Intermountain Health Deploys AI Imaging System for Breast Cancer Surgery

Intermountain Health has become the first U.S. health system to commercially deploy an FDA-approved artificial intelligence imaging system designed to help surgeons assess breast cancer margins during breast-conserving surgery.

The Utah-based nonprofit health system began using Perimeter Medical Imaging AI’s Claire OCT System at LDS Hospital and American Fork Hospital, according to Perimeter’s 08/13 announcement. The deployment moves Claire from clinical development into commercial use in U.S. operating rooms.

Claire combines optical coherence tomography (OCT), a light-based imaging technique, with an AI computer-aided detection algorithm. The system produces three-dimensional, real-time images of excised breast tissue and marks areas that are suspicious for cancer for review by the surgeon. The U.S. Food and Drug Administration (FDA) approved Claire on 03/03/2026.

AI Moves Into the Operating Room

The system is designed to address a problem that can persist after breast-conserving surgery, where surgeons remove a tumor while attempting to preserve as much healthy breast tissue as possible.

A 2024 study published in Annals of Surgical Oncology found that 21.1% of women in a commercial insurance cohort and 14.9% of women in a Medicare cohort underwent another operation within one year of breast-conserving surgery. Reoperations were associated with 24% higher one-year healthcare costs in both groups, according to the peer-reviewed study.

Claire is intended to provide additional information while the patient is still in surgery. Its FDA-approved use covers evaluation of margins from excised lumpectomy tissue alongside standard methods and physician interpretation. The FDA specifies that Claire should not replace standard histopathology or be used as a standalone diagnostic tool.

The device images tissue below its surface, producing high-resolution views up to 2 mm deep. Its AI then marks focal areas that are suspicious for breast cancer, giving surgeons regions to examine more closely rather than making the surgical decision itself.

The FDA's review included 206 participants in the primary effectiveness analysis. After standard-of-care intraoperative margin assessment, 35 participants had residual diseased margins. After Claire-assisted assessment, that number fell to 28, a 3.4 percentage-point absolute reduction and 20% relative reduction. The device also detected and helped remove residual positive margins missed by standard-of-care assessment in seven participants.

From Clinical Trial to Commercial Use

The commercial deployment follows a broader expansion of AI into clinical workflows. Healthcare organizations have increasingly moved beyond administrative applications toward systems designed to support diagnosis, clinical reasoning, and patient care, including Mayo Clinic's healthcare-specific AI initiative with Microsoft.

Cancer care is also becoming a testing ground for specialized AI systems. Researchers at Stanford Medicine and Mayo Clinic have developed an AI framework that uses blood-based measurements to study tumor microenvironments, another example of AI being applied to cancer-related clinical research.

For Perimeter, the Intermountain deployment represents a commercial milestone following the FDA's March approval. The company had previously described Claire as a system intended to support surgeons in identifying difficult-to-see cancer during breast-conserving procedures.

Intermountain's deployment covers LDS Hospital and American Fork Hospital under its systemwide agreement with Perimeter. The company said the technology is intended to give surgical teams additional information during the procedure, while existing pathology methods remain part of the evaluation process.

The deployment does not establish that AI can eliminate repeat breast-cancer surgery. It puts an FDA-approved imaging and AI system into routine clinical use for a specific part of the surgical workflow, where identifying residual disease before an operation ends could affect what happens next.