AI is moving us a step closer to more advanced and equitable endometrial cancer diagnostics

Endometrial cancer is the most common gynecologic cancer, with approximately 8,600 Canadians diagnosed every year. Despite significant advances in cancer diagnostic methods, existing tests sometimes lead to inconsistencies, missing high-risk patients or referring low-risk groups to unnecessary treatments.

In 2013, a landmark study identified four subtypes of endometrial cancer. To take these findings into clinical practice, in 2015 Dr. Jessica McAlpine and team at the University of British Columbia designed a diagnostic tool called ProMisE. In 2020, the World Health Organization made this molecular classification method a global standard of care. In British Columbia, endometrial cancer patients have had free access to ProMisE testing since 2022.

Dr. McAlpine wanted to continue addressing challenges in how current cancer tests identify at-risk patients, so she joined forces with Dr. Ali Bashashati, an expert in AI and bioinformatics at UBC. Working directly with pathologists in the lab and using cutting edge technology, the researchers trained an algorithm to analyze thousands of microscopic images of cancer tissue and search for patterns.

Their AI model adds a third layer of precision to existing cancer diagnostics like the one Dr. McAlpine developed in 2015: “First, there is the level of a pathologist looking at the tissue under a microscope, then the level of a more informative molecular classification, and now AI adds a new level of precision to detect prognostic differences that human eyes wouldn’t be able to notice,” explains Dr. McAlpine. This is also how the researchers discovered a new high-risk subset of endometrial cancer that would have been challenging to detect by existing tools.

Easy and affordable to implement, the AI tool could also reduce geographical barriers to cancer diagnostics. Pathologists anywhere in the world would simply require a desktop slide scanner to digitize the tissue slides, and if they don’t have one, the slides could be sent to a centre with such technology.

The digitized slides would be added into a secure cloud-based system for the AI analysis. Once analyzed, the results could be easily shared back with the pathologists. This simplified process would therefore benefit patients who currently need to travel for hours to receive endometrial cancer results and the appropriate care.

When used in combination with pathology and molecular diagnostics, Dr. McAlpine and Dr. Bashashati’s AI model could help clinicians better identify the risk and treatment pathway for patients. “Patients with higher risk of recurrence could be referred for additional tests, therapy, or follow-ups, while those considered lower risk could receive the appropriate care closer to home,” says Dr. Bashashati.

At a glance

Issue

Current endometrial cancer tests sometimes miss high-risk patients or refer low-risk groups to unnecessary treatments.

Research

A new AI tool developed by Drs. McAlpine and Bashashati at the University of British Columbia is a major step towards innovating cancer diagnostics and making tests more precise and accessible for endometrial cancer patients. Combined with existing cancer diagnostics, their AI tool could help clinicians identify a more appropriate risk group and treatment pathway for patients. This study builds on Dr. McAlpine’s longstanding work to advance endometrial cancer diagnostics and precision tools.