A software platform for visualization & multimodal registration of diffuse optical tomography and MRI of breast cancer

We have developed a software platform for multimodal integration and visualization of diffuse optical tomography and magnetic resonance imaging. Novel registration and segmentation algorithms have been integrated into the platform. The multimodal registration technique enables the alignment of non-concurrently acquired MR and DOT breast data. The non-rigid registration algorithm uses two-dimensional signatures (2D digitally reconstructed radiographs) of the reference and moving volumes in order to register them. Multiple two-dimensional signatures can robustly represent the volume depending on the way signatures are generated. An easy way to conceptualize the idea is to understand the motion of an object by tracking three perpendicular shadows of the object. The breast MR image segmentation technique enables a priori structural information derived from MRI to be incorporated into the reconstruction of DOT data. The segmentation algorithm is based on "Random walkers". Both registration and segmentation algorithms were tested and have shown promising results. The average Target Registration Error (TRE) for phantom models simulating the large breast compression differences was always below 5%. Tests on patient datasets also showed satisfying visual results. Several tests were also conducted for segmentation assessment and results have shown high quality MR breast image segmentation.