Adult kidney MRI models perform well on pediatric tumor segmentation
From Adult Renal Masses to Pediatric Wilms Tumors: Evaluating Cross-Domain Generalization of Kidney Lesion Segmentation on MRI
Computer Vision and Pattern Recognition
Summary
Doctors use MRI scans to find and outline tumors in kidneys, but models trained on adult kidney cancer data might not work well on children's kidney tumors. The authors checked if a computer model trained on adult kidney images could also identify tumors in children. They found the adult model did nearly as well as a model trained on children's data, though the child-specific model was still a bit better. This suggests adult-trained models might help when pediatric data is limited, but more work is needed to improve accuracy.
What this means in practice
- •For medical imaging teams: Use adult kidney MRI-trained segmentation models to assist pediatric tumor analysis when pediatric data is limited, improving workflow efficiency.
- •For radiology software developers: Incorporate adult-trained kidney lesion segmentation models as starting points for pediatric tumor imaging tools, enabling faster adaptation to pediatric needs.
Authors
Sarah de Boer, Myrthe A. D. Buser, Hartmut Häntze, Jawed Nawabi, Lisa C. Adams, Alida F. W. van der Steeg, Marry M. van den Heuvel-Eibrink, Annemieke S. Littooij, Keno K. Bressem, Mathias Prokop, Alessa Hering
Abstract
Deep learning segmentation models for pediatric oncology are often trained on small, single-center cohorts, whereas much larger datasets exist for adult diseases, such as kidney cancer. Whether models trained on adult renal lesions generalize to pediatric Wilms tumors remains unclear. We trained a segmentation model on single-sequence T1- or T2-weighted adult kidney MRI and evaluated it on a held-out adult test set and a pediatric Wilms tumor cohort, where it was compared with a pediatric-specific nnU-Net. Segmentation performance was quantified using the Dice similarity coefficient (DSC) and 95th-percentile Hausdorff distance. On the adult test set, the model achieved a median DSC of 0.86 for renal lesions. On the pediatric cohort, the adult-trained model achieved a median DSC of 0.85 compared with 0.90 for the pediatric-specific model. Despite differences in age, anatomy, and tumor type, the adult-trained model showed promising performance on pediatric Wilms tumors, with only a modest difference in median DSC compared with the pediatric-specific model. However, non-inferiority could not be shown, highlighting the value of pediatric training data. In future work, a fine-tuning strategy could be explored.