AI scans of routine CTs may flag lung cancer patients prone to immunotherapy side effect
MD Anderson researchers developed an AI model called CIPHER that analyzes pre-treatment chest CT scans to predict risk of immunotherapy-induced pneumonitis. The model outperformed current methods that rely on subjective imaging and clinical factors. The findings, published in the Journal for ImmunoTherapy of Cancer, could enable personalized monitoring and prevention.
The CIPHER model was trained on over 590,000 CT image slices from 2,500 lung cancer patients, first learning general patterns in lung tissue before being tested for its ability to predict pneumonitis risk. In validation, it achieved an AUC of roughly 0.83 across both an internal cohort of 347 non-small cell lung cancer patients and an independent external dataset, despite variations in scanners and imaging protocols.
High-risk patients identified by the model tended to develop pneumonitis sooner after starting immunotherapy, and predictions remained significant after adjusting for age and smoking history. This suggests the AI detects underlying lung vulnerability rather than merely flagging future cases, offering a potential advantage over subjective imaging analysis and conventional clinical risk factors.
This tool could allow oncologists to identify immunotherapy patients at elevated risk for pneumonitis before treatment begins, enabling closer monitoring or preventive strategies. If widely adopted, it may reduce severe complications and improve quality of life for the roughly 10% of lung cancer patients who experience this side effect. However, broader clinical integration would require validation across diverse populations and healthcare settings.