Complexity Reduction in Image-Based Breast Cancer Care

Abstract

The diversity of malignancies of the breast requires personalized diagnostic and therapeutic decision making in a complex situation. This thesis contributes in three clinical areas: (1) For clinical diagnostic image evaluation, computer-aided detection and diagnosis of mass and non-mass lesions in breast MRI is developed. 4D texture features characterize mass lesions. For non-mass lesions, a combined detection/characterisation method utilizes the bilateral symmetry of the breast s contrast agent uptake. (2) To improve clinical workflows, a breast MRI reading paradigm is proposed, exemplified by a breast MRI reading workstation prototype. Instead of mouse and keyboard, it is operated using multi-touch gestures. The concept is extended to mammography screening, introducing efficient navigation aids. (3) Contributions to finite element modeling of breast tissue deformations tackle two clinical problems: surgery planning and the prediction of the breast deformation in a MRI biopsy device.