Chimeric antigen receptor T cell (CAR-T) therapy has revolutionized cancer treatment, but variability in efficacy and toxicity remains a challenge. Current methods to define CAR-T subsets heavily rely on destructive intracellular staining. In this project, we use imaging flow cytometry and machine learning to train classifiers on paired brightfield and stained cell images, then transfer this knowledge to live, unstained cells. Our goal is to identify label-free morphological signatures that predict CAR-T phenotypes linked to stronger tumor killing or reduced exhaustion. This approach could reveal which subsets are best suited for different therapies and advance safer, more effective immunotherapies.

I am participating in SuperUROP to gain experience applying AI to biological engineering, particularly in using machine learning to classify T cells for cancer immunotherapy. My background in biomedical imaging, simulations, and machine learning prepared me for this work. I’m excited to take on more independent research and to build tools that connect AI with translational biology.