Explainable CNN With Fuzzy Tree Regularization for Respiratory Sound Analysis

Auscultation is an important tool for diagnosing respiratory-related diseases. Unfortunately, the quality of auscultation limited by professional level doctor and environment auscultation. Some studies have focused on automated techniques. However, existing approaches suffer from two challenges: 1) models cannot learn data distributed among multiple hospitals 2) predictions are difficult to interpret physicians. To address this issue, article proposes a novel explainable respiratory sound analysis framework with fuzzy decision tree regularization. This develops ensemble knowledge distillation technique achieves good performance in terms model efficiency accuracy. Fuzzy trees used explain produce rules that can be well accepted The effectiveness thoroughly validated Respiratory Sound database compared other approaches.