Histogram analysis of diffusion kurtosis imaging for T restaging and predicting complete response of locally advanced...

Accurate preoperative T restaging and prediction of pathological complete response (pCR) are pivotal for organ preservation strategies in locally advanced rectal cancer (LARC). However, conventional magnetic resonance imaging (MRI) shows limited accuracy after neoadjuvant therapy (NAT) because therapy-induced necrosis, edema, inflammation, and fibrosis may obscure residual tumor. Neoadjuvant immunotherapy has shown considerable promise in LARC but introduces more complex regression patterns, making post-treatment assessment even more challenging. This study aimed to evaluate the diagnostic performance of diffusion kurtosis imaging (DKI) for T restaging and pCR prediction in LARC following neoadjuvant immunotherapy.

Between July 2021 and March 2023, patients with LARC who underwent neoadjuvant immunotherapy followed by surgery were enrolled prospectively. All participants underwent DKI and conventional MRI (T2WI and DWI) at baseline and following immunotherapy (before surgery). Based on postoperative pathology, patients were categorized into ypT0-1 and ypT2-4 groups, where ypT denotes the post-NAT pathological T stage, and into pCR (ypT0N0) and non-pCR groups. Histogram parameters of DKI-derived diffusivity (D) and kurtosis (K) were extracted. Two combined models, integrating DKI histogram parameters with conventional MRI, were constructed. The diagnostic performance of these models was assessed by receiver operating characteristic (ROC) analysis.

Sixty-four patients [median age, 58 years (interquartile range, 52-62 years); 42 men] were included. Several histogram features derived from DKI parameters D and K differed significantly between groups (all P<0.05). For T restaging, the kurtosis of K before surgery (Post-K_Kurtosis) and the change ratio of the median of D (∆% D_Median) were incorporated into the combined model A. Model A achieved an AUC of 0.857 [95% confidence interval (CI): 0.762, 0.953]. For predicting pCR, the change ratio of the mean of D (∆% D_Mean) and the kurtosis of K (∆% K_Kurtosis) were included in the combined model B. Model B achieved an AUC of 0.904 (95% CI: 0.825, 0.982).

Histogram analysis of DKI showed good diagnostic performance for T restaging and predicting pCR in LARC following neoadjuvant immunotherapy.