A 3D Visualization Method for Breast Cancer Surgeons and Patients

Background: Surgical oncologists currently rely on 2D image slices, such as those from mammograms, to assess the locations and extents of their patients’ breast cancer tumors. In order to accurately evaluate the potential success of various surgical options, a surgeon must mentally translate these 2D images into more realistic, 3D image to visualize breast and tumor morphologies. For a patient without the same experience, it is even more difficult to imagine the true, 3D shape, size and location of their tumor and to participate fully in decision-making and surgical planning. Since the type of surgery chosen (nipple-sparing versus skin-sparing, mastectomy versus lumpectomy), and its ultimate clinical and cosmetic consequences, depends on not only a surgeon’s accurate pre-operative assessment, but also on the patient’s understanding; a method of clear, precise, 3D visualization in the clinic would represent a significant advancement.

Methods: Using DCE-MRI as input, we developed TumorSight (TSi) to provide surgical oncologists with an accurate, 3D representation of a patient’s breast tissue and tumor, and to allow computation of useful metrics, such as the tumor-to-breast volume ratio and distance to surgically useful anatomical landmarks such as the nipple. TSi was used to create 3D models of breast cancer patient tumor using MRIs from the SimBioSys Virtual TumorBank which currently houses thousands of patients.

Results: After creating the models, labels for tumor, as well as skin, chest wall, nipple, adipose tissue, glandular tissue, and vasculature were created. After tumor segmentation validation studies (of 49 patients) were completed, the median absolute volumetric error was 0.32 and the median maximum Hausdorff distance was 17.72mm. From these representations, we determine whether the tumor is monocentric, multifocal or diffuse, compute the tumor volume-to-breast volume ratio, and measure the distance of closest approach between the tumor and the nipple. In addition, by adding a margin to the tumor and computing the volume within a convex boundary containing the tumor+margin, TSi provides a pre-operative estimate of the extirpative volume and its fraction of the total breast volume.

Conclusion: The ability to visualize, in 3D, a tumor’s location and distribution in the breast, and to accurately measure important metrics such as tumor-to-nipple distance, multifocality, and tumor-to-breast volume, are critical to surgical planning in breast cancer. To assist in this planning, TSi provides a surgeon and their patient with a three-dimensional breast and tumor representation and reports measurements of key features of interest.