Classification of acute respiratory diseases based on deep networkand MFCC features

Early diagnosis of respiratory diseases is critical for disease management. In this study, a deep learning-based classification system was developed for the automatic analysis of respiratory sounds. Spectral features such as Mel Frequency Cepstrum Coefficients (MFCC) were extracted from the audio data and processed using a deep neural network model consisting of hybrid CNN + LSTM layers. Techniques to reduce class imbalance were used to increase the generalizability of the model. Experimental studies have shown that the developed model achieved 86% accuracy on test data and demonstrated effective and reliable performance in disease detection from respiratory sounds. The findings demonstrate that audio-based analysis can contribute to the early diagnosis of acute respiratory diseases.