Computed tomography-based delta-radiomics enabling early prediction of short-term responses to concurrent chemoradiotherapy for patients with non-small cell lung cancer

Abstract

Objective

To explore the potential of computed tomography (CT)-based delta-radiomics in predicting early short-term responses to concurrent chemoradiotherapy for patients with non-small cell lung cancer (NSCLC), in order to determine the optimal time point for the prediction.

Methods

A total of 20 patients with pathologically confirmed NSCLC were prospectively enrolled in this study, who did not receive surgical treatment between February 2021 and February 2022. For each case, a total of 1,210 radiomic features (RFs) were extracted from both planning CT (pCT) images along with each of the subsequent three weeks of CT images. Effective ΔRFs were selected using intra-class correlation coefficient (ICC) analysis, Pearson's correlation, ANOVA test (or Mann-Whitney U-test), and univariate logistic regression. The area under the curve (AUC) of the receiver operating characteristic (ROC) curve was used to evaluate the potential to predict short-term responses of different time points.

Results

Among the 1,210 ΔRFs for 1–3 weeks, 121 common features were retained after processing using ICC analysis and Pearson's correlation. These retained features included 54 and 58 of all time points that differed significantly between the response and non-response groups for the first and third months, respectively (P ​< ​0.05). After univariate logistic regression, 11 and 44 features remained for the first and third months, respectively. Finally, eight ΔRFs (P ​< ​0.05, AUC ​= ​0.77–0.91) that can discriminate short-term responses in both at 1 and 3 months with statistical accuracy were identified.

Conclusion

CT-based delta-radiomics has the potential to provide reasonable biomarkers of short-term responses to concurrent chemoradiotherapy for NSCLC patients, and it can help improve clinical decisions for early treatment adaptation.