Researchers at The University of Texas MD Anderson Cancer Center have developed an artificial intelligence model that can identify lung cancer patients at increased risk of developing immunotherapy-induced pneumonitis before treatment begins, potentially giving clinicians an opportunity to monitor high-risk patients more closely or intervene before serious complications develop.
Pneumonitis is a potentially life-threatening form of lung inflammation that occurs in approximately 10% of lung cancer patients receiving immunotherapy. Current risk-assessment approaches rely on clinical factors and subjective imaging analysis that may not fully capture a patient’s underlying vulnerability.
The new model, called the Checkpoint-Inhibitor Pneumonitis Hazard EstimatoR, or CIPHER, analyzes routine chest CT scans taken before immunotherapy begins and looks for patterns within lung tissue associated with future pneumonitis risk.
Researchers trained CIPHER using more than 590,000 CT image slices from 2,500 patients with lung cancer.
Rather than training the model directly on known pneumonitis cases, the researchers first taught CIPHER to recognize patterns within lung tissue and then evaluated whether those patterns could distinguish patients who later developed pneumonitis after receiving immunotherapy.
The approach was then tested using pretreatment CT scans from 347 patients with non-small cell lung cancer treated at MD Anderson and validated against an independent external dataset.
CIPHER achieved an area under the curve of approximately 0.83 in both cohorts, outperforming conventional models based on clinical risk factors and radiomics-based approaches.
The model also maintained its performance despite differences across datasets in patient populations, CT scanner types, and imaging protocols, an important consideration for technology intended to operate across different hospitals and clinical environments.
Patients identified by CIPHER as high risk also tended to develop pneumonitis sooner after beginning immunotherapy.
Researchers said that result suggests the model may be detecting underlying signs of lung vulnerability rather than simply identifying characteristics correlated with future cases.
CIPHER’s predictions remained significant after researchers accounted for variables including age, smoking history, tumor histology, and prior thoracic radiation exposure.
The study was published in the Journal for ImmunoTherapy of Cancer and was led by Jia Wu, Ph.D., associate professor of Imaging Physics and Thoracic/Head and Neck Medical Oncology at MD Anderson, alongside co-senior authors Ajay Sheshadri, M.D., associate professor of Pulmonary Medicine, and Mehmet Altan, M.D., associate professor of Thoracic/Head and Neck Medical Oncology.
One potentially important aspect of the research is that CIPHER relies on imaging already routinely collected before treatment rather than requiring a new diagnostic procedure.
That could eventually make it possible to integrate the technology into existing oncology workflows if future prospective studies validate its performance.
An example visualization accompanying the research shows AI-generated highlighting of lung abnormalities identified by CIPHER on a chest CT scan, illustrating how the model can identify regions associated with elevated risk before treatment begins.
Researchers cautioned that additional work is required before the model can be incorporated into routine clinical care.
Future prospective studies will need to evaluate CIPHER across larger and more diverse patient populations and determine whether its risk predictions improve clinical outcomes when used to guide monitoring or treatment decisions.
The research team also plans to study whether CIPHER performs similarly across other cancer types treated with immunotherapy.
Additional research could combine imaging information with biomarkers or other patient data to further improve predictions.
Researchers also see the possibility of using similar AI approaches to identify patients at risk for other immunotherapy-related toxicities.
If validated, these systems could help clinicians determine which patients need closer monitoring, identify candidates for prevention studies, and better understand why certain patients experience severe treatment-related side effects.
The research was supported by the National Institutes of Health, Cancer Prevention and Research Institute of Texas, and institutional funding from MD Anderson.
KEY QUOTE:
“Pneumonitis remains one of the most challenging complications of immunotherapy because it can be difficult to predict before symptoms appear. Our model was able to identify signals associated with future risk using information that already exists in routine CT scans.
What makes this approach particularly interesting is that it was not designed to look for pneumonitis itself. Instead, the model learned patterns within lung tissue and identified subtle abnormalities associated with future risk. That suggests routine imaging may contain much more information about treatment toxicity than we previously recognized.”
Jia Wu, Ph.D., Associate Professor Of Imaging Physics And Thoracic/Head And Neck Medical Oncology At MD Anderson
