Breast Lesion Segmentation Using a 3D Active Contours Without Edges Approach with a GPGPU Implementation
Breast cancer is the second most frequently diagnosed type of cancer in women, just after skin cancer, and is the second leading cause of cancer deaths. Thanks to improvements in early detection and treatments, the death rates of breast cancer have been reduced con- siderably. This work describes a 3D breast lesion seg- mentation algorithm that can be used for preoperative staging, evaluation after chemotherapy, and screening of women at high risk of developing breast cancer. The segmentation algorithm is a 3D implementation of the Active Contours Without Edges (ACWE) method, in- troduced by Chan and Vese . The proposed algo- rithm uses the General-Purpose Computing on Graph- ics Processing Units (GPGPU’s) to accelerate the orig- inal model by parallelizing two main steps of the seg- mentation process, the computation of the Signed Dis- tance Function (SDF) and the evolution of the seg- mented curve. The segmentation process is visualized in real time with a graphical user interface (GUI) de- veloped with Qt and OpenGL.