Evaluation of a Novel Computer-Aided Diagnosis System for Non-Mass Enhancing Breast Lesions Based on Combination of Spatial and Temporal Information of MRI Scans
Dynamic contrast-enhanced (DCE) breast MRI is being applied for detection, diagnosis, and staging of breast cancer. As breast DCE-MRI is recently recommended as a screening option of breast cancer for women at high risk, differentiation of malignant and benign tumors is becoming a more important function in breast DCE-MRI. Analyzing spatiotemporal enhancement patterns remains a challenging task because of the complex character of both kinetics and morphological aspects of these lesions. Methods such as shape-based analysis not relying on the precise boundary of the lesion and detecting representative kinetic time-series become imperative. We propose to apply modern state-of-the art-techniques to the segmentation, shape description and classification of these lesions.
1. Segmentation based either on continuous force analysis or dynamic texture segmentation or textons.
2. Preprocessing: Velocity Zernike moments
3. Classification and evaluation: SVM/Hidden Markov Models/Adaboost.
4. Writing down thesis.