Performance of magnetic-resonance imaging radiomics in prediction of response after neoadjuvant chemotherapy in head and neck squamous cell carcinoma: A systematic review and meta-analysis

This is the authors' abstract. We don't add key points for this paper.

Objective: To evaluate performance of MRI-based radiomics for evaluating response to neoadjuvant chemotherapy (NACT) in head and neck squamous cell carcinoma (HNSCC) patients. Method: We performed a comprehensive search across four databases until July, 2025 to include studies evaluating diagnostic performance of MRI radiomics in response assessment. Methodological quality of studies was assessed using METhodological RadiomICs Score (METRICS). Using a random-effects bivariate model diagnostic values of area under the curve (AUC), sensitivity, and specificity were calculated. Meta-regression and subgroup analyses were used to explore the source of heterogeneity. Results: Twenty-one studies were included. The overall predictive performance of MRI-radiomics in the validation set was good, with a summary AUC of 0.84 (95% CI: 0.81-0.87) in differentiation of complete or partial response (CR/PR) and non-responders as those with stable or progressive disease (SD/PD). High heterogeneity was observed, but there was no evidence of publication bias (Deeks' test p = 0.20). Subgroup analysis resolved the heterogeneity and showed similar performance of model across different therapeutic regimens, study designs, study qualities, and MRI magnitudes of strength. For classifying only CR as responders and PR + SD + PD as non-responders, MRI-radiomics achieved an AUC of 0.83 (95% CI: 0.79-0.86) with low heterogeneity. Conclusion: MRI radiomics demonstrated high predictive value in evaluating NACT response for HNSCC patients, showing consistent performance across different clinical scenarios. However, the current evidence base is largely retrospective, geographically limited, and predominantly derived from nasopharyngeal cancer cohorts; therefore, broader external validation and prospective multicenter designs are required before routine clinical implementation.