AACR 2026 | Using AI to predict liver decompensation risk in HCC with cirrhosis

So what I propose to you, so we will just talk about HCC, so hepatocellular carcinoma, and to speak, you know, about the two parts of the disease because most of the time HCC will occur on cirrhosis. So this is an important part for the patient. And also they will present HCC at the same time. So I propose you to go more in detail from two parts of my talk. One regarding the importance of cirrhosis, decompensation, how to present this and to better identify the patient with, for example, artificial intelligence. So this is a paper we just published. And the other part will be new data that I will show about how to identify the patient who would respond to immunotherapy in the setting of HCC, using blood from the patient, and the fact that we identify new signatures that can help to identify the patient who will respond to immunotherapy. So as I said just before, when we are speaking about HCC, so most of the time you will have cirrhosis. And so when you have an advanced HCC, now it’s recommended to propose immunotherapy to this patient. And we know that in the series of patients treated in the real world, let’s say, that some of them will present liver decompensation under treatment. And this can be really bad for them because when they present, for example, bleeding or ascites, this can lead to the discontinuation of the treatment. But we know that we have some treatment that we can offer to these patients in order to prevent these decompensations, such as beta-blockers that are used in patients with cirrhosis and without HCC. And so the idea is to identify this patient with clinically significant portal hypertension that could benefit from beta-blockers so we can prescribe, in addition to treatment, for example, these beta blockers in order to prevent decompensation. And so in order to identify this specific population of patients, what we did is that we took the baseline CT scan of the patient because all our patients with cancer, we are used to doing an imaging before starting the treatment. And we use a foundation model, so artificial intelligence, on these CTs in order to identify a higher group of patients associated with liver decompensation under treatment with atezolizumab-bevacizumab. And thanks to this technique, we were able to distinguish two groups of patients using the baseline CT. So the paper is now published in the Journal of Hepatology this year. And so if you want to have a look more in detail and also to use this model, you can have access to it. And the idea is now the next step of these projects will be to try to propose beta blockers to this high-risk group of patients in order to see if we succeed in decreasing the occurrence of liver decompensation on the treatment.

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