Startups participating in the NVIDIA Inception program are using artificial intelligence to address multiple gaps in breast cancer care, from difficulty accessing screening and a shortage of radiologists to delays in test results used to select treatment. These applications are being developed in the United States, where approximately 40 million mammograms are performed annually, while a projected shortage of radiologists over the next decade is expected to increase pressure on the interpretation and diagnosis system.
NVIDIA says the companies it highlights use its computing infrastructure, including graphics processing units, CUDA, and MONAI, to build tools ranging from automated imaging to analysis of pathology slides and the integration of clinical and genetic data.
More Consistent Imaging, Closer to the Patient
iSono Health is developing ATUSA, an FDA-cleared wearable, three-dimensional, quantitative ultrasound system. The system captures a standardized volume of the breast in approximately two minutes per breast, compared with up to 45 minutes for a traditional manual ultrasound examination.
The system uses artificial intelligence to automate image capture and was trained on thousands of complete examinations containing more than 1.5 million ultrasound frames. The company says the system is 28% more sensitive than two-dimensional manual ultrasound. Standardizing the imaging method may also help compare annual examinations and reduce variation associated with the operator’s experience.
ATUSA is commercially available through partner clinics in California, Texas, Georgia, Tennessee, and Washington, D.C. iSono Health is conducting a multicenter study involving 3,200 patients to validate the platform’s performance, with primary research sites at UC Davis and Vanderbilt University Medical Center. The company is working to expand its tools to include lesion detection, segmentation, and classification, and then integrate three-dimensional ultrasound with mammography, magnetic resonance imaging, and clinical information.
Reducing the Burden on Radiologists
Whiterabbit.ai is developing WRDensity, an FDA-cleared tool for automatically assessing breast density from mammograms, which has been used in the care of hundreds of thousands of patients. It also offers WRRisk to support clinical decision-making and estimate the long-term risk of developing breast cancer.
The company is researching a new generation of artificial intelligence tools for mammography that could help radiologists detect more cancers while automating the review of images that show no indications of disease. The practical idea is not to replace the physician, but to reduce the burden of searching for rare cases; co-founder and Chief Technology Officer Jason Su says a breast radiologist is trying to find approximately one cancer case among every 200 mammograms.
From Image to Treatment Decision
Ataraxis AI focuses on the period after diagnosis, when treatment decisions may require tissue tests that take two to four weeks. The company’s models analyze digital pathology slides and standardized clinical variables to predict the tumor’s response to treatment and the risk of recurrence.
One model predicts the likelihood that a tumor will shrink in response to chemotherapy before surgery, while another model estimates after surgery the risk of disease recurrence within five years and the potential benefit of chemotherapy. The company says both models have been validated at more than ten institutions and in several clinical trials, and are used clinically, with deployment on local infrastructure, external data centers, or the cloud.
Integrating Data into a Three-Dimensional Model
SimBioSys is developing precision medicine technology that creates three-dimensional models of breast tumors, blood vessels, and soft tissue to support surgery and treatment planning. It has also developed a tool for estimating the risk of cancer recurrence based on volumetric magnetic resonance imaging data, pathology results, and clinical data.
The most important change lies in combining multiple types of data rather than analyzing each source separately. However, the material itself points to a fundamental limitation: some of the technologies mentioned are still investigational and have not received FDA approval for commercial use. The results of ongoing studies and clinical validation will also remain decisive before these tools are widely adopted, particularly when the predictions directly influence treatment decisions.
Why Does This News Matter?
These applications show that the practical value of artificial intelligence in breast cancer is not limited to detecting an abnormal image. It also extends to making imaging more reproducible, stratifying risk, and accelerating the analysis of data that precedes a treatment decision. Nevertheless, clinical adoption is not determined solely by the speed of artificial intelligence models; regulatory clearances, validation across multiple institutions, and prediction accuracy among different patient groups are all questions that remain open.