Early Warnings: CI MED Team Harnesses Machine Learning Insights to Improve Endometrial Cancer Diagnosis

3/2/2024 Beth Hart

Written by Beth Hart

New machine learning tools under development by researchers at Carle Illinois College of Medicine and Carle Health could help address disparities in the diagnosis of endometrial cancer in younger women and women of color. A clinician-student team is leveraging technology to identify patterns in patients’ health care visits that could alert doctors of increased risks of endometrial cancer in populations that are often underdiagnosed.

Endometrial cancer (EC) develops in the lining of the uterus, and the first warning sign is usually abnormal bleeding. While endometrial cancer is most common in postmenopausal women, death rates are rising disproportionately among women of color and women who haven’t yet gone through menopause.

<em>Dr. Megan Hutchcraft</em>
Dr. Megan Hutchcraft

CI MED Clinical Assistant Teaching Professor and Gynecological oncologist Dr. Megan Hutchcraft of the Carle Cancer Institute says a conclusive diagnosis is possible through in-office tissue sampling, but many EC cases undergo late diagnosis due to two factors: patients may be slow or reluctant to report irregular bleeding to their provider, and some clinicians miss the warning signs or rely too heavily on ultrasound imaging technology in their diagnostic workup. Clinicians frequently use transvaginal ultrasound to measure endometrial thickness as a threshold test when endometrial cancer is suspected. But Hutchcraft says the guideline measurements established for postmenopausal women don’t apply to younger women and are less accurate in women of color, contributing to disparities in timely diagnosis.

<em>Dr. Megan Hutchcraft</em>
Ameek Bindra

In an effort to improve how clinicians investigate possible EC in these populations, Hutchcraft teamed up with CI MED student Ameek Bindra, who is leveraging her background in machine learning and data science to help uncover new insights. “Our goal is to develop technology that can help guide health care providers to this possible diagnosis to prompt earlier endometrial sampling,” Hutchcraft said.

The team is breaking new ground by using artificial intelligence and machine learning algorithms to examine demographic and clinical data obtained from premenopausal patients who have subsequently been diagnosed with endometrial cancer. “Our objective is to discern whether health care usage patterns exist preceding endometrial cancer diagnoses in patients and to explore additional factors influencing endometrial cancer suspicion and diagnosis,” Bindra said. “Hopefully, we can find a way to integrate these models in clinical practice. Contributing to a cause that could improve the prognosis of EC, particularly in women of color, makes this very meaningful work to me.”

In the meantime, Hutchcraft is educating future physicians and residents under her supervision and her patients to consider the potentially serious implications of abnormal bleeding across patient populations. Irregular bleeding before menopause, or any bleeding after menopause requires gynecologic evaluation, and persistent bleeding or bleeding in the presence of risk factors requires endometrial biopsy,” Hutchcraft said. Obesity is the most common risk factor for EC, followed by polycystic ovarian syndrome (PCOS) and a family history of uterine cancer.

Hutchcraft is also raising the issue within the medical community to address disparities in EC diagnosis. She co-authored an editorial titled, “Missed Opportunities for Young Women with Endometrial Cancer: A Call to Action,” published in the journal Gynecologic Oncology Reports in December of 2023.

Share this story

This story was published March 2, 2024.