ASH 2024 | Automated image analysis pipeline for DLBCL histological slides

So, we see a lot of promise in imaging analysis for personalized healthcare and precision medicine. And the reason is that, you know, although molecular genomics is probably the gold standard for personalized healthcare, this is not fast at all in the clinical work settings. And additionally, it’s restricted to very well developed countries. So the idea was, can we actually develop a system that can infer genomic properties through imaging in diffuse large B-cell lymphoma? And whether this can inform us in the future about global risk for our patients and also particularities about their predisposition to respond to different immunotherapies like CAR-T cells or bispecific antibodies. So the last two years we have been working with a PhD in physics to develop a standardized pipeline for image analysis with DLBCL biopsies. So what the system does is to make it quite straightforward to transform unstructured imaging data into structured information like tabular data that you can then correlate with any clinical endpoint of interest. So it starts by identifying tumor areas and non-tumor areas, then it segments cell nuclei and from each of these nuclei extracts a hundred or so parameters based on morphology, size, intensity and also spatial correlation with other cells. And then it also classifies cells in lymphoma-like cells and microenvironment-like cells. And all this information now, we are planning the next steps for the next ASH to correlate it, obviously, with clinical endpoints like CAR T-cell response.

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