Hepatocellular carcinoma (HCC) is the most common type of primary liver cancer and one of the deadliest cancers worldwide. While medical treatments for HCC have evolved, predicting which patients will benefit from specific therapies remains a major challenge. This project leverages artificial intelligence to uncover novel biomarkers that can predict how patients respond to treatments such as anti-angiogenic drugs and immune checkpoint inhibitors.
By applying deep learning (DL) techniques to whole slide images (WSI) of liver biopsies, the project aims to identify histological patterns that correlate with treatment outcomes. Recent research has shown that DL models can predict genetic signatures directly from tissue images, including the ABRS gene signature, which has been associated with progression-free survival in patients receiving a combination of atezolizumab and bevacizumab. Building on these advances, this project will validate and refine these AI models to predict HCC treatment response from histopathology images.
The study involves analyzing biopsy samples from HCC patients treated with various therapies. Using advanced methods like Multiple Instance Learning, AI algorithms will extract features from these images to identify potential biomarkers that correlate with patient outcomes. These biomarkers could provide a more accessible alternative to expensive genetic tests, enabling clinicians to personalize treatment plans based on a patient’s unique tumor profile.
Within a hospital setting, this AI-driven approach could help reduce the workload in pathology departments and improve diagnostic accuracy, ensuring that patients receive optimal, tailored care.