Search papers, labs, and topics across Lattice.
The paper introduces CARS, a framework for generating synthetic chest X-ray images with targeted clinical feature perturbations while preserving anatomical structure, to address the underrepresentation of critical clinical feature combinations in existing datasets. CARS applies controlled insertion and deletion of pathological findings based on clinical feature vectors. Fine-tuning chest X-ray models on CARS-generated images improves precision-recall performance, reduces predictive uncertainty, and improves model calibration compared to prior feature perturbation approaches, as validated by expert radiologists and quantitative metrics.
Synthetic chest X-rays, generated with targeted clinical feature perturbations and anatomical grounding, can significantly boost the performance and trustworthiness of diagnostic models, filling critical gaps in real-world datasets.
The clinical deployment of AI diagnostic models demands more than benchmark accuracy - it demands robustness across the full spectrum of disease presentations. However, publicly available chest radiographic datasets systematically underrepresent critical clinical feature combinations, leaving models under-trained precisely where clinical stakes are highest. We present CARS, a clinically aware and anatomically grounded framework that addresses this gap through principled synthetic image generation. CARS applies targeted perturbations to clinical feature vectors, enabling controlled insertion and deletion of pathological findings while explicitly preserving anatomical structure. We evaluate CARS across seven backbone architectures by fine-tuning models on synthetic subsets and testing on a held-out MIMIC-CXR benchmark. Compared to prior feature perturbation approaches, fine-tuning on CARS-generated images consistently improves precision-recall performance, reduces predictive uncertainty, and improves model calibration. Structural and semantic analyses demonstrate high anatomical fidelity, strong feature alignment, and low semantic uncertainty. Independent evaluation by two expert radiologists further confirms realism and clinical agreement. As the field moves toward regulated clinical AI, CARS demonstrates that anatomically faithful synthetic data generation for better feature space coverage is a viable and effective strategy for improving both the performance and trustworthiness of chest X-ray classification systems - without compromising clinical integrity.
