| Summary: | 碩士 === 國立中正大學 === 資訊工程所 === 93 === Tumor angiogenesis is the process that correlates to tumor growth, invasion, and metastasis. Breast cancer angiogenesis has been the most widely studied and now serves as a paradigm for understanding the biology of angiogenesis and its effects on tumor outcome and the patient’s prognosis. Most studies on characterization of tumor angiogenesis focus on pixel/voxel counts. However, in breast cancer, vascular morphology and tortuosity can provide more information that helps the physician diagnose more accurately. This paper presents a computer-aided diagnostic (CAD) system that can quantify vascular morphology and tortuosity using 3-D power Doppler ultrasound (US) on breast tumors. The method to extract morphological and tortuous information from angiography and to relate them to tumor diagnosis results is proposed. At first, a 3-D thinning algorithm helps narrow down the vessels into their skeletons. These measurements of vascular morphology significantly rely on the traversing of the vascular trees produced from skeletons and the estimation of vascular tortuosity is applied to trunks in vessels. Our study of 3-D assessment of vascular morphological and tortuous features regards number of vessels (NV), total length of vessels (LT), number of branches in all vessels (NB), average radius in trunks (Ravg), number of cycles (NC), vascularity index (VI), distance metric (DM), inflection count metric (ICM), and sum of angle metric (SOAM). Finally, the extracted features for all the 3-D training datasets are then fed into the neural network to compute the training process. The trained neural network is then further used for the breast tumor diagnosis. Investigations into 221 solid breast tumors include 110 benign and 111 malignant cases. The true positive fraction and false positive fraction of the proposed system are 90.0% and 16.6%, respectively. Our scheme focuses on the vascular architecture without involving the technique of tumor segmentation. The results show that the proposed method is feasible and have a good agreement with the diagnosis of the pathologists.
|
|---|
