Biography

Pierre Baudot is Principal Data Scientist at Median Technologies (Eyonis, Sophia-Antipolis, France), where he works on AI research for cancer detection and risk prediction from medical imaging. He holds a PhD in Neuroscience/Cognitive Science on neural network coding and learning and previously held research positions at CNRS, the Max Planck Institute for Mathematics in the Sciences, and INSERM, where he specialised into Algebraic Topology and Statistics and developed Information Topology, a research program providing novel mathematical foundations of information theory along with statistical characterisation of high dimensional patterns, before applying these methods to AI-ML — work recognized by the 2017 K2 Trophy in Mathematics and Applications. His current research focuses on unsupervised and self-supervised learning for cancer detection on CT, and the development of AI software as medical device supporting clinicians in screening and diagnosis, with expertise in uncertainty quantification and evaluation frameworks for medical AI applications.

Synopsis

Vision-language foundation models (VLMs) have demonstrated broad competence across medical imaging tasks, raising the question of whether they can match purpose-built, specialised AI systems for high-stakes 3D CT screening tasks. This talk presents a retrospective benchmark comparing eight VLM configurations — MedGemma, M3FM, CT-CHAT, MedSigLIP, and COLIPRI (each evaluated zero-shot and/or fine-tuned) — against seven specialized, CT-native cancer detection AI models on the NLST held-out test set (n=2,065; 54 cancers). Specialized models consistently exceed AUC 0.94, led by the Eyonis LCS ensemble at 0.98, while VLMs range from near-chance zero-shot performance to AUC 0.84-0.91 depending on fine-tuning and prompt design — notably, COLIPRI’s malignancy-grounded zero-shot prompting reaches AUC ~0.91 at a fraction of the compute cost (~15 TFLOPs/scan) of larger VLMs (up to 6,700 TFLOPs/scan).

The key driver of this gap is CT-native domain pretraining, not model scale or general vision-language capability: COLIPRI, the only VLM in our benchmark pretrained natively on CT volumes, is also the only one to approach specialised-level AUC, while Pillar-0’s CT-native vision-only encoder matches top CNN specialists using nothing more than a linear probe on frozen embeddings.

To ground these results in clinical relevance, I will also draw on two companion studies benchmarking against radiologists directly. In a 12-reader study, native MedGemma (AUC 0.70) fell short of clinically relevant performance, while fine-tuned MedGemma (AUC 0.83) reached only the level of less-experienced readers (mean radiologist AUC 0.90; range 0.80-0.94). By contrast, in an independent multi-reader evaluation, the specialised Eyonis LCS AI outperformed all radiologists.

Combined with the markedly higher computational burden of generic vision-language models, this gap currently favours CT-native specialised systems for safer near-term clinical use, with fewer diagnostic errors and unnecessary procedures. For the SAFER community, this underscores that faithful evaluation of foundation models must be paired with genuine domain adaptation — or, better still, dedicated 3D CT pretraining from the outset — not just scale, before deployment in screening pipelines.