Multi-task heart murmur detection based on progressive layered extraction
LI Huhao
LI Shilong
SHOU Du
LI Li
ZHAO Qijun
PAN Fan
Abstract：To address the problem that most of the existing murmur detection methods are limited to a single task and fail to fully u-tilize the correlation between heart murmur signals,we proposed a multi-task heart murmur model based on progressive layered extrac-tion(PLE),which could simultaneously complete the detection of small-segment heart murmurs and the automatic segmentation of heart sound signals.The model could effectively integrate the spatiotemporal information of heart sound signals by sharing the underlying features,thereby improving the accuracy of murmur detection and heart sound segmentation.On the CirCor Digiscope dataset 2022,the average recall rate and average F1 score of the model for small segment murmur detection reached 0.8033 and 0.6096,respectively,and the average recall rate and average F1 score for heart sound segmntation reached 0.9160 and 0.9100,respectively.The results showed that this model could provide new method and idea for the detection and analysis of heart murmur,offer potential technical sup-port for the early screening and diagnosis of cardiovascular diseases.
Keywords：Heart murmursDeep learningSmall-segment murmur detectionMulti-task learningProgressive layered extrac-tion
Publication Date：2025-10-30
Online Publishing Date：2026-09-11(First online date of this platform, not the publication date of the document)
Pages：9( 288-296 )