Universal CT Representations from Anatomy to Disease Phenotype through Agglomerative Pretraining

TL;DR

FlexiCT is a comprehensive CT foundation model trained through agglomerative pretraining on a large dataset, enabling versatile analysis and capturing disease-related features.

Contribution

This work introduces FlexiCT, a novel large-scale CT foundation model trained via multi-stage pretraining for diverse medical imaging tasks.

Findings

FlexiCT matches or exceeds prior methods on multiple benchmarks.

Embeddings organize scans along tumor stage gradients.

Supports slice, volume, and vision-language analysis.

Abstract

Computed tomography (CT) is a central to three-dimensional medical imaging, yet CT-based artificial intelligence remains fragmented across task-specific models for segmentation, classification, registration, and report analysis. Here we present FlexiCT, a family of CT foundation models trained by agglomerative continual pretraining on 266,227 CT volumes from 56 publicly available datasets, forming a large-scale public resource for CT representation learning. FlexiCT uses agglomerative pretraining across three stages: two-dimensional axial pretraining, three-dimensional anatomical pretraining and report-guided semantic alignment. This training strategy supports slice-level, volume-level and vision-language analysis. Across five downstream task families (segmentation, classification, registration, vision-language understanding and clinical retrieval), FlexiCT matches or exceeds prior task-specific approaches on multiple benchmarks. Its embeddings further organize CT scans along gradients associated with various tumor stages, suggesting that CT foundation models can capture imaging features relevant to disease phenotype characterization. Code is available at https://github.com/ricklisz/FlexiCT