CLIP-Driven Universal Model for Organ Segmentation and Tumor Detection

TLDR

Public datasets have advanced automated organ segmentation and tumor detection, yet their small size, partial labeling, and limited tumor diversity constrain models to specific organs/tumors and neglect anatomical semantics, hindering generalization to new domains. The authors propose a CLIP‑Driven Universal Model that integrates CLIP text embeddings into segmentation networks to overcome these limitations. Using CLIP‑based label encoding, the model learns structured feature embeddings that enable segmentation of 25 organs and 6 tumor types, trained on 3,410 CT scans from 14 datasets and evaluated on 6,162 external scans from 3 additional datasets. The model ranks first on the MSD leaderboard, achieves state‑of‑the‑art results on BTCV, runs six times faster than dataset‑specific models, generalizes better across sites, and shows stronger transfer learning on novel tasks.

Abstract

An increasing number of public datasets have shown a marked impact on automated organ segmentation and tumor detection. However, due to the small size and partially labeled problem of each dataset, as well as a limited investigation of diverse types of tumors, the resulting models are often limited to segmenting specific organs/tumors and ignore the semantics of anatomical structures, nor can they be extended to novel domains. To address these issues, we propose the CLIP-Driven Universal Model, which incorporates text embedding learned from Contrastive Language-Image Pre-training (CLIP) to segmentation models. This CLIPbased label encoding captures anatomical relationships, enabling the model to learn a structured feature embedding and segment 25 organs and 6 types of tumors. The proposed model is developed from an assembly of 14 datasets, using a total of 3,410 CT scans for training and then evaluated on 6,162 external CT scans from 3 additional datasets. We rank first on the Medical Segmentation Decathlon (MSD) public leaderboard and achieve state-of-the-art results on Beyond The Cranial Vault (BTCV). Additionally, the Universal Model is computationally more efficient (6× faster) compared with dataset-specific models, generalized better to CT scansfrom varying sites, and shows stronger transfer learning performance on novel tasks.