Medical Technology · us
Scanning Breast Cancer Excision Margins in the Operating Room: Claire AI Imaging System Enters Commercial Use for the First Time
U.S.-based Intermountain Health has introduced the FDA-cleared OCT-assisted system at two hospitals, allowing surgeons to examine suspicious tissue before breast-conserving surgery is complete; a clinical trial showed it can reduce some positive margins but cannot replace pathological diagnosis.
After a breast tumor is removed, surgeons face the challenge not only of excising the lesion, but also of determining whether cancer cells remain hidden at the edges of the removed tissue. If too little is removed, the patient may need another operation; if too much is removed, the appearance of the breast may be affected. Now, a system combining optical imaging and artificial intelligence has moved from clinical trials into routine surgical workflows at U.S. hospitals for the first time.
Perimeter Medical Imaging AI announced that Intermountain Health has deployed the Claire OCT System at LDS Hospital and American Fork Hospital in Utah, becoming the first healthcare system in the United States to commercially adopt the device. The partnership between the two organizations covers the entire healthcare system, but the company has not disclosed the purchase amount, planned number of installations, or expansion timeline.
Claire uses near-infrared light for optical coherence tomography (OCT), creating cross-sectional tissue images up to approximately 2 millimeters deep from breast tumor specimens that have been removed from the body. A convolutional neural network then classifies image regions as “suspicious” or “non-suspicious,” highlighting areas that require further physician review. Intermountain breast surgeon Teresa Reading said the device can be integrated into existing workflows and directs her attention to specific areas of concern; however, this remains one user’s preliminary experience rather than systematic effectiveness data from a commercial setting.
The FDA cleared Claire in March 2026. Its pivotal study enrolled a total of 613 participants across 11 centers; the primary effectiveness analysis in the final version included 206 patients who used the device. After the original intraoperative margin assessment had been completed, the number of patients who still had positive margins fell from 35 to 28, an absolute reduction of 3.4 percentage points and a relative reduction of 20%, meeting the prespecified primary endpoint. No adverse events attributed to the device were identified in the safety data.
The findings also have clear limitations. The study’s primary analysis compared outcomes before and after Claire was used within the device group, rather than simply comparing the device group with the control group; false alerts generated by the system also led to the removal of an average of 0.5 additional tissue samples per patient that were ultimately found to be cancer-free. Clinical practice must still determine how this additional excision and the small number of missed classifications affect reoperation rates, operating time, costs, and long-term cosmetic outcomes.
Claire is positioned as an assistive tool, not an autonomous diagnostic system. It must be used alongside physician interpretation and existing margin-assessment methods, and it cannot replace postoperative histopathological examination of permanent sections; its current indicated population also excludes groups such as patients whose primary diagnosis is lobular carcinoma. The significance of the first commercial deployments, therefore, lies not in declaring that AI has solved the margin problem, but in beginning to build evidence on how a regulated imaging system changes decision-making in real-world operating rooms.