analyzes the saah ECG waveforms to produce saah ECG data. This saah ECG data is sent to display 1380. Additionally, as shown in block 1370, the signal processor further analyzes the saah to obtain endocardium and epicardium data. This data is compared to recorded normal and abnormal data. The signal processor then produces automatic pattern recognition diagnosis (APD) information, and this information is sent to display 1380. APD information is, for example, patterns and/or colors that allow a user to easily and quickly determine that normal or abnormal endocardium and/or epicardium data was found.
Systems and methods for detecting ECG subwaveforms are described in the '204 Patent, which is incorporated by reference in its entirety.
System for Detecting ECG Subwaveforms
In various embodiments, an electrocardiography (ECG) system for detecting one or more subwaveforms within the P, Q, R, S, T, U, and J waveforms or in an interval between the P, Q, R, S, T, U, and J waveforms is provided. Returning to
Two or more electrodes 810 are placed near a beating heart and receive electrical impulses from the beating heart. Two or more electrodes 810 are shown in
Detector 820 is electrically connected to two or more electrodes 810. Detector 820 detects the electrical impulses from at least one pair of electrodes of the two or more electrodes 810. Detector 820 converts the electrical impulses to an ECG waveform for each heartbeat of the beating heart. Detector 820, for example, samples the electrical impulses. In various embodiments, detector 820 further amplifies the ECG waveform. In various embodiments, detector 820 further performs analog to digital (A/D) conversion on the ECG waveform. In various embodiments, detector 820 provides an ECG waveform with a higher signal-to-noise (S/N) ratio than conventional ECG devices.
Signal processor 830 is electrically connected to detector 820. Signal processor 830 receives the ECG waveform from detector 820. Signal processor 830 detects or calculates one or more subwaveforms within P, Q, R, S, T, U, and J waveforms of the ECG waveform or in an interval between the P, Q, R, S, T, U, and J waveforms that represent the depolarization or repolarization of anatomically distinct portions of muscle tissue of the beating heart. Signal processor 830 produces a processed ECG waveform that includes the one or more subwaveforms for each heartbeat.
Signal processor 830 can be a separate device, can be software running on a device of detector 820 or display device 840, or can be software running on a remote server and communicating with detector 820 and display device 840 through one or more communication devices. Signal processor 830 can be a separate device that includes, but is not limited to, an application specific integrated circuit (ASIC) or a field programmable gate array (FPGA) or a general purpose processor. A general purpose processor can include, but is not limited to, a microprocessor, a micro-controller, or a computer such as the system shown in
Display device 840 receives the processed ECG waveform for each heartbeat and displays the processed ECG waveform for each heartbeat. The processed ECG waveform is called a saah ECG waveform, for example. As described above, display device 840 can be an electronic display device including, but not limited to, a cathode ray tube (CRT) device, light emitting diode (LED) device, or Liquid crystal display (LCD) device. Display device 840 can also be a printer device or any combination of an electronic display device and a printer. Additionally, display device 840 can include a memory device to record saah ECG waveforms, saah ECG data and conventional ECG waveforms and data. The memory device can be, but is not limited to, a volatile electronic memory, such as random access memory (RAM), a non-volatile electronic memory, such as electrically erasable programmable read-only memory (EEPROM or Flash memory), or a magnetic hard drive.
In various embodiments, the detected one or more subwaveforms include at least one subwaveform representing depolarization of the sinoatrial node (SAN), the atria (right atrium (RA) and left atrium (LA)), the atrioventricular node (AVN), the bundle of His (HIS), or the bundle branches (BB) of the beating heart.
In various embodiments, the display device 840 further displays the ECG waveform for comparison with the processed ECG waveform.
In various embodiments, signal processor 830 further calculates timing information about the one or more subwaveforms, timing information about the intervals between the one or more subwaveforms, and timing information about the one or more subwaveforms and their relation to the P, Q, R, S, T, U, and J waveforms of the ECG waveform for each heartbeat. Display device 840 further receives this timing information from signal processor 830. Display device 840 displays the timing information about the one or more subwaveforms, the timing information about the intervals between the one or more subwaveforms, and the timing information about the one or more subwaveforms and their relation to the P, Q, R, S, T, U, and J waveforms of the ECG waveform for each heartbeat.
In various embodiments, the ECG system further includes a memory device (not shown). The memory device receives the ECG waveform and the processed ECG waveform from the signal processor.
In various embodiments, the memory device further includes normally processed ECG waveform data. Normally processed ECG waveform data is stored on the memory device using signal processor 830 or a general-purpose processor (not shown). Signal processor 830 further compares the processed ECG waveform to the normally processed ECG waveform data and calculates a status condition based on the comparison. The status conditions are, for example, normal, suspicious, or abnormal.
In various embodiments, the ECG system includes a second display device (not shown) surrounding a rotating button (not shown). Signal processor 830 further sends a colored pattern to the second display device based on the status condition. The second display device provides automatic pattern recognition diagnosis (APD).
Method for Detecting ECG Subwaveforms
In step 1610 of method 1600, electrical impulses are detected between at least one pair of electrodes of two or more electrodes placed proximate to a beating heart using a detector. The electrical impulses are converted to an ECG waveform for each heartbeat of the beating heart using the detector.
In step 1620, the ECG waveform for each heartbeat is received from the detector using a signal processor. One or more subwaveforms within P, Q, R, S, T, U, and J waveforms of the ECG waveform for each heartbeat or in an interval between the P, Q, R, S, T, U, and J waveforms that represent the depolarization or repolarization of an anatomically distinct portion of muscle tissue of the beating heart are detected using the signal processor. A processed ECG waveform that includes the one or more subwaveforms for each heartbeat is produced using the signal processor.
In step 1630, the processed ECG waveform is received from the signal processor and the processed ECG waveform is displayed using a display device.
Computer Program Product for Detecting ECG Subwaveforms
In various embodiments, computer program products include a tangible computer-readable storage medium whose contents include a program with instructions being executed on a processor so as to perform a method for detecting subwaveforms within the P, Q, R, S, T, U, and J waveforms of an ECG waveform of a heartbeat or in an interval between the P, Q, R, S, T, U, and J waveforms. This method is performed by a system that includes one or more distinct software modules.
Detection module 1710 detects electrical impulses between at least one pair of electrodes of two or more electrodes placed proximate to a beating heart. Detection module 1710 converts the electrical impulses to an ECG waveform for each heartbeat of the beating heart.
Processing module 1720 receives the ECG waveform for each heartbeat. Processing module 1720 detects one or more subwaveforms within P, Q, R, S, T, U, and J waveforms of the ECG waveform for each heartbeat or in an interval between the P, Q, R, S, T, U, and J waveforms that represent the depolarization or repolarization of an anatomically distinct portion of muscle tissue of the beating heart. Processing module 1720 produces a processed ECG waveform that includes the one or more subwaveforms for each heartbeat.
Display module 1730 receives the processed ECG waveform. Display module 1730 displays the processed ECG waveform.
Automated ECG Analysis and Diagnosis
As described above, to date the accuracy rate of automated ECG analysis and diagnosis systems has been a problem in clinical applications. There are at least three technical reasons for this. 1. The conventional ECG waveform is morphological and generally no consistent mapping points can be found. 2. Conventional ECG measurements have not provided information specific different parts of the heart muscle. 3. Automated ECG waveform analysis has generally resulted in a high number of false positives for both normal and abnormal populations. However, ECG remains one of the most extensively used clinical tools, despite the lack of accurate systems for automated ECG analysis and diagnosis. As a result, there is a significant need for such systems. Recent advancements have addressed the conventional ECG waveform measurement problem. Specifically, the systems of the '204 Patent and the '930 Patent have allowed the different frequency domain signals from different parts of the heart muscle to be measured.
Additional systems, however, are needed to further address the technical problems of analyzing the shape and form of these frequency domain signals and distinguishing disease conditions from false positives in normal and abnormal populations.
Conventional ECG Waveform Analysis Problems
Analysis of conventional ECG waveforms for clinical diagnosis has been limited by a number of problems for more than a century. (1) It has been difficult to correctly confirm the start point of the P wave. The reason for this is that the start point of the P wave is on “a parallel equipotential line,” which needs to be determined. This has traditionally been determined through guessing. If the starting point of each heartbeat cannot be correctly identified, then all subsequent parameter measurements will be wrong. The normal value for the conduction time from the SA node (the starting point of the P wave) to the atrium is only around 30 ms. As a result, a small mistake in the location of the starting point can make a big difference in the measurement of this conduction time.
(2) At the PR interval, it has been difficult to identify the specific PA, AH, or HV intervals within PR interval. When the PR interval is abnormal, in particular, it can only be estimated and cannot be measured. Also, often the end of the PR interval cannot be confirmed, as there is no equipotential and the starting point of QRS wave appears to be an upward arc angle.
(3) It has been difficult to identify a difference between the ST segment of a normal person and the ST segment of an abnormal person. In other words, the ST segment appears to be exactly abnormal for normal people and exactly normal for abnormal people. Also, and the J point often disappears, making it impossible to determine. As a result, the standards for the ST segment often cannot be applied.
In summary, at an abnormal moment, signals of a conventional ECG waveform often shift positions, and the waveform is changed to a different shape, making it difficult or impossible to estimate. If a conventional ECG waveform is changed in such a way and human intelligence or experience is still relied on to diagnose, a large amount of accuracy is lost.
Automated ECG Analysis and Diagnosis Using AI
As described above, additional systems are needed to further address the technical problems of analyzing the shape and form of the frequency domain signals of a conventional ECG waveform and distinguishing disease conditions from false positives in normal and abnormal populations using the conventional ECG waveform.
In various embodiments, these technical problems are addressed by 1. applying artificial intelligence (AI) algorithms to characterize the shape and form of the frequency domain signals of a conventional ECG waveform; 2. comparing the characterized shape and form of the frequency domain signals to a database of characterized signals from normal and abnormal populations using human like AI algorithms and non-human like AI algorithms; and 3. annotating the conventional ECG waveform with diagnosis information based on the comparison.
In step 1802, electrical heart signals are obtained from the two or more electrodes.
In step 1803, the electrical heart signals are detected and amplified.
In step 1804, the amplified signals are processed. For example, the signals from a number of different conventional ECG leads are combined.
In step 1805, the combined signals form a conventional ECG waveform.
In step 1806, the conventional ECG waveform is displayed or printed, for example.
In step 1810, the detected signals of step 1803 are obtained and processed to produce two or more frequency domain signals.
In step 1811, the two or more frequency domain signals are processed for characteristics of cardiac electrophysiological signals using one or more AI algorithms.
In step 1812, the cardiac electrophysiological characteristics of the two or more frequency domain signals are compared to databases of similar cardiac electrophysiological characteristics for normal and abnormal populations using the system of
In step 1820, the cardiac electrophysiological characteristics of the two or more frequency domain signals are compared to the databases using human like AI algorithms. These human like AI algorithms can include, but are not limited to, an expert and cardiologist system 1821, an ECG diagnosis system 1822, an intelligent signals unsupervised feature learning system 1823, a general ECG problem solver 1824, and a semantic ECG waveform system 1825.
In step 1830, the cardiac electrophysiological characteristics of the two or more frequency domain signals are compared to the databases using non-human like AI algorithms. These human like AI algorithms can include, but are not limited to, a neural network algorithm 1831 and a deep learning algorithm 1832. The neural network algorithm 1831 can include a shallow network 1833. The deep learning algorithm 1832 can include a deep network 1834. This deep network 1834 can include, but is not limited to, a convolution all neural network (CNN) 1835, a deep belief net (DBN) 1836, or a restricted Boltzmann machine (RBM).
In step 1840, the results from steps 1820 and 1830 are combined to provide diagnosis information for the conventional ECG waveform.
In step 1806, this diagnosis information is displayed on the conventional ECG waveform.
The system of
In various embodiments, the diagnosis information presented in step 1806 can include, but is not limited to, diagnosis markers or more accurate timing information.
Analysis and Diagnosis System
The systems of the '204 Patent and the '930 Patent have used different signal processing methods to detect the harmonic signals and discontinuity points of a conventional ECG waveform. In various embodiments, artificial intelligence (AI) in conjunction with a database of normal and abnormal ECG data is used to detect the harmonic signals and discontinuity points of a conventional ECG waveform and to annotate cardiac electrophysiological signals in the ECG waveform as normal or abnormal.
A voltage signal is detected between two electrodes 1910 by detector 1920. Detector 1920 also amplifies the voltage signal. Detector 1920 converts the electrical impulses to an ECG waveform for each heartbeat of the beating heart. Detector 1920 converts the voltage signal to a digital voltage signal using an analog to digital converter (A/D), for example. Detector 1920 provides the detected and amplified voltage signal from each pair of electrodes 1910 directly to display device 1940 to display the ECG waveform. The ECG waveform includes conventional P, Q, R, S, T, U, and J waveforms, for example. Detector 1920 also provides the detected and amplified voltage signal from each pair of electrodes 1910 directly to processor 1930.
Processor 1930 can be a separate electronic device that can include, but is not limited to, an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), or a general purpose processor or computer, such as the system of
Processor 1930 receives the ECG waveform for at least one heartbeat from detector 1920. Processor 1930 converts the ECG waveform to a frequency domain waveform. Processor 1930 separates the frequency domain waveform into two or more different frequency domain waveforms. Processor 1930 converts the two or more different frequency domain waveforms into a plurality of time domain cardiac electrophysiological subwaveforms and discontinuity points between these subwaveforms of the ECG waveform.
Processor 1930 compares the plurality of subwaveforms and discontinuity points to a database (not shown) of subwaveforms and discontinuity points for cardiac electrophysiological signals of ECG waveforms for a plurality of known and normal and abnormal patients. Processor 1930 identifies at least one subwaveform or one or more discontinuity points of the plurality of subwaveforms and discontinuity points as a normal or abnormal electrophysiological signal of the ECG waveform based on the comparison.
Display device 1940 is an electronic display device, a printer, or any combination of the two. Display device 1940 displays the ECG waveform for the at least one heartbeat of the beating heart. Display device 1940 also displays one or more markers at the location of the at least one subwaveform or the one or more discontinuity points on the ECG waveform and identifies the one or more markers as a normal or abnormal cardiac electrophysiological signal. For example,
In various embodiments, processor 1930 converts the ECG waveform to a frequency domain waveform, separates the frequency domain waveform into two or more different frequency domain waveforms, and converts the two or more different frequency domain waveforms into a plurality of time domain cardiac electrophysiological subwaveforms and discontinuity points using an artificial intelligence algorithm. The artificial intelligence algorithm includes a multivariable calculus algorithm, for example.
In various embodiments, processor 1930 compares the plurality of subwaveforms and discontinuity points to a database of subwaveforms and discontinuity points using a human like artificial intelligence algorithm.
In various embodiments, the human like artificial intelligence algorithm includes an expert and cardiologist system. This system evaluates the comparison based on morphological rules developed from cardiologists. In other words, the shape differences are compared using rules based on pattern recognition and doctors' experiences
In various embodiments, the human like artificial intelligence algorithm includes an ECG diagnosis system. This system evaluates the comparison based on morphological patterns and their correlation to specific diseases.
In various embodiments, the human like artificial intelligence algorithm includes an intelligent signals unsupervised feature learning system. This system evaluates the comparison based on morphological patterns learned over time by the system.
In various embodiments, the human like artificial intelligence algorithm includes a general ECG problem solver. This system evaluates the comparison based on one or more known conditions or conflicting conditions including, but not limited to, heart failure (HF), atrium fibrillation (AF), atrial conductor block, premature atrial contraction (PAC), premature ventricular condition (PVC), atrial tachycardia, and ventricular tachycardia.
In various embodiments, the human like artificial intelligence algorithm includes a semantic ECG waveform system. This system evaluates the comparison based on additional signal processing rather than pattern recognition.
In various embodiments, processor 1930 compares the plurality of subwaveforms and discontinuity points to a database of subwaveforms and discontinuity points using a non-human like artificial intelligence algorithm.
In various embodiments, the human like artificial intelligence algorithm includes a neural network. The neural network can be a shallow network, for example.
In various embodiments, the human like artificial intelligence algorithm includes a deep learning algorithm. The deep learning algorithm can include a deep network, for example. The deep network can include, but is not limited to, convolution all neural network (CNN), a deep belief net (DBN), or a restricted Boltzmann machine (RBM).
Method for Identifying and Annotating Cardiac Electrophysiological Signals
In step 2010 of method 2000, electrical impulses are received from a beating heart using two or more electrodes.
In step 2020, the electrical impulses are detected from at least one pair of electrodes of the two or more electrodes and converted to an ECG waveform for each heartbeat of the beating heart using a detector.
In step 2030, the ECG waveform for at least one heartbeat is received from the detector, the ECG waveform is converted to a frequency domain waveform, the frequency domain waveform is separated into two or more different frequency domain waveforms, and the two or more different frequency domain waveforms are converted into a plurality of time domain cardiac electrophysiological subwaveforms and discontinuity points between these subwaveforms of the ECG waveform using a processor.
In step 2040, the plurality of subwaveforms and discontinuity points are compared to a database of subwaveforms and discontinuity points for cardiac electrophysiological signals of ECG waveforms for a plurality of known and normal and abnormal patients using the processor.
In step 2050, at least one subwaveform or one or more discontinuity points of the plurality of subwaveforms and discontinuity points are identified as a normal or abnormal electrophysiological signal of the ECG waveform based on the comparison using the processor.
In step 2060, the ECG waveform for the at least one heartbeat of the beating heart is displayed, one or more markers at a location of the at least one subwaveform or one or more discontinuity points is displayed on the ECG waveform, and the one or more markers are identified as a normal or abnormal cardiac electrophysiological signal using a display device.
Endocardium, ST-T Interval and Ventricle Display Methods
The ECG was invented more than 100 years ago. The ECG is a map of electrical signals within the heart. Every doctor and nurse needs to learn and master this basic tool. The existing ECG only consists of the P-QRS-T waveform; of which, P-wave represents the atrium, the QRS represents ventricular depolarization, and the T-wave represents ventricular repolarization. However, the heart simply contains too much information, and this ECG waveform is too simple. Therefore, in a Clinical setting, the ECG is not extensive enough. In various embodiments, systems and methods add several functional modules, allowing for additional quantitative functions without changing the traditional qualitative ECG.
Function Introduction
In addition to the traditional ECG waveform, the aiECG includes the following functions: [1] an artificial intelligence color waveform, [2] artificial intelligence digital data, [3] artificial intelligence assessment and judgment, [4] endocardium & epicardium digital data, [5] ST-T interval digital data, an [6] Evolved Frank-Starling law diagram, and [7] a data cloud system.
In various embodiments, AI data is embedded into the instrument. The functional modules of the instrument have received preliminary clinical application verification through before and after results including the following: cardiac surgery, electrophysiological surgery, drug treatment, pacemaker installation, PCI and stent treatment, animal studies, and so on. These studies proved to be effective, accurate, rarely showed false positives and false negatives, and had high correlation with disease.
Due to its simple operation, portability, economic use, and functions, this instrument can add many practical functions to the traditional ECG waveform. The instrument's main technical parameters include: an the input impedance, acuity 2.5 M Ω, an electrode offset tolerance of 500 mV, defibrillation and isolation protection cable type of VA001DNA/VA001DNI, a reset time 10 s or less, a common mode rejection ratio (CMRR)>100 dB, a patient leakage current of <10 μA, a sensitivity of 2.5/5/10/20, an internal noise of 20 μVp-v or less, interference between ≤−38 dB or less, frequency response of 0.05 to 150 Hz+0.4 dB/−3 dB, and a sample rate: 5,000-80,000 sample/s.
Various embodiments include an endocardium & epicardium mapping module. In existing ECG's, it is not possible to test the endocardium and epicardium electrical potential on the body surface. According to their order from resting potential to action potential, using quantum and spectral methods, the myocardial layer on the time (x) axis of the endocardium & epicardium APD time course is measured using the conducting law of positive and negative potential, in order to obtain the specific potential values.
This embodiment provides body surface scanning of the endocardium and epicardium. The signal processing principle of the original six leads I, II, V3, V4, V5, and V6 uses a square filter (described below) to obtain the membrane potential signals in each lead. Artificial intelligence is then automatically applied to detect normality and abnormality. Endocardium and Epicardium values are then displayed in their corresponding six lead categories ABCDEF.
Various embodiments include a whole ST-T digital number module. A traditional ECG only contains morphological waveforms, only qualitative indicators, and no quantitative indicators. In the diagnosis of the clinical application, the ST segment with elevating or descending is the only basis for assessing CAD, AMI or ACS.
Presently, the biggest deficiency is due to two areas: (1) Many CAD or AMI patients' ST segment does not elevate or descend; and (2) Many normal people show elevating, but the disease shows poor correlation. It has been confirmed from a large set of test data, that the ST-T digital data contains very obvious changes, while the traditional ECG does not display significant changes. The ST-T interval's raw data is divided into 16 quadrants in the time-course (x) axis; the myocardial potential and ST-T waveform shape has a fixed correlation ranging slowly from low to high, to T peak, and then rapidly falls from high to low. All of the displayed values are positive in a normal patient. If these values are reduced or are negative, then these values display abnormality. The human eye cannot distinguish such delicate changes in the traditional ECG waveform shape. Quantified standard ST-T data values are easier to read and judge than morphological waveforms.
The time duration length of the ST-T interval is represented by 16 numeric values, which is the average data value of detection. Averaged normal healthy persons values are above 0.10 with variance of plus or minus 0.05. The data values of unhealthy individuals is lower than normal persons; abnormal individuals are in the negative value range. Artificial intelligence is applied in automatic navigation, identification, extraction of ST segment ventricular muscle ultra-low frequency signals. Invasive methods cannot be used in testing of the inner Myocardium.
Various embodiments include an evolved Frank-Starling Law module. Unlike the Frank-Starling Law, this is not a cardiac output, but a trajectory morphological map of myocardial APD during the cardiac conduction cycle (every beat), and three-dimensional information is shown in two dimensions. Normality resembles the form of the heart; abnormality loses this form. The most common abnormal results are displayed as an inverted type, where the red and blue segments are upside down. Abnormality can also display as separation, the parts may be much smaller, too straight, or crisscross. Therefore, this format is easy to read.
A new way to replace data in heart science is to use a “graphical” alternative to the ECG waveform. Combines the principles of electrophysiology and hemodynamics creates a graphic “law”. The left and right ventricle are expressed in color, and the law is expressed in form, which is easy for everyone to read, it is easy to judge, and can be understood immediately within a second. It resembles the upside down “tear drop”; red for the left Ventricle, and blue for the right Ventricle. (The left ventricle is the working part of the ventricle, the wall thickness of the myocardium, the systolic and diastolic heart.) The wall of the right ventricle is thin and the cavity is small).
The displayed left and right ventricle form is consistent, the size is variable, the shape remains more or less the same, the direction remains the same, the left and right colors are blue and red respectively in normal individuals, and the proportion of different size may have to do with individual's health, age, body surface area, disease, etc. Regarding myocardial construction, the left and right ventricle form directly reflects the left and right ventricular pressure gradient characteristics. In pathological cases, the surface area and volume density of myocardial fibers are often associated with changes. As a result, normal and abnormal people will be completely different, and abnormal changes are shown in all shapes and color correspondence with sides.
Statistics, number theory, Fourier transform theory and calculus have played a critical role in shaping the foundation of AI systems. In ECG AI systems two crucial tasks are performed to achieve AI results. First, the focus is placed on fixed-point representations and signal processing algorithms focused pm fixed-point representations. The most popular digitization of information, itself mimics the quantum states of tiny particles such as photos, protons as well as atoms inside quantum physics. Fixed-point representations and signal processing algorithms possess inherent stability or in more accurate terms: fixed-point representations are basically immune to noise and interference.
Second, statistics mathematics provides the basis of modern digital signal processing algorithms which were initially developed for computerized radar systems. Statistics based signal processing algorithms constitute the primary tools in ECG AI systems. Number theory in a sense becomes the true body of AI's logical representations. The integers (or numbers' fixed-point representations), firstly including only positive integers or natural numbers, and are the very basis of information meaningful to human beings. The innovations of various embodiments are a combination of both statistics mathematics and number theory. This innovative combination offers a number of advantages: better capability of ECG AI software in detecting cardiac diseases and more secure as well as more power efficient ECG AI instruments.
While it is well known that Fourier transform theory is the standard tool for peeking into the insights of any signals that are not random, in various embodiments, it has been found that ECG signals or waves are close to perfect periodic signals in most normal situations or almost-periodic signals in all other situations. Fourier series theory, rather than Fourier transform theory, therefore, provides some important insights which could not be found if only Fourier transform theory is considered.
To be more specific, Fourier transform theory, also called Fourier integral theorem is based on the conception of calculus plus using complex domain mathematics to work around the assumption that Fourier series theory only deals with infinite duration sinusoids. Besides those sinusoids' infinite duration in time domain, Fourier series require the summation of multiple harmonic frequencies goes to infinite high as well. As the interesting things unfolded with Fourier series, people have also noticed that Fourier series cannot accurately represent the discontinuity points of any signals no matter how many sinusoids are summed up.
This interesting observation is named as Gibbs phenomenon, which says when using a summation of sinusoids as basic functions to approximate periodic signals, there are always oscillating small waves to be seen in the immediate vicinity of discontinuity points. Now back to ECG signals. If the summation of sinusoids is limited to be a finite number of harmonic frequencies, which is very true in any practical situations anyway, those oscillating small waves in the immediate vicinity of discontinuity points—for example, R-wave in ECG signals behaves much like a discontinuity point—can actually represent the true physiological events in our heart systems. Therefore, in various embodiments, it was discovered that Fourier series theory can be very beneficial to building ECG AI systems if only band-limited signals are involved. As for infinite duration time domain requirement, it is already solved by nanometer microchip technologies that provide almost infinite large memories.
Fourier series theory unexpectedly provides a powerful mathematical tool to reveal the details in the vicinity of those extremely important discontinuity points—in most normal situations inside ECG signals QRS waves are likely discontinuity points which in a peculiar sense can grantee the stability of ECG's periodic nature. Other significant points of ECG signals are those so-called turning points. From the strict mathematical conception, there is a fundamental difference between a discontinuity point and a turning point: the former point at which derivatives from different directions approach different values and this derivative discontinuity turns out to be a “nasty” nature enough to break many theories, and the latter at which point all derivatives from different directions do approach the same value. Based on accumulated experience we have observer many significant turning points inside ECG signals.
The J-point at which the QRS Complex ends might be one of the most important turning points, which indicates the heart enters a quiet period with regard to heart system's electrical activities and this period is referred to as isoelectric line. On the other hand, if abnormalities exist in one's heart system one or more discontinuity points may turn into turning points. This kind of abnormalities often occur to Q wave or S wave when normally sharp-angular waves become wide-angular shaped waves with multiple turning points in sight.
Because derivatives of any curves are actually the slope estimates in the vicinity of those turning points therefore the slope analysis is of vital importance in understanding underlying electrical activities for many physiological events. ECG signals are the recording voltages on our body surface and the slope estimates are directly related to current or the underlying electrical activities. In summary, the first derivative of ECG signals provides the current and the second derivative shows the current changing rate.
Both of them can give clear signs when the underlying physiological events start or end and combining with surface voltage levels significant insights can be gained into those physiological events.
In appearance, the slope analysis of ECG signals seems to be a straightforward simple task. In reality, however, noise and interference are always existing with ECG signals, therefore, all possible signal processing means need to be used to perform the slope analysis. In fact, many different filtering techniques are needed when applying the slope analysis to all those turning points.
A novel electrocardiograph (ECG) signal processing solution based on rectangular-shaped digital lowpass filters is provided. This rectangular-shaped filter based method stems from the modeling methodology for periodic ECG P-Q-R-S-T waves described above by using the Fourier series theory. With piecewise linear lines and partial sinusoidal curves being mathematical models and their computations under the condition of finite harmonics, rectangular-shaped filters become the natural solution as a practical equivalence to computing bandlimited Fourier series. The advantages of this new solution include resulting in clear small waves and better preserving the sharp angles as well as positions of the P-Q-R-S-T waves.
The bandlimited Fourier series spectrum is simply a series of N harmonics uniformly spreading from 0 to N Hz in the frequency domain. Therefore, a rectangular-shaped digital filter allows those N harmonics to pass but attenuates any frequency components higher than N Hz.
As a result, a rectangular-shaped digital filter whose passband ends at NHz and stopband starts before (N+1) Hz is provided. Then the output of such a rectangular-shaped digital filter represents an indirect computing methodology for the bandlimited Fourier series. To design rectangular-shaped digital filters there are just a couple of technical difficulties to deal with. The first is that the filter order needs to be rather high—according to Kaiser 1974 simplified formula the linear-phase finite-impulse-response (FIR) filter order should be estimated as:
where δp and δs represent the passband and stopband filtering error, respectively, and Δf is the transition band width defined as the distance from passband second edge to stopband first edge. For example, if δp=δs=10−2 and Δf=1 Hz, the filter order M has to be at least 1850. However, in real applications, the basis frequency may be as low as 0.75 Hz which corresponds to heartbeat rate of 45 per minute. Therefore, lowpass rectangular-shaped digital filters with orders up to 3000 need to be used.
The design requirement of such high orders of FIR filters imposes the second technical difficulty that no software can actually design those filters in a straightforward manner. The computing noise due to finite-word effects may raise the distortion to such a high level that renders those filters useless. Fortunately, Neuvo et al. in 1984 proposed a frequency-transformation technique called interpolated FIR filter design procedure to solve the second technical problem. In short, the iFIR filter design starts with a couple of low-order filters in the range of a few hundreds, then these two prototype filters are transformed into orders in thousands. Designing FIR filters with orders of a few hundred is quite straightforward and the frequency-transformation to achieve iFIR filters merely adds zero-valued coefficients to them, thereby no additional computing noise being created when implementing the final filters.
System for Displaying ST-T Interval Quantitative Indicators
In various embodiments, artificial intelligence (AI) in conjunction with a database of normal and abnormal ECG data is used to provide quantitative indicators for the ST-T interval of an ECG waveform during measurement of the ECG waveform. This system is referred to as an aiECG system or a system for performing aiECG, for example.
Returning to
A voltage signal is detected between two electrodes 1910 by detector 1920. Detector 1920 also amplifies the voltage signal. Detector 1920 converts the electrical impulses to an ECG waveform for each heartbeat of the beating heart. Detector 1920 converts the voltage signal to a digital voltage signal using an analog to digital converter (A/D), for example. Detector 1920 provides the detected and amplified voltage signal from each pair of electrodes 1910 directly to display device 1940 to display the ECG waveform. The ECG waveform includes conventional P, Q, R, S, T, U, and J waveforms, for example. Detector 1920 also provides the detected and amplified voltage signal from each pair of electrodes 1910 directly to processor 1930.
Processor 1930 can be a separate electronic device that can include, but is not limited to, an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), or a general-purpose processor or computer, such as the system of
Processor 1930 receives the ECG waveform for at least one heartbeat from detector 1920. Processor 1930 converts the ECG waveform to a frequency domain waveform. Processor 1930 separates the frequency domain waveform into two or more different frequency domain waveforms. Processor 1930 converts the two or more different frequency domain waveforms into a plurality of time domain cardiac electrophysiological subwaveforms and discontinuity points between these subwaveforms of the ECG waveform.
Processor 1930 compares the plurality of subwaveforms and discontinuity points to a database (not shown) of subwaveforms and discontinuity points for cardiac electrophysiological signals of ECG waveforms for a plurality of known and normal and abnormal patients. Processor 1930 identifies the ST-T interval from the plurality of subwaveforms and discontinuity points based on the comparison, divides the ST-T interval into N number of equally spaced sections, and calculates an average data value of detection for each section.
Display device 1940 is an electronic display device, a printer, or any combination of the two. Display device 1940 displays the ECG waveform for the at least one heartbeat of the beating heart. Display device 1940 also displays for the at least one heartbeat a table that includes an average data value of detection for each section of the ST-T interval.
In various embodiments, the average data value of detection for each section is a potential energy. The potential energy is calculated from the following equation, for example.
In this equation, the average data value of detection for a section is calculated from the ECG signal f(t) of the section.
In various embodiments, processor 1930 separates the frequency domain waveform into two or more different frequency domain waveforms using one or more rectangular-shaped filters. The one or more rectangular-shaped filters include a finite-impulse-response (FIR) filter or an interpolated FIR filter.
In various embodiments, N equally spaced sections includes 16 equally spaced sections.
In various embodiments, an average data value of detection above 0.10 with a variance of plus or minus 0.05 for a section is a normal value.
In various embodiments, an average data value of detection equal to or below 0.10 with a variance of plus or minus 0.05 for a section is an abnormal value.
In various embodiments, wherein a normal average data value of detection is displayed in black and an abnormal average data value of detection is displayed in red.
Table 2220 shows an average data value of detection for each section of the ST-T interval of one heartbeat of the 52 year-old normal male before PCI. At least, the first four columns of table 2220 show abnormal values (below 0.10) for four sections of the ST-T interval. In other words, the quantitative indicators of table 2220 identify an abnormality. The negative values are depicted in red, but all abnormal values can also be depicted in red.
Table 2230 shows an average data value of detection for each section of the ST-T interval of one heartbeat of the 52 year-old normal male after PCI. Table 2230 shows that after PCI the first four columns return to normal values (above 0.10) for four sections of the ST-T interval. In other words, a comparison of tables 2220 and 2230 shows that average data values of detection for sections of the ST-T interval can be used to detect heart abnormalities even when the ECG waveform shows no abnormalities.
In various embodiments, processor 1930 further identifies a first set of one or more subwaveforms or one or more discontinuity points representing an endocardium of the heart and a second set of one or more subwaveforms or one or more discontinuity points representing an epicardium of the heart. Processor 1930 calculates an average positive potential for the first set and an average negative potential for the second set. Finally, display device 1940 further displays the average positive potential for the endocardium and the average negative potential for the epicardium.
In various embodiments, display device 1940 further displays a suspicious potential range and an abnormal potential range for the average positive potential for the endocardium and a suspicious potential range and an abnormal potential range for the average negative potential for the epicardium.
In various embodiments, display device 1940 displays the average positive potential for the endocardium and the average negative potential for the epicardium for one or more electrodes of the two or more electrodes. The one or more electrodes include leads I, II, V3, V4, V5, and V6, for example.
In various embodiments, processor 1930 further identifies a first set of one or more subwaveforms or one or more discontinuity points representing a right ventricle of the heart and a second set of one or more subwaveforms or one or more discontinuity points representing a left ventricle epicardium of the heart. Processor 1930 calculates a first trajectory of myocardial action potential duration (APD) for the first set and a second trajectory of myocardial APD for the second set. Finally, display device 1940 further displays the first trajectory for the right ventricle and the second trajectory for the left ventricle.
In various embodiments, display device 1940 displays the first trajectory and the second trajectory in the form of a heart for normal values.
In various embodiments, display device 1940 displays the first trajectory in blue and the second trajectory in red.
Method for Displaying ST-T Interval Quantitative Indicators
In step 2710 of method 2700, electrical impulses are received from a beating heart using two or more electrodes.
In step 2720, the electrical impulses are detected from at least one pair of electrodes of the two or more electrodes and converted to an ECG waveform for each heartbeat of the beating heart using a detector.
In step 2730, the ECG waveform for at least one heartbeat is received from the detector, the ECG waveform is converted to a frequency domain waveform, the frequency domain waveform is separated into two or more different frequency domain waveforms, and the two or more different frequency domain waveforms are converted into a plurality of time domain cardiac electrophysiological subwaveforms and discontinuity points between these subwaveforms of the ECG waveform using a processor.
In step 2740, the plurality of subwaveforms and discontinuity points are compared to a database of subwaveforms and discontinuity points for cardiac electrophysiological signals of ECG waveforms for a plurality of known and normal and abnormal patients using the processor.
In step 2750, an ST-T interval is identified from the plurality of subwaveforms and discontinuity points based on the comparison, the ST-T interval is divided into N number of equally spaced sections, and an average data value of detection is calculated for each section using the processor.
In step 2760, the ECG waveform is displayed for the at least one heartbeat of the beating heart and also for the at least one heartbeat a table is displayed that includes an average data value of detection for each section of the ST-T interval using the display device.
Further, in describing representative embodiments of the present invention, the specification may have presented the method and/or process of the present invention as a particular sequence of steps. However, to the extent that the method or process does not rely on the particular order of steps set forth herein, the method or process should not be limited to the particular sequence of steps described. As one of ordinary skill in the art would appreciate, other sequences of steps may be possible. Therefore, the particular order of the steps set forth in the specification should not be construed as limitations on the claims. In addition, the claims directed to the method and/or process of the present invention should not be limited to the performance of their steps in the order written, and one skilled in the art can readily appreciate that the sequences may be varied and still remain within the spirit and scope of the present invention.
This application is a continuation in part of U.S. patent application Ser. No. 16/114,025, filed Aug. 27, 2018, which is a continuation in part of U.S. patent application Ser. No. 15/998,487, filed Aug. 16, 2018, which is a continuation in part of U.S. patent application Ser. No. 15/961,952, filed Apr. 25, 2018, now U.S. Pat. No. 10,092,201, which is a continuation in part of U.S. patent application Ser. No. 15/904,543, filed Feb. 26, 2018, now U.S. Pat. No. 10,085,663, which is a continuation in part of U.S. patent application Ser. No. 15/393,135, filed Dec. 28, 2016, now U.S. Pat. No. 9,999,364, which is a continuation in part of U.S. patent application Ser. No. 14/749,697, filed Jun. 25, 2015, now U.S. Pat. No. 9,538,930 (hereinafter the “'930 Patent”), which is a continuation in part of U.S. patent application Ser. No. 14/662,996, filed Mar. 19, 2015, now U.S. Pat. No. 9,339,204 (hereinafter the “'204 Patent”), which is a continuation of PCT Application No. PCT/US15/20828, filed Mar. 16, 2015, which claims the benefit of U.S. Provisional Patent Application Ser. No. 62/008,435, filed Jun. 5, 2014; this application also claims the benefit of U.S. Provisional Patent Application Ser. No. 62/571,180, filed Oct. 11, 2017 and U.S. Provisional Patent Application Ser. No. 62/701,841, filed Jul. 23, 2018; U.S. patent application Ser. No. 16/114,025 claims the benefit of U.S. Provisional Patent Application Ser. No. 62/551,759, filed Aug. 29, 2017; U.S. patent application Ser. No. 15/998,487 claims the benefit of U.S. Provisional Patent Application Ser. No. 62/546,461, filed Aug. 16, 2018; U.S. patent application Ser. No. 15/961,952 claims the benefit of U.S. Provisional Patent Application Ser. No. 62/489,540, filed Apr. 25, 2017; U.S. patent application Ser. No. 15/904,543 claims the benefit of U.S. Provisional Patent Application Ser. No. 62/463,662, filed Feb. 26, 2017; U.S. patent application Ser. No. 15/393,135 claims the benefit of U.S. Provisional Patent Application Ser. No. 62/271,704, filed Dec. 28, 2015, and U.S. Provisional Patent Application Ser. No. 62/271,699, filed Dec. 28, 2015; and U.S. patent application Ser. No. 14/749,697 claims the benefit of U.S. Provisional Patent Application Ser. No. 62/017,185, filed Jun. 25, 2014, the content of all of which is incorporated by reference herein in their entireties.
| Number | Name | Date | Kind |
|---|---|---|---|
| 20160183827 | Xue | Jun 2016 | A1 |
| Number | Date | Country | |
|---|---|---|---|
| 62701841 | Jul 2018 | US | |
| 62571880 | Oct 2017 | US | |
| 62551759 | Aug 2017 | US | |
| 62546461 | Aug 2017 | US | |
| 62489540 | Apr 2017 | US | |
| 62463662 | Feb 2017 | US | |
| 62271704 | Dec 2015 | US | |
| 62271699 | Dec 2015 | US | |
| 62017185 | Jun 2014 | US | |
| 62008435 | Jun 2014 | US |
| Number | Date | Country | |
|---|---|---|---|
| Parent | PCT/US2015/020828 | Mar 2015 | US |
| Child | 14662996 | US |
| Number | Date | Country | |
|---|---|---|---|
| Parent | 16114025 | Aug 2018 | US |
| Child | 16157178 | US | |
| Parent | 15998487 | Aug 2018 | US |
| Child | 16114025 | US | |
| Parent | 15961952 | Apr 2018 | US |
| Child | 15998487 | US | |
| Parent | 15904543 | Feb 2018 | US |
| Child | 15961952 | US | |
| Parent | 15393135 | Dec 2016 | US |
| Child | 15904543 | US | |
| Parent | 14749697 | Jun 2015 | US |
| Child | 15393135 | US | |
| Parent | 14662996 | Mar 2015 | US |
| Child | 14749697 | US |
