Predicting Stereotactic Body Radiation Therapy Response Using an AI-Based Tumor Vessel Biomarker

Area of Science:

• Radiology
• Oncology
• Artificial Intelligence

Background:

• Abnormal tumor vasculature in non-small cell lung cancer (NSCLC) leads to hypoxia, treatment resistance, and poor prognosis.
• Radiation therapy can alter tumor vessels, but outcomes vary due to vascular heterogeneity.
• A noninvasive method to quantify vascular abnormality is crucial for predicting treatment response.

Purpose of the Study:

• To develop and validate a deep learning-based imaging biomarker, the Vessel Risk Score (VRS).
• To quantify tumor vascular abnormality from contrast-enhanced CT scans in NSCLC patients.
• To assess VRS's ability to predict radiation therapy response and prognosis.

Main Methods:

• A deep learning model was trained on multi-institutional data from 126 NSCLC patients treated with hypofractionated radiotherapy.
• The model learned vascular morphology patterns to quantify heterogeneity.
• VRS generalizability was evaluated in an external cohort of 128 early-stage NSCLC patients treated with SBRT.

Main Results:

• VRS significantly outperformed vessel density in predicting SBRT response.
• Lower VRS was associated with treatment response (0.494 vs. 0.578).
• High VRS correlated with shorter progression-free survival (PFS) and was the sole significant predictor in multivariate analysis.

Conclusions:

• The AI-derived VRS is a noninvasive, reproducible measure of tumor vascular abnormality.
• VRS offers improved prediction of radiation therapy response and prognosis in NSCLC compared to vessel density.
• This AI approach holds potential for prognostic assessment in other cancers where vascular morphology is key.