Dr Andrew Srisuwananukorn on the Potential for AI in Differentiating prePMF and ET

An artificial intelligence (AI) model was able to differentiate between primary myelofibrosis (prePMF) and essential thrombocythemia (ET) with 92.3% accuracy by examining digital whole-slide images, according to a study presented at the 2023 American Society of Hematology Annual Meeting and Exposition.

AI-based decision support tools have potential to increase diagnostic accuracy for physicians in the community who may not see patients with prePMF or ET often, said Andrew Srisuwananukorn, MD, of Ohio State University Comprehensive Cancer Center, lead author of the study.

In this interview, Srisuwananukorn discussed the potential benefits and considerations for the development of AI algorithms to help diagnose these conditions.

Transcript

What is the potential value of implementing AI to assist in appropriately diagnosing patients with prePMF and ET in the clinical setting?

I view the benefit of a potential algorithm such as the one that we've created to be used ubiquitously across multiple centers. I find that the value of such a tool might be helpful in community practices that don't necessarily see myeloproliferative neoplasms on a consistent basis. These algorithms are cheap and affordable to be used and are equitable across different countries.

As AI use becomes more common, what can be done to ensure these algorithms are developed effectively and ethically?

I think there are 2 aspects that we really should be considering as we develop these AI algorithms. Number 1, it's important for us to understand that the algorithm was developed on a patient cohort, and we really want that patient cohort to be representative of the general population. It might accidentally learn a feature of the cohort that has no basis in biology, so it's important that our algorithm is representative of all patient cohorts of at-risk populations.