Print  Close

Automatic recognition of heart disease based on phonocardiogram

Ting-Yu Yan Co-Author
Deaprtment of Applied Mathematics, National Sun Yat-sen University
 
Yu-Jung Huang Co-Author
I-Shou University
 
Ming-chun Yang Co-Author
Department of Pediatrics, E-Da Hospital, Kaohsiung, Taiwan,
 
Wei-Chen Lin Co-Author
Department of Medical Research, E-DA Hospital
 
Meihui Guo First Author
National Sun Yat-Sen University
 
Meihui Guo Presenting Author
National Sun Yat-Sen University
 
Wednesday, Aug 6: 8:40 AM - 8:45 AM
1034 
Contributed Speed 
Music City Center 
Heart sound recognition is crucial for early cardiovascular disease detection, but auscultation alone often leads to diagnostic challenges, even for experienced clinicians. To address this, we propose a convolutional recurrent neural network (CRNN) model combined with machine learning, utilizing MFCC, SFTF, and Deep Scattering features. Applied to 512 datasets from E-Da Hospital, our CRNNA + LightGBM model achieved 92.2% accuracy (specificity: 96.2%, sensitivity: 88%), outperforming physicians by 9.7% in accuracy and 24% in sensitivity.

Using self-attention mechanisms, we visualized the model's focus areas, which closely matched physicians' auscultation regions, demonstrating its ability to act as a diagnostic proxy. Validation on the 2016 PhysioNet/CinC Challenge database further confirmed the model's robustness, achieving 95% accuracy (specificity: 93%, sensitivity: 98%).

Keywords

CRNNA


Deep scattering

Heart sound classification

Light GBM

MFCC

PCG 

Main Sponsor

Section on Statistical Learning and Data Science