APED-CXR: Adaptive Privileged Evidence Distillation for Chest X-Ray Grounding with Image-Only Inference

Md Shihabul Islam Shovo; Nishi Kanta Paul; Jamil Fayyad

cs.CVOct 21, 2026

APED-CXR: Adaptive Privileged Evidence Distillation for Chest X-Ray Grounding with Image-Only Inference

Authors: Md Shihabul Islam Shovo, Nishi Kanta Paul, Jamil Fayyad

Organizations: NIMISHES Lab, Dhaka, Bangladesh · American University of Sharjah, Sharjah, UAE

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.

Figures & tables

Figure 1: Overview of APED-CXR. The student predicts disease probabilities $\hat{\bm{y}}_{c}$ and boxes $\hat{\bm{B}}_{c}$ from evidence maps $\bm{A}_{c}$ using the image alone. In training, a frozen teacher verifies proposals $\{\bm{B}_{j}\}$ ; its support scores $\bm{p}_{T}$ supervise the box-pooled scores $\bm{r}_{S}$ . The switch $\oslash$ selects proposals from the current student $S_{t}$ (APED-CXR) or the warm-up checkpoint $S_{0}$ (static comparator C2). Maps and boxes are schematic.

Mode	Inputs	AUROC $\uparrow$	Top-1 $\uparrow$	$\rho_{\mathrm{IoU}}$ $\uparrow$	All-bad $\downarrow$	ms
Image only	$\bm{x}$ , $c$ , $B$	0.716	0.46	0.37	0.41	318
Report only	$\mathcal{R}$ , $c$ , $B$	0.689	0.43	0.31	0.46	86
Report rules	$\mathcal{R}$ , $c$ , $B$	0.664	0.40	0.28	0.38	$<$ 1
Generation	$\bm{x}$ , $\mathcal{R}$ , $c$	0.652	0.39	0.30	0.19	1,380
Image+report	$\bm{x}$ , $\mathcal{R}$ , $c$ , $B$	0.781	0.55	0.46	0.27	333

Table 1: Teacher scoring of grounding-val proposals from image

\bm{x}

, report

\mathcal{R}

, class

c

, and box

B

; generated boxes are mapped to proposals by overlap. Acceptable: IoU

\geq

0.5 with a same-class box. AUROC: acceptable vs. other; Top-1: best-scored is acceptable;

\rho_{\mathrm{IoU}}

: Spearman correlation with IoU; All-bad: mean support if none is acceptable. Time: ms per image–class pair.

ID	Spatial supervision	Micro-gF1 $\uparrow$	$\Delta$ vs. static [95% CI]	Macro-gF1 $\uparrow$	AUC $\uparrow$	Teacher GPU-h $\downarrow$
C0	None (image-level labels only)	0.140 $\pm$ 0.014	$-$ 0.165 [ $-$ 0.188, $-$ 0.144]	0.101	0.812	0
C1	Manual boxes only	0.284 $\pm$ 0.009	$-$ 0.017 [ $-$ 0.030, $-$ 0.004]	0.201	0.809	0
C2	Teacher on fixed warm-up proposals ( $S_{0}$ )	0.302 $\pm$ 0.005	–	0.215	0.813	5.9
C3	Teacher on refreshed random boxes	0.311 $\pm$ 0.022	+0.006 [ $-$ 0.010, +0.022]	0.223	0.811	57.5
C5	Teacher on a fixed dense proposal bank	0.308 $\pm$ 0.008	+0.004 [ $-$ 0.016, +0.022]	0.224	0.813	47.4
C6	Report rules on current proposals ( $S_{t}$ )	0.311 $\pm$ 0.009	+0.006 [ $-$ 0.011, +0.023]	0.222	0.810	0

Table 2: Box-level grounding on the held-out PadChest-GR test partition (IoU

\geq

0.5; mean

\pm

SD over five seeds for C1, C2, and Full, three otherwise).

\Delta

: paired difference from C2 over common seeds (patient-bootstrap 95% CI). All rows except C0 use the manual box loss.

Figure 2: (a) Grounding-val micro-gF1 across refreshes and (b) per-class gF1 on the PadChest-GR test partition (mean

\pm

SD over seeds).

Model	Params	GMACs	ms	GB	gF1 $\uparrow$	AUC $\uparrow$
MedGemma	4B	–	1,910	9.6	0.189	–
APED-CXR	11.1M	5.7	9.6	0.38	0.338	0.815

Table 3: Inference cost per image (A100, batch 1; ms, peak GB) and PadChest-GR test results; MedGemma generates boxes zero-shot from the image and class names.

Medical Image Grounding

Medical image grounding links clinical phrases, questions, and diagnoses to the image regions or slices that support them.

See all 80 papers

X-Ragent: Tool-Augmented Reinforcement Learning for Evidence-Grounded Chest X-Ray Reasoning
Oct 12, 2026Yi Wang, Zhenlong Yuan, Yuchan Guo +9Medical VLMs Medical Imaging
Institute of Computing Technology · University of Chinese Academy of Sciences · Carnegie Mellon University +6
CXRMate-2: Structured Multimodal Temporal Embeddings and Tractable Reinforcement Learning for Clinically Acceptable Chest X-ray Radiology Report Generation
Apr 21, 2026Aaron Nicolson, Elizabeth J. Cooper, Hwan-Jin Yoon +7Radiology Report Generation Radiology VLMs
Australian e-Health Research Centre, CSIRO Health and Biosecurity, Brisbane & Melbourne, Australia · Princess Alexandra Hospital, Metro South Health, Brisbane, Australia · Queensland Digital Health Centre, University of Queensland, Brisbane, Australia +2
CARE-X: Towards Clinically Useful Radiology VLMs with Auxiliary Supervision, Reward-Aligned Learning, and Tool-Augmented Measurement
Aug 4, 2026Mercy Prasanna Ranjit, Anirban Porya, Sathvik Joel +8Radiology Report Generation Chest X-Ray Classification
Microsoft Research India · Medha AI, Narayana Health, India. · RTIICS, Narayana Health, India

Oct 12, 2026J/K move · Enter open · S save

X-Ragent: Tool-Augmented Reinforcement Learning for Evidence-Grounded Chest X-Ray Reasoning

Yi Wang, Zhenlong Yuan, Yuchan Guo, Maolei Huang+8

Institute of Computing Technology · University of Chinese Academy of Sciences · Carnegie Mellon University+6

Medical VLMs Medical Imaging RL for Tool Use Agentic RL

Open paper AskPDF arXiv

Chest X-ray (CXR) interpretation often requires integrating visual findings with task-specific evidence rather than relying on direct visual recognition alone. While tool-augmented medical agents can supply such evidence via external modules, they often lack explicit mechanisms for learning when auxiliary evidence is needed and for preventing inapplicable quantitative evidence from entering diagnostic reasoning. To address this, we propose X-Ragent, a post-training framework for tool-augmented reinforcement learning that decouples learned auxiliary-evidence seeking from validity-gated quantitative measurement acquisition for evidence-grounded CXR reasoning. We construct Rad-CoT, a 7.7K-sample corpus combining causally ordered tool-augmented trajectories with report-style diagnostic reasoning, and use it to initialize Qwen3-VL-8B via agentic supervised fine-tuning (SFT). We then optimize the policy with Group Relative Policy Optimization (GRPO) using an accuracy-gated reward that applies process-level terms for measurement consistency and auxiliary-tool cost only to correct trajectories, preventing plausible process behavior from compensating for an incorrect diagnosis. Across three CXR benchmarks, X-Ragent achieves the strongest results, including an 8.6-point improvement on the fully held-out ChestAgentBench. By seeking evidence selectively and admitting only valid measurements, X-Ragent supports more faithful and evidence-grounded CXR reasoning.

Abstract

Medical Image Grounding

Explore similar work

X-Ragent: Tool-Augmented Reinforcement Learning for Evidence-Grounded Chest X-Ray Reasoning

CXRMate-2: Structured Multimodal Temporal Embeddings and Tractable Reinforcement Learning for Clinically Acceptable Chest X-ray Radiology Report Generation

CARE-X: Towards Clinically Useful Radiology VLMs with Auxiliary Supervision, Reward-Aligned Learning, and Tool-Augmented Measurement