Detection of Valvular Heart and Respiratory Diseases using and AI technique

A low-cost artificial intelligence-based mobile solution with a digital stethoscope is developed for the early detection of heart and lung diseases. This solution will classify the user's heart/lung sounds into normal and abnormal sounds. Our solution currently works for commonly known heart and lung diseases such as murmur and COPD. A digital stethoscope consists of a stethoscope head and a microphone. This is used to record real-time heart/lung sounds with the help of the microphone jack in the mobile phone into the mobile app. The mobile app is then connected to firebase where the heart/lung sound is stored. The audio recorded in the mp4 format is manually converted into a .wav format and then uploaded onto the computer system. The system contains the training model which extracts the features using MFCC and trains them using CNN with the help of datasets collected from trusted sources. The GUI on the computer system outputs whether the user has a murmur or not or COPD or not depending on the heart or lung disease detection. We have obtained high accuracy ranging from 95% to 100% while training the model. This solution can be used as an alternative to visiting clinics to detect murmur or COPD in a person. This is an effective and low-cost way of detecting the aforementioned diseases in the comfort of the home. The use of this solution can be quite helpful in rural areas where there is a lack of medical assistance and facilities. This can be used by frontline workers in rural areas who are not well versed in the medical field.