How SimBioSys® Uses AI to Bring 3D Precision to Breast Cancer Surgery

Breast cancer is the most common cancer among women globally, impacting over 2.3 million patients annually. Traditional presurgical planning has relied on black-and-white MRI images and brief radiology summaries. This approach limits a surgeon's ability to fully understand tumor shape, position, and context within the breast.

As a result, critical surgical decisions—including whether to perform breast-conserving surgery or a mastectomy—have sometimes been based on incomplete or hard-to-interpret data, leading to variable outcomes and unnecessary loss of healthy tissue. Re-excision rates in breast cancer surgery remain above 20 percent, and data presented at the 2025 American Society of Breast Surgeons Annual Meeting highlighted significant variability in surgical decisions, even among experienced providers.

Clinical validation forms the foundation of TumorSight Viz's adoption in real surgical settings. SimBioSys is actively collecting prospective clinical and operational outcomes, with early indicators proving encouraging. TumorSight Viz has demonstrated strong concordance with radiologist annotations, accurate delineation of tumor size relative to breast volume, and consistent performance across more than 1,600 retrospective cases spanning over nine institutions—supporting its reliability in real-world surgical planning. 

Against this backdrop, clinical sites report that TumorSight Viz improves confidence and efficiency in preoperative planning by providing standardized, anatomy-specific insight. Formal outcome studies are underway to further quantify its impact on surgical decision-making and downstream outcomes.

One compelling example that highlights TumorSight Viz's real-world impact is the case of Brooke Davis, whose experience was featured nationally on ABC News and “Good Morning America” in collaboration with NVIDIA. TumorSight Viz provided a clearer, 3D understanding of her tumor anatomy, helping her care team evaluate breast-conserving options that weren't fully apparent through traditional imaging alone. 

Across clinical sites, similar cases demonstrate how improved anatomical clarity can support more informed surgical decisions, preserve healthy tissue, and positively influence recovery and long-term quality of life. For patients, this means more personalized surgical planning and greater confidence that decisions are being made with the fullest possible picture of their disease.

High-resolution 3D segmentation of full breast MRI volumes within clinically relevant timeframes wouldn't be feasible without GPU acceleration. SimBioSys uses MONAI across its medical imaging pipeline, including data preprocessing, AI model development, validation, and packaging and delivery of model artifacts. MONAI's domain-specific tooling significantly accelerated development of 3D segmentation workflows and helped ensure consistency and rigor as the company scaled.

NVIDIA CUDA™ acceleration underpins nearly everything SimBioSys does—from training segmentation models to performing high-resolution 3D inference. GPU-accelerated libraries enable the company to process large volumetric datasets efficiently while maintaining the precision required for surgical planning. Drop-in CUDA-X replacements for traditional numerical libraries like cuCIM, cuML, and cuBLAS have accelerated processing times with essentially no code changes. These improvements were critical in addressing early feedback from surgeons that processing times were a hindrance to effective clinical use. NVIDIA GPUs allow SimBioSys to generate interactive, color-coded 3D models in minutes rather than hours—making them essential for clinical use rather than limited to offline research.

Beyond segmentation, CUDA-X-accelerated simulations of gravitational effects on breast deformation, blood flow through vessels, and estimates of drug delivery are enabling new solutions for clinical problems facing surgical and medical oncologists.