Development and validation of a prediction model based on two-dimensional dose distribution maps fused with computed tomography images for noninvasive prediction of radiochemotherapy resistance in non-small cell lung cancer.

Background: There are individualized differences in the prognosis of radiochemotherapy for non-small cell lung cancer (NSCLC), and accurate prediction of prognosis is essential for individualized treatment. This study proposes to explore the potential of multiregional two-dimensional (2D) dosiomics combined with radiomics as a new imaging marker for prognostic risk stratification of NSCLC patients receiving radiochemotherapy.

Methods: In this study, 365 patients with histologically confirmed NSCLC, who had computed tomography (CT) scans before treatment, received standard radiochemotherapy, and had Karnofsky Performance Scale (KPS) scores ≥70 were included in three medical institutions, and 145 cases were excluded due to surgery, data accuracy, poor image quality, and the presence of other tumors. Finally, 220 patients were included in the study. Efficacy evaluation criteria for solid tumors are used to evaluate efficacy. Complete and partial remission indicate the radiochemotherapy-sensitive group, and disease stability and progression indicate the radiochemotherapy-resistant group. We combined all the data and then randomised them into a training cohort (154 cases) and a validation cohort (66 cases) in a 7:3 ratio. Radiomics and dosiomics features were extracted for gross tumor volume (GTV), GTV-heat, and 50 Gy-heat and screened. 2D dosiomics model (DM_GTV and DM_50Gy), radiomics model (RM_GTV), 2D radiomics-dosiomics model (RDM), and combined models were constructed, and the predictive performances for radiochemotherapy resistance were compared. Subsequently, the predictive performance of various models for radiochemotherapy resistance was compared by receiver operating characteristic (ROC) curves and calculating accuracy, sensitivity and specificity. The multi-omics and clinical models were integrated for patient risk stratification.

Results: DM_50Gy had better predictive performance than RM_GTV and DM_GTV, with the area under the curve (AUC) of the ROC in the training and validation cohorts for DM_50Gy were 0.764 [95% confidence interval (CI): 0.687-0.841] and 0.729 (95% CI: 0.568-0.889). And the RDM performed significantly better than the single radiomics and dosiomics models, with AUC of 0.836 (95% CI: 0.773-0.899) and 0.748 (95% CI: 0.617-0.879), respectively. Hemoglobin level and T stage were independent predictors in the clinical model. The combined model containing independent predictors further improved the predictive performance in both the training and validation cohorts, with AUC of 0.844 (95% CI: 0.781-0.907) and 0.753 (95% CI: 0.618-0.887). Grouping of patients according to the critical value of the combined model revealed significant differences in progression-free survival (PFS) and overall survival (OS) between the high-risk and low-risk groups (P<0.05).

Conclusions: Compared to the traditional radiomics model, the 2D dosiomics model demonstrates superior predictive performance. The combined model based on clinical data, radiomics, and dosiomics has improved the prediction of radiochemotherapy resistance in NSCLC and effectively performed survival stratification. Through precise risk assessment, doctors can better understand which patients may develop resistance to treatment and optimize treatment plans accordingly.