ASH 2025 | Lymphovision: an AI model for histology-based lymphoma classification and subtyping

I think this is something I’m really very excited about because I think this is the cusp of the next big thing in lymphomas. So Lymphovision is really utilizing a typical H&E stain done on patients’ biopsies and specifically using an AI-generated large language model learning way of training what the image looks like and correlating that with is it lymphoma versus not, and then correlating that with whether or not the patient has a good outcome or a bad outcome, or also what is the cell of origin and is it associated with one picture or image versus another? So right now where Lymphovision is effective is taking the H&E stain and saying lymphoma versus not lymphoma, and then if lymphoma, identifying about 10 different subtypes of lymphoma, and then on top of that being able to say in large cell lymphoma, for example, the cell of origin is GCB versus ABC and then possibly even the dark zone, double hit, triple hit subtypes. And I think the reason this is very valuable, in seconds you have this answer compared to the usual many days that it takes to complete the immunohistochemistry and often the FISH analysis. So it’s really going to accelerate our ability to bring treatment to patients sooner. And the reason that really matters is many studies have shown those patients that don’t get treated or need to get treated right away and can’t go on trials are the ones that do the worst. This will help to fix that problem.

This transcript is AI-generated. While we strive for accuracy, please verify this copy with the video.