GUI Toolkit for Pulmonologists: Age-Gender Specific Lung Sound Analysis and Disease Identification Using Sequence Modeling

The objective diagnosis of respiratory diseases, by observing pathological sounds is a promising area of research in respiratory health. A GUI-based toolkit to aid a pulmonologist in decision-making is presented in this paper. Age and Gender-specificmachine learning models are trained and analyzed in this work for the classification of wheeze and crackle. Sequence models like LSTM, GRU, and Transformer were implemented, demonstrating high accuracy in classifying intricate lung sound features for diagnosing various respiratory diseases. In this work, MATLAB is used to enable user-friendly visualization, featuring filtered spectrograms and time- domain lung sound signals. In experimental evaluations, age and gender-specific models (child, adult, female) outperformed general models in the classification of wheeze and crackle. GRU and LSTM models achieved 69% accuracy, while the Transformer model outperformed with 750%.