XAI-Driven Spectral Analysis of Cough Sounds for Respiratory Disease Characterization
A new explainable AI technique reveals distinct cough patterns for COPD patients, which standard models couldn't detect.
Researchers from Universidad de Valladolid published a paper titled 'XAI-Driven Spectral Analysis of Cough Sounds for Respiratory Disease Characterization.' They used occlusion maps with a Convolutional Neural Network (CNN) to analyze cough spectrograms. The XAI method successfully identified significant spectral differences in patients with COPD, a result not found when analyzing raw spectrograms. This demonstrates how explainable AI can uncover disease-specific acoustic signatures to improve diagnostic capabilities.
Why It Matters
It paves the way for more accurate, interpretable, and non-invasive AI tools for early respiratory disease detection.