Radiation Pneumonitis Risk Assessment Using Fractal Analyses in NSCLC Patients Treated with Curative-Intent Radiotherapy

<b>Objectives:</b> This study evaluated the utility of complex morphometric analyses for predicting radiation pneumonitis (RP) and proposed a quantitative prognostic framework for patients with non-small cell lung cancer (NSCLC) undergoing curative-intent radiotherapy (RT). Imaging biomarkers, including box-counting fractal dimension (BoxFD), lacunarity, and minimum spanning tree fractal dimension (MSTFD), were assessed for their prognostic significance. <b>Materials and Methods:</b> We retrospectively analyzed 166 NSCLC patients who received curative-intent RT and had both pre-treatment and follow-up chest CT scans. Among them, 85 received RT alone and 81 underwent concurrent chemoradiotherapy (CCRT). Fractal features were measured to build a Random Forest model (RFM) predicting RP of grade ≥ 2, and the most important features were used to construct a decision tree model. <b>Results:</b> RP of grade ≥ 2 occurred in 19 patients (22.3%) in the RT alone group and 44 patients (54.3%) in the CCRT group. Lacunarity increased significantly post-RT in both groups, while BoxFD and MSTFD showed no significant changes. In the RFM, pre-RT MSTFD and lung dose parameters (V10 in RT alone; V5-V20 in CCRT) were identified as key predictors. Decision tree models based on these features achieved high predictive performance, with AUROC of 0.83 and 0.85, and F1 scores of 0.92 and 0.76 for RT alone and CCRT groups, respectively. <b>Conclusions:</b> Fractal imaging biomarkers demonstrated promising prognostic value for predicting grade ≥ 2 RP in NSCLC patients. The proposed decision tree model may serve as a practical tool for early identification of high-risk patients, facilitating personalized treatment strategies and informing future research.