Improving YOLO-based breast mass detection with transfer learning pretraining on the OPTIMAM Mammography Image Database.
Authors
Affiliations (5)
Affiliations (5)
- Department of Engineering and System Science, National Tsing-Hua University, No. 101, Section 2, Kuang-Fu Road, Hsinchu, 30013, Taiwan. Electronic address: [email protected].
- Institute of Nuclear Engineering and Science, National Tsing Hua University, No. 101, Section 2, Kuang-Fu Road, Hsinchu, 30013, Taiwan. Electronic address: [email protected].
- Institute of Nuclear Engineering and Science, National Tsing Hua University, No. 101, Section 2, Kuang-Fu Road, Hsinchu, 30013, Taiwan. Electronic address: [email protected].
- Institute of Nuclear Engineering and Science, National Tsing Hua University, No. 101, Section 2, Kuang-Fu Road, Hsinchu, 30013, Taiwan. Electronic address: [email protected].
- Department of Medical Imaging, Taichung Veterans General Hospital, No. 1650, Taiwan Boulevard Sect. 4, Taichung, 407219, Taiwan; Department of Medical Imaging and Radiological Sciences, Central Taiwan University of Science and Technology, No. 666, Buzih Road, Beitun District, Taichung, 40601, Taiwan; Department of Post-Baccalaureate Medicine, College of Medicine, National Chung Hsing University, No. 145, Xinda Road, South Dist., Taichung, 402202, Taiwan. Electronic address: [email protected].
Abstract
Early detection of breast cancer through mammography significantly improves survival rates. However, high false positive and false negative rates remain a challenge. Deep learning-based computer-aided diagnosis systems can assist in lesion detection, but their performance is often limited by the availability of labeled clinical data. This study systematically evaluated the effectiveness of transfer learning, image preprocessing techniques, and the latest You Only Look Once (YOLO) model (v9) for optimizing breast mass detection models on small proprietary datasets. We examined 133 mammography images containing masses and assessed various preprocessing strategies, including cropping and contrast enhancement. We further investigated the impact of transfer learning using the OPTIMAM Mammography Image Database (OMI-DB) compared with training on proprietary data. The performance of YOLOv9 was evaluated against YOLOv7 to determine improvements in detection accuracy. Pretraining on the OMI-DB dataset with cropped images significantly improved model performance, with YOLOv7 achieving a 13.9 % higher mean average precision (mAP) and 13.2 % higher F1-score compared to training only on proprietary data. Among the tested models and configurations, the best results were obtained using YOLOv9 pretrained OMI-DB and fine-tuned with cropped proprietary images, yielding an mAP of 73.3 % ± 16.7 % and an F1-score of 76.0 % ± 13.4 %, under this condition, YOLOv9 outperformed YOLOv7 by 8.1 % in mAP and 9.2 % in F1-score. This study provides a systematic evaluation of transfer learning and preprocessing techniques for breast mass detection in small datasets. Our results demonstrating that YOLOv9 with OMI-DB pretraining significantly enhances the performance of breast mass detection models while reducing training time, providing a valuable guideline for optimizing deep learning models in data-limited clinical applications.