Ultrasound and AI Bias: Challenges in Endometrial Cancer Detection for Black Women

No results found.

A recent study published in JAMA Oncology reveals that a widely used algorithm for detecting endometrial cancer may not be reliable for Black women. The research emphasizes the need for tissue biopsies in Black patients exhibiting concerning symptoms, rather than relying solely on transvaginal ultrasounds, which are considered less effective in this demographic. Endometrial cancer, the most common gynecological cancer in the U.S., is particularly aggressive among Black women, who often face later diagnoses. Analyzing data from 1,500 Black women who underwent hysterectomy at ten health centers, the researchers found that nearly 10% of patients with endometrial cancer had a thinner endometrium than the commonly accepted cutoff of 4 millimeters, indicating a significant risk of missed diagnoses.

Dr. Kemi Doll, lead author and gynecologic oncologist, criticized the algorithm’s purported accuracy of 99% to 100% in ruling out cancer, stating that it does not hold true in real-world scenarios for Black women. The study identified four complicating factors affecting detection in this group, including reduced visibility of tumors, the prevalence of fibroids, and technician skill levels. Given these challenges, Dr. Doll recommends that tissue biopsies should be the first line of testing for Black patients presenting with symptoms, rather than an optional follow-up. This finding aligns with previous studies highlighting racial bias in ultrasound diagnostics, which found Black women are at a higher risk of false-negative results. As nearly 70,000 women are expected to be diagnosed with uterine cancer this year, the implications of this research could reshape diagnostic protocols to ensure more equitable and accurate cancer care for Black women.

Read More

Contact Support

If you're interested in posting an article and need assistance, please don't hesitate to contact our support team. We're here to help you through the process, answer any questions you may have, and ensure that your article is published smoothly and effectively.

support@ehealthcommunity.org