The objective of the study was to assess the correlation between radiomics features extracted from cone-beam CT with the treatment response of liver tumors treated with stereotactic body radiation.

The planning CT and cone-beam CT were prospectively collected for 76 patients with liver cancer who received five fractions of stereotactic body radiation therapy. Pearson correlation test was used to identify interchangeable radiomics features between cone-beam- and planning CT from a total of 547 extracted radiomics features. Principal components analysis was used for cone-beam CT delta radiomics to characterize therapy-induced tumor change. The Mann-Whitney U-test was used to identify features with correlation to treatment response: local efficacy versus local non-efficacy; complete versus partial response in both raw and principal components analysis-based cone-beam CT radiomics features. The area under the receiver operating characteristic curve was used for assessing feature performance.

A total of 345 cone-beam CT radiomics features were interchangeable with planning CT. The mean value of LHH_GLSZM_LGZE showed an ascending trend in five fractions during the course of treatment. Among these features, PCA3_LHH_Histogram_Energy showed the best performance in predicting local efficacy, with an AUC of 0.879 (0.744–1.00, 95 % CI). For identifying complete response from partial response, CBCT2_LHL_GLSZM_GLV showed the best value, with the highest AUC of 0.884 (0.773–1.00, 95 % CI).

Radiomics features extracted from cone-beam CT images have the potential for assessing the response to treatment in advance and can serve as an early biomarker for liver tumor stereotactic body radiation therapy.