The non-invasive assessment of renal masses remains a critical challenge in urologic oncology. Here, the authors develop RenalCLIP, a vision-language foundation model for renal mass assessment and classification using CT scans from 8,809 patients across Chinese and international cohorts, outperforming other models in diagnostic classification at even 20% of the training data.
- Yuhui Tao
- Zhongwei Zhao
- Shuo Wang
