AI imaging technology detects early endometrial cancer with high accuracy

Endometrial cancer is the most common type of gynecologic cancer, with more than 69,000 cases diagnosed in the United States in 2025 and increasing by up to 3% annually. Diagnosis requires a painful and often invasive biopsy that carries the risk of false-negative results. A multidisciplinary research team at Washington University in St. Louis and Siteman Cancer Center, based at Barnes-Jewish Hospital and Washoe Medicine, is investigating a fast, safe, non-invasive imaging method combined with machine learning for the accurate detection and diagnosis of precancerous lesions and early cancer.

The team, led by Qing Zhu, the Edwin Murty Professor of Engineering in the McKelvey School of Engineering at Washington University in St. Louis, conducted a preliminary investigation using optical coherence tomography (OCT), which detects differences in how tissue reflects light and acquires high-resolution 3D images 1 to 2 millimeters deep. Using a custom catheter probe developed in Zhou’s lab, the team captured images of the entire endometrial cavity in less than 3 seconds, creating a visual inspection. Biopsy. It is the first catheter-based 3D imaging study that integrates functional, structural and radiological features to evaluate the endometrium. The research results were published in imagine npjg June 3, 2026.

To obtain images of the patient’s tissue, the team collaborated with WashU Medicine physicians led by Lindsay Kuroki, MD, associate professor of obstetrics and gynecology, and Ian Hagman, MD, PhD, professor of pathology, immunology and obstetrics and gynecology. They, along with Cho, are research members at Siteman Cancer Center, where Kuroki also treats patients. The team obtained OCT images from 57 uteruses after hysterectomies in 2025. Of these, 34 contained high-risk precancerous lesions or early-stage cancers.

3D OCT images provided a close-up view of tissue microstructure and optical properties, revealing clear differences between normal endometrium, benign endometrium, high-risk precancerous lesions, and endometrial cancer at different stages.

First authors, Sanskar Thakur, a doctoral student in Zhou’s lab, and Yiqiao Lin, who received his PhD in biomedical engineering from UW in 2025, developed an imaging feature extraction pipeline and a machine learning model to classify findings into two groups: normal and benign, and pre-cancerous and cancer using 26 extracted imaging features. Their model achieved an exploratory sensitivity of 94% and specificity of 87%.

“Current endometrial biopsy practice has an estimated false-negative rate of approximately 10% (about 90% sensitivity), largely due to sampling limitations and interpretive variability,” Zhou said.

Qing Zhu is the Edwin Murty Professor of Engineering at the McKelvey School of Engineering, Washington University in St. Louis

“There is currently no reliable screening for endometrial cancer,” said co-author David Mutch, the Ira C. and Judith Gall Professor and vice chair of the Department of Obstetrics and Gynecology at Washoe Medicine, a Siteman Research Member and principal investigator of the National Cancer Institute-funded Specialized Program for Endometrial Cancer Research Excellence (SPORE). “This technology, developed by Dr. Zhu and her colleagues, should allow us to better screen this cancer and at least detect it earlier in its development,” Matsch added. “It’s really new and cutting-edge technology.”

Going forward, Zhu said the team plans to evaluate the catheter in living patients to demonstrate the translational potential of the AI-powered OCT technology.