Compact chest X-ray model improves disease localization using report hints
APED-CXR: Adaptive Privileged Evidence Distillation for Chest X-Ray Grounding with Image-Only Inference
Computer Vision and Pattern Recognition
Summary
Chest X-rays can show signs of disease, but marking exactly where those signs are usually needs expert-drawn boxes, which are hard to get. The authors developed a smaller computer model that learns from a bigger model which has read related doctor reports. During training, the big model helps the small one focus on the right image areas, but at test time only the X-ray image is needed. This approach helps the small model better pinpoint disease locations without losing accuracy in detecting the diseases.
What this means in practice
- •For hospital it teams: Integrate compact CXR models that accurately highlight disease areas using only images to speed up radiology workflows.
- •For medical device developers: Develop lightweight AI tools for portable X-ray devices that provide disease localization without needing accompanying reports at inference.$Commercial implications: Enables selling improved portable X-ray software that helps clinicians identify disease regions quickly without expensive multimodal inputs.
Authors
Md Shihabul Islam Shovo, Nishi Kanta Paul, Jamil Fayyad
Abstract
Compact chest X-ray (CXR) models that localize findings suit deployment, but box annotations are limited and report-aware multimodal models are expensive at inference. We introduce APED-CXR, which distills report-informed spatial evidence from a frozen multimodal teacher into a student with 11.1M parameters that predicts diseases and boxes from shared evidence maps and requires the image alone at test time. During training, the teacher uses the radiograph, report, and queried class to evaluate regions proposed from the student's current evidence maps, and its support scores serve as training targets for the student's pooled evidence scores within these regions. Trained on PadChest and evaluated on held-out PadChest-GR, refreshing the student's proposals for teacher evaluation improves box-level localization F1 from 0.302 with fixed warm-up proposals to 0.338 (IoU >= 0.5; gain 0.036, 95% CI [0.022, 0.049]), while maintaining diagnostic AUC. This gain depends on the teacher checking current proposals: with equally many random boxes, a fixed dense box set, teacher-generated boxes, or report rules instead of the teacher, F1 stays near the fixed-proposal level, and proposals two refreshes old recover only part of the gain. On NIH ChestX-ray14, a secondary test set, box recall also rises from 0.423 to 0.447.