BEST IN PHYSICS (JOINT IMAGING-THERAPY) - Registration of Magnetic Resonance, Reconstructed 3D Ultrasound Imaging and Whole-Mount Breast Pathology for Therapy Assessment of Breast Cancer
R Vlad1*, N Samavati2, J Moseley1, H Tadayyon3, S Iradji4, G Stanisz3 4, G Czarnota3 4, K Brock1 3 (1) Radiation Medicine Program, Princess Margaret Hospital, Toronto, ON,Canada (2) Institute of Biomaterials and Biomedical Engineering, University of Toronto, Toronto, ON, Canada, (3) Medical Biophysics,University of Toronto, Toronto, ON, Canada, (4) Imaging Research, Sunnybrook Health Sciences Centre, Toronto, ON, Canada,WE-C-BRA-3 Wednesday 10:30:00 AM - 12:30:00 PM Room: Ballroom A
To develop a procedure to register 3D whole-mount histology of the breast with multimodality imaging, including MRI and ultrasound. This work provides a platform for developing methods for therapy assessment in breast cancer using cell death characterized through quantitative ultrasound.
Ultrasound scans, DCE-MRI, MRI-exvivo mastectomy sample and whole-mount histological slices were obtained. The ultrasound scans were collected using a freehand 2D-ultrasound probe and a tracking tool affixed to the probe for coordinate determination. These coordinates were used to reconstruct the 3D- ultrasound volume. All image data sets were rigidly registered, the tumor was contoured on each data set, and each set of contours was converted to a volumetric mesh. There are considerable deformations between imaging modalities due to different breast positions at the time of scanning and between imaging and pathology. Part of these deformations were estimated using the MRI exvivo scans of the breast collected before histological processing, and applying a biomechanical-based model deformation algorithm to calculate the deformation map from DCE-MRI-invivo to MRI-exvivo.
The rigid registration method resulted in large differences between the residual tumor volumes estimated from DCE-MRI(147%), US(56%) and histology(100%), considering the histology representation as the ground truth. After applying the deformation map to the DCE-MRI, the volume of the residual tumor in DCE-MRI decreased by 30% approaching the representation of the residual tumor in histology and ultrasound. Modest improvements to the Dice Index were seen from DCE-MRI-to histology and from DCE-MRI-to-US after applying the deformation map.
The project provides metrics of comparison between the volumes of residual tumor assessed from ground truth histology and the volumes assessed from ultrasound and DCE-MRI in breast cancer patients. Since tumors are stiffer than surrounding breast tissue, future work will need to consider including tumor elastic properties in calculating the deformation map from MRI-invivo to MRI-exvivo.