Habitat-based radiomics analysis for evaluating immediate response in colorectal cancer lung metastases treated by radiofrequency ablation

Abstract

Abstract Purpose To create radiomics signatures based on habitat to assess the instant response in lung metastases of colorectal cancer (CRC) after radiofrequency ablation (RFA). Methods Between August 2016 and June 2019, we retrospectively included 515 lung metastases in 233 CRC patients who received RFA (412 in the training group and 103 in the test group). Multivariable analysis was performed to identify independent risk factors for developing the clinical model. Tumor and ablation regions of interest (ROI) were split into three spatial habitats through K-means clustering and dilated with 5 mm and 10 mm thicknesses. Radiomics signatures of intratumor, peritumor, and habitat were developed using the features extracted from intraoperative CT data. The performance of these signatures was primarily evaluated using the area under the receiver operating characteristics curve (AUC) via the DeLong test, calibration curves through the Hosmer-Lemeshow test, and decision curve analysis. Results A total of 412 out of 515 metastases (80%) achieved complete response. Four clinical variables (cancer antigen 19–9, simultaneous systemic treatment, site of lung metastases, and electrode type) were utilized to construct the clinical model. The Habitat signature was combined with the Peri-5 signature, which achieved a higher AUC than the Peri-10 signature in the test set (0.825 vs. 0.816). The Habitat+Peri-5 signature notably surpassed the clinical and intratumor radiomics signatures (AUC: 0.870 in the test set; both, p < 0.05), displaying improved calibration and clinical practicality. Conclusions The habitat-based radiomics signature can offer precise predictions and valuable assistance to physicians in developing personalized treatment strategies.

Keywords

• Colorectal neoplasms
• Lung metastasis
• Radiofrequency ablation
• Radiomics
• Habitat imaging
• Peritumoral micro-environment