A Dual-Stream Signal Processing Framework for Velcro Crackle Detection in Idiopathic Pulmonary Fibrosis Using Electronic Stethoscope Instrumentation
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
Idiopathic pulmonary fibrosis (IPF) is a chronic progressive interstitial lung disease characterized by velcro crackles on auscultation. Early detection of such velcro crackles is crucial for timely diagnosis and intervention. Digital signal processing analysis of respiratory sounds using electronic stethoscope instrumentation offers a non-invasive, convenient, and cost-effective measurement approach for early detection of IPF, demonstrating broad clinical application prospects. However, existing studies have yet to establish a standardized measurement dataset of velcro crackles and lack efficient signal processing algorithms tailored to their unique acoustic measurement characteristics. This study proposes a specialized cross-domain multi-perspective signal processing framework purposefully designed for detecting velcro crackles in IPF patients. The algorithm integrates time-domain and frequency-domain features through a dual-stream measurement architecture, incorporating Mel frequency cepstral coefficients (MFCC), Gammatone frequency cepstral coefficients (GFCC), and Bark frequency cepstral coefficients (BFCC) for comprehensive spectral analysis. We introduce a squeeze-and-excitation mamba module (SEMBM) specifically designed for adaptive signal enhancement of velcro crackle-specific patterns and employ a gated attention fusion module (GAFM) to effectively integrate multi-perspective features. We constructed a specialized measurement dataset comprising 547 respiratory sound recordings from 166 subjects using our self-developed electronic stethoscope instrumentation. Based on this, we combined it with a public dataset KAUH to construct a larger-scale composite dataset. Experimental validation demonstrates that our signal processing algorithm exhibits significant advantages in velcro crackle detection tasks, with all performance metrics substantially outperforming existing baseline approaches and state-of-the-art methods. In the binary classification task, the algorithm surpassed the state-of-the-art method by 11.68%. In the more challenging ternary classification task, the algorithm maintained robust performance on the combined dataset, exceeding the existing state-of-the-art result by 11.78%. Ablation studies further validated the critical role of the dual-stream architecture and feature fusion strategy. These findings indicate that the proposed algorithm holds promise as an effective solution for early diagnosis of IPF.
Fields of Study & Subjects
Phonocardiography and Auscultation Techniques
Music and Audio Processing
Artificial Intelligence in Healthcare
Publication details
- Journal / source
- IEEE Transactions on Instrumentation and Measurement
- Volume
- 75
- Pages
- 1–17
- Publication date
- 1 January 2026
- Language
- EN