FDA clears AI that predicts breast cancer recurrence from routine MRIs
SimBioSys won FDA clearance for an AI tool that predicts breast cancer recurrence up to 10 years out using existing MRI and clinical data.
Tools · Source: MPO Mag
What happened
The FDA just granted 510(k) clearance to TumorSight Risk by Chicago-based SimBioSys. It is a cloud-based AI software device that predicts breast cancer recurrence. The tool analyzes standard pre-treatment dynamic contrast-enhanced MRIs alongside basic clinical and pathology data. It requires no new hardware, no extra imaging, and no tissue shipping.
The software generates a proprietary 0 to 100 risk score and classifies patients as low, intermediate, or high risk. It provides prognostic estimates for recurrence-free survival at five and ten years. The tool is cleared for adult women with hormone receptor-positive, HER2-negative Stage I to IIIA breast cancer. This specific subtype accounts for about 75 percent of all breast cancers in the United States.
The clearance relies on a retrospective study of 2,129 patients across four major institutions, including Mayo Clinic and Stanford Medicine. The AI successfully stratified patients by risk. High-risk patients showed a 3.3 times greater risk of recurrence compared to low-risk patients. Each 10-point increase in the score correlated with a 22 percent increase in relative recurrence risk. The FDA classified it as a Class II prescription device.
Key facts
- 2,129 — patients in the multi-institutional validation study
- 3.3x — greater recurrence risk for high-risk patients versus low-risk
- 22% — increase in relative recurrence risk per 10-point score jump
- 75% — of US breast cancers fit the HR-positive, HER2-negative subtype
- 0-100 — proprietary risk score generated by the AI
Why it matters
This turns a standard diagnostic image into a predictive time machine without adding friction to the clinical workflow. Traditionally, recurrence risk assessment relies on expensive tissue-based genomic assays that demand logistical overhead and long turnaround times. TumorSight Risk extracts prognostic intelligence from data already sitting on the hospital server. Oncologists get objective recurrence data before surgery, allowing them to make faster, more confident decisions about chemotherapy and endocrine therapy.
This sets a new baseline for medical AI products. The FDA is showing a willingness to clear software that extracts secondary value from existing diagnostic pathways. Founders building in health-tech should take notes. If your AI requires hospitals to buy new hardware or alter their clinical routines, you face a massive adoption hurdle. Products that piggyback on existing data streams to deliver high-value insights will dominate the next wave of medical AI.
For builders
Zero-friction clinical integration wins
SimBioSys built a product that requires no extra procedures or tissue shipments. Hospitals pay for software that seamlessly integrates into existing workflows. Startups that force doctors to change their daily habits will lose to those that run silently in the background.
Spatial AI unlocks legacy imaging data
The AI extracts new prognostic value from standard MRI scans. Diagnostic imaging centers and hospitals sit on mountains of underutilized data. Founders who can build predictive models on top of this existing data will attract serious healthcare enterprise contracts.
AI as a clinical co-pilot
The FDA cleared this as an indicator to be used alongside other clinical factors, not as a standalone decision maker. Insurers and hospital networks will pay for AI that augments physician judgment and reduces unnecessary treatments. AI that claims to replace doctors entirely will face insurmountable regulatory walls.
My take
I love seeing AI products that do not ask users to change their behavior. SimBioSys nailed it by taking data hospitals already collect and turning it into a 10-year crystal ball for cancer recurrence. If you are building AI for healthcare, stop trying to invent new hardware and start mining the data already sitting on the servers.
Original reporting: MPO Mag. This is my rewrite and opinion.