Congratulations to Dr. Kerol Djoumessi for successfully defending his PhD!

In his doctoral thesis, Kerol's research focused on the question: "How can we design deep learning models that doctors can actually understand and trust?" 
Modern neural networks often achieve high diagnostic accuracy, but they usually seem like black boxes. It is difficult for clinicians to verify why a model made a particular decision, and especially when it comes to medical decisions, this lack of transparency is a big barrier for deployment.

During his PhD, Kerol developed inherently interpretable deep learning models for medical image analysis, specifically for ophthalmology. Rather than relying on post-hoc explanations applied to opaque models, his approach builds interpretability directly into the model. Over the course of his PhD, Kerol contributed five peer-reviewed papers presented at leading venues including MICCAI, MIDL, and PLOS Digital Health.

But Kerol's path is an achievement in itself. He says: "When I started in 2021, I had just graduated from AIMS South Africa with only a short introduction to machine learning. I wasn’t familiar with most deep learning concepts and had never trained a neural network before starting my PhD. What I did have, though, was a strong motivation to work on AI for medical imaging, and that motivation carried me throughout this journey."

Thanks for the inspiring work, Kerol, and we are excited to see where it takes you next! We wish you all the best for the path ahead.