AI model uses routine imaging to identify patients at risk for serious treatment-induced lung inflammation

Researchers at The University of Texas MD Anderson Cancer Center have created an artificial intelligence (AI) model that can identify lung cancer patients at higher risk of developing serious immunotherapy-induced lung inflammation prior to commencing treatment. This advance offers potential avenues for more personalized monitoring and preventive strategies.

The study, published in the Journal for ImmunoTherapy of Cancer, demonstrates that standard medical imaging may contain predictive indicators of pneumonitis, a potentially life-threatening lung inflammation affecting around 10% of lung cancer patients undergoing immunotherapy. Researchers led by Jia Wu, Ph.D., developed the Checkpoint-Inhibitor Pneumonitis Hazard EstimatoR (CIPHER), an AI model trained on over 590,000 CT image slices from 2,500 lung cancer patients.

The model learned to recognize subtle lung tissue patterns associated with future risk during pretreatment CT scans from 347 non-small cell lung cancer patients. CIPHER outperformed conventional clinical-factor and radiomics models in internal and external cohorts, achieving an area under the curve (AUC) of approximately 0.83. This suggests routine imaging may hold more information about treatment toxicity than previously recognized.

While the model did not focus on detecting pneumonitis itself, it identified subtle abnormalities linked to increased risk. However, prospective studies involving larger, more diverse patient populations are necessary to determine the model's potential integration into clinical workflows. Future research may explore the model's applicability to other cancer types treated with immunotherapy and whether combining imaging data with other biomarkers can enhance risk prediction.