Foundation models show lower accuracy on pediatric chest X-rays
From Adults to Children: Benchmarking Foundation Models for Chest X-Ray Classification under Domain Shift
Computer Vision and Pattern Recognition
Summary
Medical AI models often learn from adult chest X-ray images but may not work as well on children's images. The authors tested several big AI models trained only on adult data to see how well they classify diseases in pediatric X-rays. They found all models performed much worse on children's images, dropping accuracy by over 10%. Adjusting the models more deeply did not improve their ability to handle pediatric data. This means AI trained on adults might not easily adapt to kids without new approaches.
What this means in practice
- •For medical ai developers: Evaluate and improve model robustness when deploying chest X-ray classifiers trained on adult data to pediatric patients.
- •For hospital data teams: Plan validation strategies for AI tools to ensure safe pediatric chest X-ray diagnosis before clinical use.
Authors
Sara Inácio, Hugo Proença, João C. Neves
Abstract
Foundation models have shown significant potential across diverse fields and have been applied to medical imaging, but their robustness across patient populations remains unclear. In this work, we investigate foundation models' ability to generalize from adult to pediatric chest X-ray classification. To the best of our knowledge, this is the first benchmark to compare multiple foundation models under a common adult-to-pediatric evaluation setting while also analyzing the effect of adaptation depth. We harmonize VinDr-CXR and PediCXR into a five-label classification task and evaluate six foundation models and two CNN baselines. All models are adapted using adult data only and tested in both the in-domain and pediatric domains using linear probing, partial fine-tuning, and full fine-tuning. The benchmark reveals a clear and consistent trend: all models suffer a substantial performance degradation when transferred from adult to pediatric data, with drops exceeding 10 percentage points. While Macro-AUROC values reach approximately 92% on the adult test set, performance decreases to around 78% on the pediatric domain. Another important finding is that increasing the depth of adaptation does not significantly improve cross-population generalization. Moreover, strong in-domain performance does not translate into better performance on pediatric data.