Classifier generation method, atrial fibrillation detection device and storage medium

Abstract

The invention provides a classifier generation method, an atrial fibrillation detection device and a storage medium. The atrial fibrillation detection device and the storage medium carry out the atrial fibrillation / non-atrial fibrillation classification detection on detected signals through the classifier. The generation method of the classifier comprises the following steps of extracting a cardiac interval of a cardiac signal segment of a pre-labeled type; calculating a variation value of the heart beat interval; grouping the plurality of heart beat interval variation values, and extracting the frequency of each group of data as a feature vector; and carrying out machine learning training on the mark type and the feature vector to generate the classifier. According to the classifier generation method provided by the invention, the dimension of the original feature can be reduced, the discrimination of the sample and the robustness to noise are improved, and the classifier can be quickly obtained through training by adopting a simple machine learning model.

Description

technical field [0001] The present application relates to the technical field of supervised machine learning, and in particular to a method for generating an atrial fibrillation / non-atrial fibrillation classifier, an atrial fibrillation detection device including the classifier, and a computer storage medium. Background technique [0002] Atrial fibrillation (abbreviated as atrial fibrillation, Atrial Fibrillation, AF) is the most common sustained arrhythmia, and it is increasingly showing younger age. The heartbeat frequency in atrial fibrillation is often fast and irregular, sometimes up to 100-160 beats per minute, not only much faster than normal heartbeat, but also absolutely irregular. [0003] The current detection algorithms for atrial fibrillation can be divided into atrial activity-based analysis and RR interval-based analysis. [0004] The principle of detecting atrial fibrillation through atrial activity analysis (ECG analysis) mainly relies on the detection of ...

AI Technical Summary

Problems solved by technology

However, the existing methods of generating AF / NON-AF classifiers based on RR interval machine learning have a large dimension of the extracted feature vectors, and the training models used are mainly nonlinear deep learning models. The hardware configuration involved is relatively high, the learning speed is low, and the generated classifier has a huge structure, which is not suitable for transplantation to portable detection equipment for online application

Images