Pulmonary Hypertension Detection via Deep Learning of Heart Sounds

Pulmonary hypertension is a serious vascular disease that involves severe respiratory and cardiac complications in patients suffering from it. A more accessible and, above all, early diagnosis could improve the prognosis, reducing complications and enabling timely and effective treatment. The main aim of the following study is to find a non-invasive, accessible and reliable method to diagnose pulmonary hypertension. To do so, a dataset of heart sounds obtained from the pulmonary valve was analysed and processed to create a deep learning model capable of identifying acoustic patterns characteristic of the disease and supporting the diagnostic process automatically and accurately. The developed model was trained and tested on this dataset, yielding promising results in terms of accuracy and ability to distinguish between healthy subjects and those with pulmonary hypertension. Preliminary analyses suggest that the proposed approach could be a valuable support tool for early diagnosis, helping to improve disease identification in clinical settings. Further studies and validations on larger datasets will be necessary to confirm the reliability and generalisability of the method.