AI-based analysis of 3D organoids – Accelerating translational cancer research

My name is Christophe Deben, and I work at the Center for Oncological Research in the University of Antwerp, which is located in Belgium. I'm head of the Tumoroid Screening Lab, and I'm also involved in a university spinoff called OrBITS, where we develop AI-based image and data analysis pipelines.

I've always been very enthusiastic about new technologies and applied that enthusiasm to research technologies. In the future, doing my Ph.D., it was always very low throughputs. We spend a lot of time looking just into a single research question, and I really see that these technologies allow us to study many more different combination therapies, for example, on a much shorter time, so really advancing the research much faster.

I was very interested in the organoid technology because I really think it's a good model to study the biology of cancer, and I'm especially excited about the potential clinical applications where you can predict therapy response with these organoids. And that is also what we're focusing on now in developing new methods to do drug screening, high-throughput drug screening, using live-cell imaging and using the OrBITS image analysis to get more relevant information out of it.

So, in comparison to cell lines, you buy them from these companies. They're always the same cell lines, the same backgrounds, and eventually, they're in culture for a very long time. So, they do not really have anything to do anymore with the tumor they're derived from. And now, with the organoids, we have the potential to grow organoids from patients at the university hospital ourselves. So very low-passage organoids, keeping heterogeneity and the link to the clinic.

AI really enabled us to get more information of images we gather with a live-cell imaging instrument, and our goal was to do the label-free analysis and standardize the analysis method.

So, we don't need to manually put in the settings to characterize the organoids, but everything is fully automated, increasing reproducibility and reliability. AI really enables us to automate the full data and image analysis process. The time we acquired the live-cell imaging instruments, in the same time, a colleague was hired in our group, and his specialty was actually image segmentation and AI image analysis.

So, we really started collaborating together. We worked, I think, more than a year or two really close together to see the image analysis was correct. So really applying a multi-disciplinary approach to organoid drug screening and live-cell imaging. The Spark Cyto instrument really is the most-used instrument, actually, in our lab because it's not only a live-cell imager but also absorbance, luminescence, fluorescence, even alpha technology.

We were suddenly able to apply five different new technologies in the lab with just one instrument. So really helpful for our research. For me, what inspires me for the organoid technology is really the translation of what we do in lab to the patient. We're looking into now retrospectively how our methodology can actually predict therapy response, looking at progression-free survival of these patients, and we really see a very high correlation.

For me, the biggest value there is to, hopefully, in the future, predict therapy response, do a broad screen of therapies, and get the right therapy to the patients, really working towards personalized medicine. Looking into the future of AI and also personalized therapy, I think it can really help us to find the best models to predict therapy response, not only looking into the therapy response we get out of the organoids but also including omics data, the mutations, transcriptome, kinome.

So, really setting up the collaborations between this multi-disciplinary approach within university to get the best model to predict therapy response in these patients.