Electrocardiogram (ECG) analysis method and ECG analysis apparatus

Abstract

The invention relates to the technical field of image processing, and specifically relates to an electrocardiogram (ECG) analysis method and ECG analysis apparatus. In order to solve the problems of lack of high-quality medical labeling and lack of interpretable auxiliary analysis in the prior art, the invention discloses the ECG analysis method comprising the following steps: obtaining text semantic information vector corresponding to each ECG on basis of a plurality of ECGs acquired in advance through a text semantic information extraction model; obtaining ECG waveform vector corresponding to each ECG on basis of the plurality of ECGs through an ECG waveform vector extraction model; obtaining a first ECG analysis result according to the text semantic information vectors, the ECG waveformvectors and a first event set acquired in advance; obtaining a second ECG analysis result according to the text semantic information vectors and the ECG waveform vectors; and then, obtaining ECG event corresponding to each ECG according to the first ECG analysis result and the second ECG analysis result. The ECG analysis method disclosed by the invention is capable of directly interpreting ECGs in PDF format.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to an electrocardiogram analysis method and device. Background technique [0002] Since its birth, electrocardiography (ECG) examination technology has been widely used in clinical health examinations in cardiovascular and other fields because of its simple, efficient and non-invasive characteristics. Traditional ECG machines can generate ECG reports that include basic information such as ECG signal waveforms and rhythms; human physicians can give specific insights into ECG reports based on comprehensive analysis of patient age, signs, medical history, etc., as well as medical knowledge and case experience. judge. Electrocardiogram examination has a large business volume in hospitals at all levels. However, medical resources in my country are relatively scarce and severely unevenly distributed. There is a shortage of experienced electrocardiogram physicians in primary hosp...

AI Technical Summary

Problems solved by technology

One is the scarcity of high-quality medical annotations: for rare ECG events, it is very difficult to collect enough training samples, and end-to-end models such as complex neural networks tend to overfit the ECG in the training set, even if obtained on the experimental data. Although it has achieved good results, it is still difficult to guarantee the generalization ability of patients in real complex environments

The second is the lack of interpretability of auxiliary analysis: modern evidence-based medicine attaches great importance to giving evidence for diagnosis, while the existing public end-to-end artificial intelligence diagnosis model can only directly give the probability of specific ECG events, human physicians It is difficult to efficiently verify and confirm the judgments given by such machines, which makes these systems limited in clinical scenarios with high safety requirements

Images