Method for non-invasive diagnostic assessment of the cardiovascular system and brain of patients

Abstract

A method for the non-invasive diagnostic assessment of the cardiovascular system and brain of patients relates to the field of medicine. The technical result of the claimed invention, which lies in enabling less invasive diagnostic assessment of disorders of the cardiovascular system and brain in patients while providing the necessary diagnostic accuracy, is achieved in that the present method for the non-invasive diagnostic assessment of the cardiovascular system and brain of patients includes first recording a training set of cardiovascular system diagnostic assessment results obtained in parallel from electrocardiography and as a result of recording wireless channel state parameters, then recording a training set of brain diagnostic assessment results obtained in parallel from an electroencephalogram and as a result of recording wireless channel state parameters, and then training a machine learning model which, as a result, enables a diagnostic assessment of the cardiovascular system and the brain on the basis of wireless channel state parameters recorded.

Description

[0001] METHOD OF NON-INVASIVE DIAGNOSTICS OF THE STATE OF THE CARDIOVASCULAR SYSTEM AND BRAIN OF PATIENTS

[0002] DESCRIPTION

[0003] The invention relates to the field of medicine, namely to methods of non-invasive diagnostics of the state of the cardiovascular system and brain of patients.

[0004] A METHOD FOR TRAINING AN ELECTROCARDIOGRAPHY ANOMALY DETECTION NETWORK AND A METHOD AND DEVICE FOR EARLY WARNING OF ELECTROCARDIOGRAPHY ANOMALIES are known in the prior art [CN112022144 (A), published 04.12.2020], characterized in that it includes the following steps: obtaining electrocardiogram signals of patients with an abnormal electrocardiogram and electrocardiogram signals without anomalies, extracting training data from the obtained electrocardiogram signals, training a binary neural network by accepting the training data and using the trained binary neural network as an electrocardiography anomaly detection network, wherein for each network layer in the electrocardiography anomaly detection network, the values ​​and weights of the nodes of the network layer are binary data, and a binary operation is performed on the values ​​and weights of the nodes of the network layer to obtain the value of the node of the next network layer.

[0005] The disadvantages of the analogue are: - in this technical solution, invasive diagnostic methods are used to obtain training data, which complicate the procedures for continuous diagnostics of the patient’s body condition;

[0006] - in this technical solution, diagnostics of anomalies is performed only based on electrocardiogram data, which does not provide sufficient diagnostic accuracy;

[0007] - there is no additional training of the neural network in the event of detection of new data that were not taken into account in the training sample when training the neural network.

[0008] The closest in technical essence is the METHOD AND SYSTEM FOR NON-INVASIVE SCREENING ASSESSMENT OF PHYSIOLOGICAL PARAMETERS AND PATHOLOGIES [RU2016146176A, published 05 / 24 / 2018], which includes the following steps: forming training and test samples of patient records with a given pathology or physiological parameters that depend on the cardiac activity of patients, including records of patients of different genders and ages, wherein each record contains at least one cardiac lead of the ECG signal and information about the patient; obtaining records from the training sample, wherein for each record at least one cardiac lead of the ECG signal is processed, and the parameters of heart rate variability and the average cardiac cycle are calculated;training an artificial neural network to detect a given pathology or physiological parameters using records of the training and test samples, comparing the parameters of the processed ECG signal, the calculated parameters of heart rate variability and the average cardiac cycle, and information about patients; storing the connections and weights of the trained artificial neural network; obtaining at least one cardiac lead of the ECG signal and information about the patient; processing the obtained at least one cardiac lead of the ECG signal, calculating the parameters of heart rate variability and the average cardiac cycle; determining the physiological parameters or the presence of a given pathology, using the trained neural network, using the parameters of the processed ECG signal, the calculated parameters of heart rate variability and the average cardiac cycle, and information about the patient.

[0009] The main problem with the prototype is that this method uses invasive methods of obtaining training data, which limit the possibilities of continuous diagnostics of the patient's body condition; the neural network is trained on a training sample compiled only from ECG, which does not allow for more accurate diagnostics of pathologies. In addition, this method does not provide the ability to further train the neural network on new data that was not taken into account when forming the training sample for training the neural network. It is also worth noting that the use of a neural network, as one of the machine learning devices, does not always provide a stable and accurate result of pathology diagnostics, which is confirmed by errors of the first and second type of recognition.

[0010] The objective of the invention is to eliminate the shortcomings of the prototype.

[0011] The technical result of the invention is to reduce the invasiveness of diagnostics of pathologies of the cardiovascular system and the brain of patients while ensuring the required accuracy of diagnostics. The said technical result is achieved due to the fact that the method of non-invasive diagnostics of the state of the cardiovascular system and the brain of patients is characterized by the fact that initially a training sample of the results of diagnostics of the state of the cardiovascular system is recorded, formed on the basis of recording electrophysiological signals, simultaneously received from electrocardiography and as a result of recording the parameters of the state of wireless communication channels, then a training sample of the results of diagnostics of the brain is recorded, formed on the basis of recording electrophysiological signals, simultaneously received from the electroencephalogram and as a result of recording the parameters of the state of wireless communication channels,then, the machine learning model is trained, within the framework of which the machine learning model is trained first to diagnose the state of the cardiovascular system by training to compare the results of the patient's diagnosis based on the electrophysiological signals obtained from electrocardiography with the results of the registered parameters of the state of wireless communication channels, as a result of which the machine learning model is trained to diagnose the state of the cardiovascular system based on the results of the registered parameters of the state of wireless communication channels, and then to diagnose the state of the brain by comparing the results of the patient's diagnosis based on the electrophysiological signals obtained from electrocardiography with the results of the registered parameters of the state of wireless communication channels, as a result of which the machine learning model is trained to diagnose the state of the brain based on the registered parameters of the state of wireless communication channels,then, using the trained machine learning model, at the stage of training the machine learning model, the state of the cardiovascular system and the brain is diagnosed, where the state of the cardiovascular system and the brain is diagnosed based on the recorded parameters of the state of the wireless communication channels.,

[0012] In particular, to record the training sample of the results of cardiovascular system diagnostics, the RSSI and CSI parameters of the wireless communication channels are recorded.

[0013] In particular, to record the training sample of the results of the diagnosis of the state of the brain, the RSSI and CSI parameters of the state of the wireless communication channel are recorded.

[0014] In particular, if the trained machine learning model receives registered parameters of the state of wireless communication channels at the input, according to which the trained machine learning model does not produce a result of diagnosing the state of the cardiovascular system and the patient's brain, then these registered parameters of the state of wireless communication channels are fed to retrain the machine learning model.

[0015] Implementation of the invention.

[0016] The method for recording electrophysiological signals of the state of the cardiovascular system and the brain of patients is characterized in that initially a training sample of the results of diagnosing the state of the cardiovascular system 1 is recorded, formed on the basis of parallel recording of electrophysiological signals obtained from electrocardiography performed using an electrocardiograph, and obtained as a result of recording RSSI and CSI parameters of the state of wireless communication channels received from smart devices such as a smartphone, smart watch, smart Wi-Fi router, as well as RSSI parameters of the state of wireless communication channels received from a cellular repeater and a base station communication channel, then a training sample of the results of diagnosing the brain 2 is recorded, formed on the basis of parallel recording of electrophysiological signals obtained from an electroencephalogram performed using an electroencephalograph,and RSSI and CSI of the wireless communication channel state parameters received from Bluetooth devices installed in the right and left wireless earphones inserted into the patient's ears, for example, wireless Apple AirPods headphones. In this case, when forming a training sample for both cases, various patient behavior scenarios, different states of health of the cardiovascular system and brain (without pathologies, with various pathologies) are taken into account, and the gender and age of the patients are also taken into account, including for the formation of a training sample of the results of cardiovascular system diagnostics, data on the state of the cardiovascular system of the fetus in the womb and the pregnant woman herself are taken into account.

[0017] Next, the training of the machine learning model 3 is directly carried out, within the framework of which the machine learning model is trained first to diagnose the state of the cardiovascular system by training to compare the results of patient diagnostics based on electrophysiological signals obtained from electrocardiography with the results of the registered RSSI and CSI parameters of the state of wireless communication channels, as a result of which the machine learning model is trained based on the results of the registered RSSI and CSI parameters of the state of wireless communication channels to diagnose the state of the cardiovascular system, and then to diagnose the state of the brain, for example, nervous diseases, sleep disorders, by comparing the results of electrophysiological signals obtained from electrocardiography with the results of the registered RSSI and CSI parameters of the state of the wireless communication channel,as a result, the machine learning model is trained using the registered RSSI and CSI parameters of the wireless communication channel state to diagnose the state of the brain.

[0018] Then, using the trained machine learning model, at the stage of training the machine learning model 3, the state of the cardiovascular system and the brain 4 is diagnosed, at which the state of the cardiovascular system is diagnosed based on the registered RSSI and CSI parameters of the state of the wireless communication channels received from smart devices, such as a smartphone, smart watch, smart Wi-Fi router, as well as based on the registration of RSSI parameters of the state of the wireless communication channels received from the cellular repeater and the base station communication channel, and then the state of the brain is diagnosed based on the results of registering RSSI and CSI parameters of the state of the Bluetooth wireless communication channel.

[0019] If the trained machine learning model at the stage of training the machine learning model 3 at the stage of diagnosing the state of the cardiovascular system and the brain 4 receives at the input the registered parameters of the state of the wireless communication channels, according to which the trained machine learning model does not issue the result of diagnosing the state of the cardiovascular system and the brain of the patient, then these registered parameters of the state of the wireless communication channels are fed to retrain the machine learning model 5. The technical result of the invention is a reduction in the invasiveness of diagnosing pathologies of the cardiovascular system and the brain of patients while ensuring the required accuracy of diagnosis, which is achieved due to the fact that the recording of training samples of the results of diagnosing the state of the cardiovascular system and the brain, formed on the basis of recording electrophysiological signals, simultaneously received from electrocardiography and electroencephalogram,and the corresponding parameters of wireless communication channels, allows to significantly improve the quality of training the machine learning model and, as a result, to ensure the required accuracy of diagnostics of the state of the cardiovascular system and the brain, while this approach to training the machine learning model and subsequently diagnostics of the state of the cardiovascular system and the brain, produced on the basis of the registered parameters of wireless communication channels, allows to reduce invasiveness in diagnostics while ensuring the required accuracy of diagnostics, and also to expand the possibilities of diagnostics using non-invasive methods, since the smart devices used to record the parameters of wireless communication channels can always be worn by patients. In addition, maintaining the required accuracy while adhering to approaches in non-invasive diagnostics is achieved by retraining the machine learning model in cases wherewhen diagnostics record wireless channel parameters that were not taken into account when training the machine learning model,

[0020] Examples of achieving technical results:

[0021] In patient Ekaterina, 55 years old, at the stage of registering RSSI and CSI parameters of wireless communication channels received from smart devices, namely from a smartphone, smart watch, smart Wi-Fi router, as well as registering RSSI parameters of the state of wireless communication channels received from a cellular repeater and a base station communication channel, for diagnosing the state of the cardiovascular system and RSSI and CSI parameters of the state of the wireless communication channel received from Bluetooth devices of wireless headphones inserted into the patient's ears, for diagnosing the state of the brain using a trained machine learning model, the following pathologies of the cardiovascular system were identified, namely myocarditis, hypertension and pathologies of the brain, namely the development of epileptic disease, neurotic symptoms manifested in panic attacks and sleep disorders.After identifying these cardiovascular and brain pathologies, the patient was diagnosed based on ECG and EEG analysis, resulting in diagnoses confirmed with 97% accuracy.

[0022] Patient Elena, 32, eight months pregnant. At the stage of registering RSSI and CSI parameters of wireless communication channels received from a smart Wi-Fi router, for diagnosing the state of the cardiovascular system using a trained machine learning model, the parameters of wireless communication channels were registered in the fetus inside the pregnant woman, according to which a normal, without deviations, state of the cardiovascular system was diagnosed. At the same time, tachycardia was detected in the mother according to the registered parameters of the state of wireless communication channels. The separation of the parameters of wireless communication channels registered from the intrauterine fetus and the pregnant woman itself is made on the basis of taking into account the features of their electrocardiograms and the simultaneously registered parameters of wireless communication channels, which are laid down during the training of the machine learning model.The cardiovascular conditions of the fetus and pregnant woman identified by the trained machine learning model were then verified based on the analysis of the ECG of the fetus and pregnant woman, resulting in diagnoses confirmed with 95% accuracy.

[0023] This example reflects the implementation of the claimed method. For this purpose, a group of patients aged 25 to 65 years was collected. Using these patients, training samples of the results of diagnostics of the state of the cardiovascular system and the brain were recorded, at first the patients were connected to an electrocardiograph and electrocardiography was recorded for five days, 10 electrocardiographs with a recording period of one hour each, as a result, 50 electrocardiographs were collected, which were analyzed and according to the obtained data (deviations of the main teeth: P, Q, R, For example, patient Ivan, 39 years old, was diagnosed with the following cardiovascular pathologies, namely:

[0024] 1) hypertrophy of the heart sections, which was accompanied by an increase in the time of internal deviation, since in the hypertrophied myocardium excitation spreads longer in the area from the endocardium to the epicardium, an increase in the amplitude of the R wave, ischemia of the subendocardial layers of the heart, caused by a lack of blood flowing through the coronary arteries, conduction disturbances, deviation of the electrical axis of the heart towards the hypertrophied section, since its mass increases due to the growth of cardiomyocytes, a change in the electrical position of the heart; displacement of the transition zone (V3), manifested by a change in the ratio of the R and S waves in the third chest lead;

[0025] 2) angina pectoris, which was accompanied by changes in the terminal part of the ventricular QRS complex: depression of the S-T segment; various changes in the T wave, for example, a decrease in amplitude, biphasicity, isoelectricity or negativity; the focal nature of these changes: they were recorded in one or two leads, since the observed hypoxia is local in nature, developing in the basin of a separate branch of the coronary artery;

[0026] 3) Tachycardia, which was accompanied by an increase in heart rate, with the heart rhythm being 100-150 beats per minute, which was reflected by an increase in the automatism of the sinus node, a decrease in the R - R interval was also noted, since the T - P interval was shortened, the P - Q segment was shortened; the degree of increase in heart rate was directly proportional to the decrease in Q - T, an upward shift of the RS - T segment downward from the isoelectric line was observed.

[0027] In parallel, when taking electrocardiographs, RSSI and CSI parameters of wireless communication channels were recorded, which were taken from smart devices, namely a passive Wi-Fi router, smart watches and recorded on a mobile phone: RSSI recorded parameters were the power indicators of incoming signals, interference and noise coming from base stations of cellular communication and repeaters, measured in dBm, and CSI recorded parameters were the parameters of the signal propagated from the transmitter to the receiver and reflecting scattering, fading and reduction in signal power, namely the parameters of instantaneous, statistical CSI.

[0028] Then the patients were connected to the electroencephalograph and electroencephalogram was recorded for five days, 10 electroencephalograms with the recording period of each one hour, in total 50 electroencephalograms were collected, which were analyzed and based on the obtained data (frequency, amplitude, degree of manifestation and zonal differences, a-rhythm, R-rhythm, 0-oscillations, b-activity, namely their wave form, modulation, asymmetry, tendency to form bursts) the patient was diagnosed with the following pathologies of the cardiovascular system, namely the presence of a focus of partial epileptiform activity in the frontotemporal region of the right hemisphere, which were accompanied by the following parameters: a-rhythm is weakly modulated, fragmentary, unstable in frequency, with an amplitude of up to 50 μV, predominance in the parietal regions of the right hemisphere; a pronounced focus of b- and R-activity in the right frontotemporal region; 9-activity is slightly represented in the frontocentral regions;paroxysmal activity was characterized by synchronous generalized discharges of a complex of sharp and slow waves; hyperventilation caused a slight increase in EEG disorganization.

[0029] In parallel, when recording electroencephalographs, RSSI and CSI parameters of wireless Bluetooth communication channels were recorded, which were taken from Bluetooth devices inserted into the ears of wireless Apple AirPods headphones and recorded on a smartphone.

[0030] Based on the obtained training sample, a machine learning model, namely a convolutional neural network, was trained. For this purpose, the training sample, consisting of data sets including ECG and EEG readings and the corresponding parameters of wireless communication channels, was divided into training - 50%, control - 30% and test 20%. As a result, the neural network was trained without retraining with a quality of 97% of correct diagnosis of the state of the cardiovascular system and brain based on the parameters of wireless communication channels used in the test sample, which was not involved in training, but was used to assess the quality of training of the convolutional neural network.

Claims

FORMULA 1. The method for non-invasive diagnostics of the state of the cardiovascular system and the brain of patients is characterized by the fact that initially a training sample of the results of diagnostics of the state of the cardiovascular system is recorded, formed on the basis of recording electrophysiological signals, simultaneously received from electrocardiography and as a result of recording the parameters of the state of wireless communication channels, then a training sample of the results of diagnostics of the brain is recorded, formed on the basis of recording electrophysiological signals, simultaneously received from the electroencephalogram and as a result of recording the parameters of the state of wireless communication channels, then a machine learning model is trained, within the framework of which the machine learning model is trained first to diagnose the state of the cardiovascular system by training to compare the results of diagnostics of the patient based on electrophysiological signals,obtained from electrocardiography, with the results of the registered parameters of the state of wireless communication channels, as a result of which the machine learning model is trained based on the results of the registered parameters of the state of wireless communication channels to diagnose the state of the cardiovascular system, and then to diagnose the state of the brain by comparing the results of the patient's diagnosis based on electrophysiological signals obtained from electrocardiography with the results of the registered parameters of the state of wireless communication channels, as a result of which the machine learning model is trained based on the registered parameters of the state of wireless communication channels to diagnose the state of the brain, then using the trained, Machine learning models at the training stage of the machine learning model diagnose the state of the cardiovascular system and the brain, which diagnoses the state of the cardiovascular system and the brain based on the recorded parameters of the state of wireless communication channels.

2. The method according to item 1, characterized in that in order to record the training sample of the results of diagnosing the state of the cardiovascular system, the RSSI and CSI parameters of the state of wireless communication channels are registered.

3. The method according to item 1, characterized in that in order to record the training sample of the results of diagnosing the state of the brain, the RSSI and CSI parameters of the state of the wireless communication channel are recorded.

4. The method according to item 1, characterized in that if the trained machine learning model receives registered parameters of the state of wireless communication channels at the input, according to which the trained machine learning model does not produce a result of diagnosing the state of the cardiovascular system and the patient’s brain, then these registered parameters of the state of wireless communication channels are fed to retrain the machine learning model.

Images