Sub-regional Tumor Segmentation Based on CEUS Perfusion Characteristics: Enhancing Breast Tumor Diagnosis

Breast cancer is one of the most prevalent cancers among women globally, and its increasing incidence poses a significant threat to women's health. Despite notable advances in early diagnosis and treatment due to the continuous progress in medical technology, the high heterogeneity within breast cancer still results in considerable variability in clinical manifestations, treatment responses, and disease progression. This diversity presents new challenges in achieving precise treatment. Thus, a profound exploration and study of the heterogeneity of breast cancer are crucial for developing more effective diagnostic models, advancing treatment strategies, and enhancing cure rates. In current clinical practice, although biopsy is widely used for the diagnosis of benign or malignant breast tumors, its accuracy and comprehensiveness are somewhat limited due to the complex internal heterogeneity of breast cancer and the invasive nature of the procedure. In recent years, preoperative qualitative diagnosis of breast cancer using medical imaging technology has become a hot topic in research. Compared with other common imaging techniques such as CT and MRI, ultrasound examination is extensively employed due to its safety, convenience, and lower cost. Particularly, Contrast-Enhanced Ultrasound (CEUS) technology, with its superior temporal resolution, can vividly illustrate the details of tumor perfusion hemodynamics, effectively revealing key features such as enhancement patterns, blood supply, and vascular invasion of the tumor. This study is dedicated to using dynamic CEUS videos to cluster perfusion characteristics at the pixel level within the tumor and divide the tumor into different subregions based on the clustering results. We will explore the correlation between these perfusion subregions and the diagnosis of benign or malignant breast tumors, and based on this, develop related diagnostic models. This non-invasive diagnostic approach aims to maximally mine and utilize image data, comprehensively capturing the tumor's perfusion characteristics at the pixel level, and reducing subjective judgments in the diagnostic process. The application of this method is not only expected to improve the accuracy of breast cancer diagnosis but also to provide more information support for personalized treatment of patients, thereby promoting progress in the field of breast cancer treatment.