Can Radiomics Predict Pathologic Complete Response After Neoadjuvant Chemoradiotherapy for Rectal Cancer? A Systematic Review and Meta-Analysis of Diagnostic-Accuracy Studies

Abstract

Background: The rectal cancer treatment paradigm is rapidly changing with the advent of total neoadjuvant therapy and non-operative management approaches in responders. A good clinical response to neoadjuvant treatment documented by magnetic resonance imaging, endoscopy and clinical examination corresponds, to a large extent, to a pathologic complete response, as assessed in surgical specimens. Methods: We undertook a systematic review and meta-analysis on the MRI-based omics approach to predicting pathologic complete responses. Results: A total of 29 studies with relevant data available reporting on a total of 4486 patients were eligible for meta-analysis. The calculated values for the area under the curve in receiver operator curves of diagnostic accuracy for radiomics-only and radiomics-combined-with-clinical-data models were 0.80 and 0.88, respectively, for studies incorporating baseline imaging data only. The value for studies using delta radiomic data was 0.86, and those for studies using data from the post-neoadjuvant setting were 0.75 and 0.83, respectively, for the radiomics-only and radiomics-combined-with-clinical-data models. Conclusions: Radiomics-based prediction models for pathologic complete response assessment might further enable individualized treatment decisions to be made in patients with rectal cancer.