An artificial intelligence processing circuit for ECG signals based on heart beat differential coding

Abstract

The invention discloses an electrocardiographic signal artificial intelligence processing circuit based on heart beat differential coding, which belongs to the technical field of electrocardiographic signal processing. The scheme is as follows: the preprocessing module is connected with the differential coding module through the heartbeat memory, the differential coding module is connected with the activation module and the neural network classifier in sequence, and the differential network memory and the non-differential network memory are connected with the neural network classifier through the template update module. The heart beat template memory is connected, and the heart beat template memory is also connected with the differential encoding module. The invention utilizes the waveform characteristics of the differential heartbeat and the adaptive threshold to realize the dynamic wake-up of the normal and abnormal heartbeat, thereby reducing the number of startup times of the artificial intelligence classifier and ultimately reducing the overall system power consumption. The invention utilizes the historical classification result of the artificial intelligence classifier to generate a patient heart beat template, improves the generalization accuracy of model processing, does not need to additionally mark the patient's heart beat in the whole process, and can learn the patient's self-specificity in real time.

Description

technical field [0001] The invention belongs to the technical field of electrocardiographic signal processing, and in particular relates to an electrocardiographic signal artificial intelligence processing circuit based on heart beat differential coding. Background technique [0002] According to the 2019 Global Health Estimates Report published by the World Health Organization (WHO), cardiovascular disease (CVD) is the leading cause of death worldwide. In order to accurately detect CVD patients, electrocardiogram (ECG) plus manual arrhythmia analysis by professional physicians is generally used in medical treatment, but this type of solution is difficult to use in the field of home medical equipment because of cumbersome equipment and high labor costs . In recent years, the field of home medical care has become more and more popular with the advent of the era of intelligence. In terms of ECG heart health monitoring, daily ECG monitoring equipment with intelligent analysis ...

AI Technical Summary

Problems solved by technology

Generally speaking, these five requirements often restrict each other, and it is difficult to meet the above requirements at the same time with the design directly based on the existing hardware. ASIC chips for physiological signal recognition are pursuing lower power consumption levels to meet long battery life requirements

[0004] In addition, applying existing technical achievements to actual daily heart health monitoring equipment also needs to solve the bottleneck of generalization accuracy: in the field of ECG monitoring, due to the differences in physiological structure between patients, each person's cardiac cycle activity There are inherent differences, so that the ECG waveforms that record cardiac cycle activity are patient-specific. In addition, due to differences in the position of cardiac electrodes or age changes, the patient's own ECG records also have patient-specificity that changes over time. sex

However, for arrhythmia identification, the artificial intelligence classifier can only access the ECG records of patients in the database for a period of time during the learning phase, so the current artificial intelligence classifiers often show extremely high recognition accuracy on the database. However, for patients outside the database, the generalization accuracy may drop significantly.

Some existing solutions such as online learning also have some inherent problems: 1) They all need to collect a part of the ECG records of the actual user (patient), and then mark the classification label of each heartbeat on the ECG waveform by a professional physician; 2) It is impossible to learn the latest heartbeat characteristics (ie ECG characteristics) of the patient in real time, and the accuracy rate may decrease as the user wears it again in the daily use of the user

Images