Learning Robust Visual Features in Computed Tomography Enables Efficient Transfer Learning for Clinical Tasks

TL;DR

The paper introduces VoxelFM, a 3D CT foundation model trained with self-distillation, which learns robust visual features enabling efficient transfer to various clinical tasks without the need for language supervision or backbone fine-tuning.

Contribution

Proposes VoxelFM, a self-distilled 3D CT foundation model that outperforms existing models across multiple tasks using frozen features and lightweight probes.

Findings

VoxelFM matches or exceeds performance of existing models across seven clinical tasks.

It surpasses language-supervised models even without language alignment during training.

Current CT foundation models are more effective as feature extractors than as vision-language encoders.

Abstract

There is substantial interest in developing artificial intelligence systems to support radiologists across tasks ranging from segmentation to report generation. Existing computed tomography (CT) foundation models have largely focused on building generalist vision-language systems capable of tasks such as question answering and report generation. However, training reliable vision-language systems requires paired image-text data at a scale that remains unavailable in CT. Moreover, adapting the underlying visual representations to downstream tasks typically requires partial or full backbone fine-tuning, a computationally demanding process inaccessible to many research groups. Instead, foundation models should prioritise learning robust visual representations that enable efficient transfer to new tasks with minimal labelled data and without backbone fine-tuning. We present VoxelFM, a 3D CT foundation model trained with self-distillation using the DINO framework, which learns semantically rich features without language supervision. We evaluated VoxelFM across seven categories of clinically relevant downstream tasks using frozen backbone representations with lightweight probes: classification, regression, survival analysis, instance retrieval, localisation, segmentation, and report generation. VoxelFM matched or outperformed four existing CT foundation models across all task categories. Despite receiving no language supervision during pre-training, VoxelFM surpassed models explicitly trained with language-alignment objectives, including on report generation. Our results indicate that current CT foundation models perform significantly better as feature extractors for lightweight probes rather than as vision encoders for vision-language models. Model weights and training code are publicly available.