Medical Image Analysis and Causal Machine Learning
Principal Investigator
About this CDAS Project
Study
PLCO
(Learn more about this study)
Project ID
PLCOI-1872
Initial CDAS Request Approval
Apr 23, 2025
Title
Medical Image Analysis and Causal Machine Learning
Summary
This project focuses on advancing X-ray imaging analysis through causal machine learning to improve early diagnosis and mechanistic understanding of lung diseases and cancers. While X-rays are widely accessible, their diagnostic utility is often limited by low contrast, overlapping anatomical structures, and reliance on correlation-driven AI models. We propose a novel causal framework to address these challenges, combining deep learning for feature extraction with causal inference to identify robust, interpretable relationships between X-ray biomarkers, clinical factors, and disease outcomes.
Aims
The research aims to: (1) Enhance diagnostic accuracy by training causal-aware models to detect subtle lung lesions while correcting for biases from X-ray technical variations (e.g., dose, positioning); (2) Discover causal drivers of disease progression (e.g., tumor growth patterns linked to biomarkers) using causal discovery algorithms on paired X-ray and clinical data; (3) Predict treatment responses by simulating counterfactual X-ray outcomes (e.g., post-intervention tumor regression) to guide personalized therapy.
Collaborators
University of Exete
University of Strathclyde
University of Birmingham
University of Glasgow
NHS Greater Glasgow and Clyde
Glasgow Royal Infirmary
Queen Elizabeth University Hospital