Breast Ultrasound Tumor Detection Based on Improved YOLOv8s-OBB Algorithm

Breast cancer is the most common cancer among women, and early detection and treatment can significantly improve survival rates. Due to its non-radiative nature, effectiveness, and low cost, ultrasound imaging has become a primary tool for breast cancer screening. However, breast ultrasound images typically have low resolution, high noise, and blurred boundaries, making it easy to misdiagnose when relying solely on manual diagnosis. To address these issues, this paper proposes an improved YOLOv8s-OBB algorithm named MS-YOLO for breast ultrasound tumor detection. The model introduces oriented bounding boxes (OBB) to more accurately detect the shape and orientation of tumors and employs deformable convolutional networks (DCNv4) to enhance the model's ability to capture irregular tumor boundaries and complex structures. Additionally, the designed Multi-Scale Feature Focus Pyramid Network (MSFPN) improves the ability to fuse multi-scale information. Experimental results show that MS-YOLO outperforms traditional methods on the BUSI breast ultrasound dataset, particularly in detecting complex backgrounds, demonstrating higher detection accuracy and robustness. This study provides an efficient and accurate solution for breast ultrasound tumor detection.