Causal and Active Learning-Based Counterfactual Chest X-ray Generation for Supporting Clinical Decision-Making in Lung Disease

Yifei Zhu, Greta Mohr, Lei Zhang, Christopher Sainsbury, Feng Dong, John D Maclay, David J Lowe, David Lagnado, Xujiong Ye
Proceedings of the Fifth Conference on Causal Learning and Reasoning, PMLR 323:1635-1655, 2026.

Abstract

Lung diseases such as lung cancer are major contributors to global morbidity, requiring accurate diagnostic decisions for optimal patient outcomes. While deep learning has advanced medical imaging, the lack of causal inference limits its clinical utility. This study proposes a causal generative framework for counterfactual analysis of Chest X-rays, guided by expert model supervision to ensure clinical plausibility. To solve data imbalance and enhance robustness, we introduce a recurrent active learning strategy that utilises "forgetting rates" to select informative samples. Experimental results demonstrate effectiveness improvements of 9.25% on the MIMIC dataset and 13.40% on ChestXray8. Furthermore, two-stage human expert evaluations confirm that the model generates highly realistic synthetic data that maintains a clinical heavy-tailed distribution. These high-quality counterfactuals not only improve diagnostic accuracy but also facilitate confidence calibration for clinicians through interpretable evidence. Our findings demonstrate that integrating causal modeling with expert supervision and active learning provides a robust, clinically meaningful tool for pulmonary diagnostic decision-making.

Cite this Paper

BibTeX
@InProceedings{pmlr-v323-zhu26b, title = {Causal and Active Learning-Based Counterfactual Chest X-ray Generation for Supporting Clinical Decision-Making in Lung Disease}, author = {Zhu, Yifei and Mohr, Greta and Zhang, Lei and Sainsbury, Christopher and Dong, Feng and Maclay, John D and Lowe, David J and Lagnado, David and Ye, Xujiong}, booktitle = {Proceedings of the Fifth Conference on Causal Learning and Reasoning}, pages = {1635--1655}, year = {2026}, editor = {Mazaheri, Bijan and Hanson, Niels Richard}, volume = {323}, series = {Proceedings of Machine Learning Research}, month = {06--08 Apr}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v323/main/assets/zhu26b/zhu26b.pdf}, url = {https://proceedings.mlr.press/v323/zhu26b.html}, abstract = {Lung diseases such as lung cancer are major contributors to global morbidity, requiring accurate diagnostic decisions for optimal patient outcomes. While deep learning has advanced medical imaging, the lack of causal inference limits its clinical utility. This study proposes a causal generative framework for counterfactual analysis of Chest X-rays, guided by expert model supervision to ensure clinical plausibility. To solve data imbalance and enhance robustness, we introduce a recurrent active learning strategy that utilises "forgetting rates" to select informative samples. Experimental results demonstrate effectiveness improvements of 9.25% on the MIMIC dataset and 13.40% on ChestXray8. Furthermore, two-stage human expert evaluations confirm that the model generates highly realistic synthetic data that maintains a clinical heavy-tailed distribution. These high-quality counterfactuals not only improve diagnostic accuracy but also facilitate confidence calibration for clinicians through interpretable evidence. Our findings demonstrate that integrating causal modeling with expert supervision and active learning provides a robust, clinically meaningful tool for pulmonary diagnostic decision-making.} }
Endnote
%0 Conference Paper %T Causal and Active Learning-Based Counterfactual Chest X-ray Generation for Supporting Clinical Decision-Making in Lung Disease %A Yifei Zhu %A Greta Mohr %A Lei Zhang %A Christopher Sainsbury %A Feng Dong %A John D Maclay %A David J Lowe %A David Lagnado %A Xujiong Ye %B Proceedings of the Fifth Conference on Causal Learning and Reasoning %C Proceedings of Machine Learning Research %D 2026 %E Bijan Mazaheri %E Niels Richard Hanson %F pmlr-v323-zhu26b %I PMLR %P 1635--1655 %U https://proceedings.mlr.press/v323/zhu26b.html %V 323 %X Lung diseases such as lung cancer are major contributors to global morbidity, requiring accurate diagnostic decisions for optimal patient outcomes. While deep learning has advanced medical imaging, the lack of causal inference limits its clinical utility. This study proposes a causal generative framework for counterfactual analysis of Chest X-rays, guided by expert model supervision to ensure clinical plausibility. To solve data imbalance and enhance robustness, we introduce a recurrent active learning strategy that utilises "forgetting rates" to select informative samples. Experimental results demonstrate effectiveness improvements of 9.25% on the MIMIC dataset and 13.40% on ChestXray8. Furthermore, two-stage human expert evaluations confirm that the model generates highly realistic synthetic data that maintains a clinical heavy-tailed distribution. These high-quality counterfactuals not only improve diagnostic accuracy but also facilitate confidence calibration for clinicians through interpretable evidence. Our findings demonstrate that integrating causal modeling with expert supervision and active learning provides a robust, clinically meaningful tool for pulmonary diagnostic decision-making.
APA
Zhu, Y., Mohr, G., Zhang, L., Sainsbury, C., Dong, F., Maclay, J.D., Lowe, D.J., Lagnado, D. & Ye, X.. (2026). Causal and Active Learning-Based Counterfactual Chest X-ray Generation for Supporting Clinical Decision-Making in Lung Disease. Proceedings of the Fifth Conference on Causal Learning and Reasoning, in Proceedings of Machine Learning Research 323:1635-1655 Available from https://proceedings.mlr.press/v323/zhu26b.html.

Related Material