The Lymphoma Channel on VJHemOnc is an independent medical education platform, supported with funding from AstraZeneca (Diamond), BMS (Gold), Johnson & Johnson (Gold), Takeda (Silver) and Galapagos (Bronze). Supporters have no influence on the production of content. The levels of sponsorship listed are reflective of the amount of funding given.
EHA 2025 | LymphoPath: an AI-driven digital pathology platform for risk stratification in DLBCL
Adrian Mosquera-Orgueira, MD, PhD, University Hospital of Santiago de Compostela, Santiago, Spain, discusses the development of LymphoPath, an AI-driven digital pathology platform for risk stratification in diffuse large B-cell lymphoma (DLBCL). Dr Mosquera-Orgueira highlights that the development of digital biomarkers will enable faster risk stratification of DLBCL using hematoxylin and eosin slides at the time of diagnosis. This interview took place at the 30th Congress of the European Hematology Association (EHA) in Milan, Italy.
These works are owned by Magdalen Medical Publishing (MMP) and are protected by copyright laws and treaties around the world. All rights are reserved.
Transcript
For a few years we’ve been discussing how to best risk-stratify DLBCL for newer trials with immunotherapy, for example, and there are lots of developments on molecular analysis. You can see the GUIDANCE trial, for example, is based on molecular characterization of DLBCL and other transcriptomic signatures that are all of them very difficult to take into the clinical setting, particularly because this takes time and lymphoma is a very aggressive disease, particularly DLBCL...
For a few years we’ve been discussing how to best risk-stratify DLBCL for newer trials with immunotherapy, for example, and there are lots of developments on molecular analysis. You can see the GUIDANCE trial, for example, is based on molecular characterization of DLBCL and other transcriptomic signatures that are all of them very difficult to take into the clinical setting, particularly because this takes time and lymphoma is a very aggressive disease, particularly DLBCL. So there are new ways needed to better risk-stratify these patients.
So one of the approaches we have been following during the past three years was to try to link histological images to prognostication. And for that, what we have done is we have digitized all the hematoxylin and eosin slides from our hospital. These are roughly 500 cases. And we have started to analyze how a digital pathology tool, which actually detects the cell nucleoli and extracts lots of metrics from each cell and from the interaction between the tumor and the microenvironment, how does this correlate with risk?
So our initial analysis, which is based on 118 patients, has provided a very strong foundation for the risk stratification of DLBCL based on this approach. We’ve seen very dramatic differences between risk groups predicted by this AI-driven approach based on histological images. And we believe this is a new path to go forward because actually, we need something very fast in DLBCL. And if you can establish risk accurately with hematoxylin and eosin slides that are available from the time of diagnosis, we’ll develop these digital biomarkers, which can be readily applicable as soon as a new DLBCL patient is diagnosed. You would not need to wait a little bit until the molecular data and you can start treatment in a clinical trial. So very promising data. Obviously, we need additional validations and to expand our cohort, but this is going to be done.
This transcript is AI-generated. While we strive for accuracy, please verify this copy with the video.
Read more...
