Multi-Class Heart Sounds Classification Using Machine Learning and Transfer Learning Approaches Kalp Seslerinin Makine grenmesi ve Transfer grenme Yaklasimlari ile oklu Siniflandirilmasi

BOZ C., KOÇYİĞİT Y.

33rd IEEE Conference on Signal Processing and Communications Applications, SIU 2025, İstanbul, Türkiye, 25 - 28 Haziran 2025, (Tam Metin Bildiri) identifier

  • Yayın Türü: Bildiri / Tam Metin Bildiri
  • Doi Numarası: 10.1109/siu66497.2025.11112497
  • Basıldığı Şehir: İstanbul
  • Basıldığı Ülke: Türkiye
  • Anahtar Kelimeler: Deep learning, Heart sound classification, Machine learning, Phonocardiogram, Spectrogram, Transfer learning
  • Manisa Celal Bayar Üniversitesi Adresli: Evet

Özet

Early detection and accurate diagnosis of cardiovascular diseases (CVDs) are of great importance for public health. This study examines machine learning and transfer learning-based deep learning approaches for classification of phonocardiogram (PCG) signals. Heart sound recordings obtained from the GitHub and Pascal datasets were preprocessed using various techniques to generate spectral representations, which were then classified. k Nearest Neighbors (kNN), Support Vector Machines (SVM), Random Forest (RF), and transfer learning-based VGG16 (finetuning) and ResNet50 (fine-tuning) models were compared. The RF model achieved the highest accuracy at 99.82%, while among transfer learning approaches VGG16 (fine-tuning) performed best with 96.51% accuracy. This work aims to identify the most suitable method for PCG signal analysis by comparing different classification techniques.