CAPTOR: a vision-language foundation model for Chest-Abdomen-Pelvis Oncologic CT
Purpose : To develop and evaluate CAPTOR, a deep-learning vision-language foundation model for oncologic chest-abdomen-pelvis (CAP) CT interpretation. Materials & Methods : CAPTOR (Chest-Abdomen-Pelvis foundaTion model for Oncologic Radiology) is a contrastive vision-language model, trained using a…
The article presents CAPTOR, a vision-language foundation model designed for interpreting oncologic chest-abdomen-pelvis (CAP) CT scans, aiming to improve the accuracy and efficiency of cancer diagnosis and treatment planning.
What happened
The article describes the development and evaluation of CAPTOR, a deep-learning vision-language foundation model tailored for oncologic CT interpretation. The model was trained on a diverse dataset from multiple regions and tested on various tasks, demonstrating superior performance compared to existing models.
Why it matters
CAPTOR's improved accuracy in disease detection and survival prediction can enhance diagnostic precision and treatment planning in oncology, potentially leading to better patient outcomes. Its performance highlights the value of domain-specific pretraining in medical AI applications.
- Clinically aligned pretraining significantly improves the performance of AI models in medical imaging.
- CAPTOR's success demonstrates the potential of vision-language models in oncology applications.
- Open-sourcing code and models promotes transparency and accelerates research in the field.
The bigger picture
The model's superior performance in oncologic CT interpretation could lead to more accurate and efficient diagnostic tools, benefiting both clinicians and patients. The release of code and model weights encourages further research and application in medical imaging.
CAPTOR represents a significant advancement in medical AI, offering improved accuracy in oncologic CT interpretation through specialized pretraining and evaluation on critical tasks, with potential to enhance clinical workflows and patient care.
Technical details
- Contrastive vision-language model architecture
- Pretraining on a multicenter retrospective dataset
- Evaluation on disease detection, survival prediction, and image retrieval tasks
- Bootstrap resampling for confidence interval estimation
- Permutation tests for pairwise comparisons
FAQ
What is CAPTOR?
CAPTOR is a vision-language foundation model designed for interpreting oncologic chest-abdomen-pelvis (CAP) CT scans, developed to improve cancer diagnosis and treatment planning.
How was CAPTOR evaluated?
CAPTOR was evaluated on three downstream tasks: disease detection, survival prediction using TumorImagingBench, and image retrieval, with performance compared to state-of-the-art models.
What datasets were used to train and test CAPTOR?
CAPTOR was trained on a multicenter retrospective dataset from North America, Europe, and Asia, and tested on internal and public test sets stratified by anatomical coverage and indication.
This summary and analysis were generated with AI from the original article and may contain errors. Read the original for the full story.
Read the full story
Originally published by arXiv - cs.CV (CV) on Oct 14, 2026.
Comments
Sign in to join the discussion.