In the age of digital technology, there is a great interest in artificial intelligence-based disease diagnosis algorithms. There are many image-based studies, especially for lung diseases. Cardiovascular and respiratory diseases are the world's top two causes of death according to the World Health Organization. For this reason, for the diagnosis of these diseases, there is a need for systems with a high accuracy rate, as well as reliable and fast-responding systems. In addition to image-based diagnoses, it is possible to diagnose diseases with audio signals. In the study, three data sets, cough, breath, and /a/ sound, were collected from patients hospitalized in the ward who was diagnosed with Asthma, COPD, and Pneumonia, the characteristics of the sound signals were extracted and then a spectrogram image was created. Artificial neural network (ANN) model training was performed on the sound signals whose features were extracted. As a result of the ANN model training, it has been verified with the graphics that any user gives high accuracy in detecting the disease with cough, breath, and /a/ sound.