Feb 24, 2025

It will take a village: making AI work for Pathology

Programs: Postgraduate, Research: Artificial Intelligence in healthcare, Research: Cancer, Disruptive Innovation, Alumni

By Jenni Bozec

“My role in digital pathology is just starting but I can already see how important it is to collaborate across countries and industries to make it work. I believe the more people we can get involved, the more collaboration we can get from different institutions, and the larger the data sets we can accumulate, the more we improve the quality of AI in the future”, comments Dr. Ellen Yang.

Having completed a residency in Diagnostic and Molecular Pathology in the Department of Laboratory Medicine and Pathobiology (LMP), she then did a fellowship at Memorial Sloan Kettering (MSK) in New York. Recently returning to Toronto, she accepted a position as a Breast Pathologist at Mount Sinai Hospital and Assistant Professor in LMP.

Yang already had an interest in AI and worked on related research projects during her residency. Her fellowship at MSK allowed her to explore AI in Pathology further which has resulted in several papers, including one published in Nature Medicine. Her research in this area was also recognized by an award from the International Society of Breast Pathology at the latest United States and Canadian Academy of Pathology (USCAP) meeting.

During her residency, she worked with Dr. Ming-Sound Tsao on the reproducibility of assessment of lepidic (non-invasive) patterns in lung cancer and quickly realized how a software tool could improve this challenging diagnosis. Consensus between pathologists in diagnosing this disease can be varied, due to overlapping features of some invasive patterns, chronic lung injury and tissue processing artefacts. They aimed to increase consensus in diagnosis by developing a program that could identify relevant patterns in the nuclear stains and assist Pathologists.

“AI in Pathology is not something of the future, it’s here right now,” Yang says, “It has so much potential as an ancillary tool. It helps make us faster and more efficient by highlighting abnormal tissue, but it cannot replace the expertise of the Pathologist.”

Yang’s experiences at MSK allowed her to test drive the use of AI. She conducted several projects where she compared the output of AI tools such as Paige Breast and Virchow foundation model with a Pathologist’s diagnosis.

“The program at MSK highlights the area of interest and tells you whether there is suspicion of cancer so you, as the Pathologist, can look more closely and make a decision. If it is suspicious, it can give a breakdown of the mitotic count and the hotspot. I would check that against a manual count from the Pathologist”.

When testing a smaller, specialized AI program specific to breast cancer, Yang noticed some discrepancies and cases the AI would miss. In her next study, she used a program trained on a much larger data set and saw a big difference. “This software was trained on over 1.5 million scans, one of the largest teaching sets so far in our field. Many of the cases, particularly the rare breast cancers, were still recognized. It showed how quickly these AI models are developing and becoming more accurate”.

What Yang realized was the scale and collaboration needed for AI to become fully embedded in Pathology practice. “AI needs scale. We need to collaborate across industries, institutions, and countries to get large enough datasets and to really see the benefits of AI. The more data the AI has, the more accurate, and therefore, useful, it becomes for us”.

Now that she is part of Sinai Health and back at LMP, Yang hopes to be part of the driving force for digital pathology. Sinai Health recently received a donation to support its digital transformation.

“Hearing from others who are at the forefront of digital pathology, it isn’t just about the technology, it’s the hardware, workflows and people, including a certain amount of culture shock as we incorporate tools like this into our work. It takes a village to succeed, and I am excited to see that happen here in Toronto”.