Patient-Aware Feature Alignment for Robust Lung Sound Classification:Cohesion-Separation and Global Alignment Losses

2025-05-28

AI Summary

Weak cross-patient generalization in lung sound classification—caused by inter-subject physiological and pathological variability—hampers clinical deployment. To address this, we propose a patient-aware feature alignment framework that jointly optimizes intra-patient feature consistency and inter-patient discriminability via two novel losses: patient cohesion-separation loss and global patient alignment loss, thereby mitigating feature space fragmentation. The framework integrates deep feature learning with intra-class compactness, inter-class separation, and patient-level centroid alignment, enabling end-to-end supervised training. Evaluated on the ICBHI dataset, our method achieves 64.84% accuracy for four-class classification (crackles, wheezes, both, normal) and 72.08% for binary classification (abnormal vs. normal), outperforming state-of-the-art approaches. This improvement enhances robustness and clinical applicability for early diagnosis of respiratory diseases.

Technology Category

• Machine Learning: Multi-class/Multi-label Learning & Extreme Classification
• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning

Application Category

• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Lung sound classification is vital for early diagnosis of respiratory diseases. However, biomedical signals often exhibit inter-patient variability even among patients with the same symptoms, requiring a learning approach that considers individual differences. We propose a Patient-Aware Feature Alignment (PAFA) framework with two novel losses, Patient Cohesion-Separation Loss (PCSL) and Global Patient Alignment Loss (GPAL). PCSL clusters features of the same patient while separating those from other patients to capture patient variability, whereas GPAL draws each patient's centroid toward a global center, preventing feature space fragmentation. Our method achieves outstanding results on the ICBHI dataset with a score of 64.84% for four-class and 72.08% for two-class classification. These findings highlight PAFA's ability to capture individualized patterns and demonstrate performance gains in distinct patient clusters, offering broader applications for patient-centered healthcare.

Problem

• Address inter-patient variability in lung sound classification
• Develop patient-aware feature alignment for robust diagnosis
• Improve classification accuracy using cohesion-separation and global alignment

Innovation

• Patient-Aware Feature Alignment (PAFA) framework
• Patient Cohesion-Separation Loss (PCSL)
• Global Patient Alignment Loss (GPAL)