A comparative study of YOLO models and a transformer-based YOLOv5 model for mass detection in mammograms


COŞKUN D., KARABOĞA D., BAŞTÜRK A., AKAY B., NALBANTOĞLU Ö. U., DOĞAN S., ...Daha Fazla

Turkish Journal of Electrical Engineering and Computer Sciences, cilt.31, sa.7, ss.1294-1313, 2023 (SCI-Expanded) identifier identifier

  • Yayın Türü: Makale / Tam Makale
  • Cilt numarası: 31 Sayı: 7
  • Basım Tarihi: 2023
  • Doi Numarası: 10.55730/1300-0632.4048
  • Dergi Adı: Turkish Journal of Electrical Engineering and Computer Sciences
  • Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Academic Search Premier, Applied Science & Technology Source, Compendex, INSPEC, TR DİZİN (ULAKBİM)
  • Sayfa Sayıları: ss.1294-1313
  • Anahtar Kelimeler: Breast cancer, computer-aided detection, data augmentation, deep learning, transformer-based YOLO, YOLO
  • Sivas Cumhuriyet Üniversitesi Adresli: Evet

Özet

Breast cancer is a prevalent form of cancer across the globe, and if it is not diagnosed at an early stage it can be life-threatening. In order to aid in its diagnosis, detection, and classification, computer-aided detection (CAD) systems are employed. You Only Look Once (YOLO)-based CAD algorithms have become very popular owing to their highly accurate results for object detection tasks in recent years. Therefore, the most popular YOLO models are implemented to compare the performance in mass detection with various experiments on the INbreast dataset. In addition, a YOLO model with an integrated Swin Transformer in its backbone is proposed for mass detection in mammography images within the study. The performance of YOLOv5 models and a transformer-based YOLO model is compared to that of each other and YOLOv3 and YOLOv4 models using images with different sizes on the INbreast dataset. The best results are obtained by the transformer-based YOLO model of YOLOv5 for 832 × 832 image size. In another experiment, we compared the default anchors against the anchors provided by the YOLOv5 autoanchor function before training and saw that the anchors generated by the YOLOv5 autoanchor increased the success rates. Furthermore, various experiments were conducted to observe how data augmentation affects performance. Although a small amount of data was used in the study, high performance was obtained by YOLO algorithms, which are promising tools for cancer detection.