Exploring the Diagnostic Potential of Heart Sound Spectrograms in Depression Identification

This study introduces a unique method for detecting depression by analyzing the acoustic properties of heartbeats, positing that changes in heart sounds may reflect depressive states due to their link with the circulatory system. By collecting heart sounds from 80 subjects and using varying window sizes and overlap percentages, the research investigates how these parameters influence the quality of spectrograms and the precision of depression detection. Spectrograms were analyzed using the Fast Fourier Transform to detail the frequency dynamics of heart sounds. The results indicated a significant correlation between specific spectrogram characteristics and the presence of depressive symptoms, achieving optimal detection accuracy with a window size of 1600 and FFT length of 3200 via a Random Forest classifier. This method proved to be 92.50 % accurate, suggesting that heart sound analysis could serve as a non-invasive, precise diagnostic tool. This promising approach may pave the way for new methodologies in mental health screening, enhancing early diagnosis and treatment of depression.