Continuous-time random-walk diffusion model for predicting tumor consistency and extent of resection in patients with pituitary adenomas

To investigate the utility of continuous-time random-walk (CTRW) diffusion model parameters in predicting tumor consistency and the extent of resection (EOR) in patients with pituitary adenomas (PAs). This prospective study included 53 patients with PAs who underwent preoperative multi-b value diffusion MRI followed by endoscopic transsphenoidal surgery. CTRW-derived parameters, including the anomalous diffusion coefficient (D m ), temporal heterogeneity parameter (α), and spatial heterogeneity parameter (β), were calculated from whole-tumor regions of interest. Conventional apparent diffusion coefficient (ADC) values were obtained for comparison. Tumor consistency was classified intraoperatively as soft or hard. Predictors of EOR > 95% were evaluated using univariable and multivariable logistic regression analyses. Diagnostic performance was assessed using receiver operating characteristic (ROC) analysis. Hard tumors demonstrated significantly lower α (0.472 ± 0.170 vs. 0.649 ± 0.141, P < 0.001) and β values (0.752 ± 0.060 vs. 0.806 ± 0.093, P = 0.031) compared with soft tumors. D m and ADC values were also lower in hard tumors ( P = 0.039 and P = 0.029, respectively). Collagen content showed a significant negative correlation with all diffusion parameters, with α exhibiting the strongest association ( r = -0.492, P < 0.001). Patients achieving EOR > 95% showed significantly higher α and β values and lower Knosp grades (all P < 0.01). In multivariable analysis, α (odds ratio [OR], 1.10 per 0.1 increase; P = 0.017), β (OR, 1.20 per 0.1 increase; P = 0.012), and Knosp grade (OR, 0.29; P = 0.011) were independent predictors of EOR > 95%. A combined model incorporating Knosp grade, α, and β achieved an AUC of 0.941, significantly outperforming Knosp grade alone ( P = 0.027). CTRW-based diffusion analysis provides clinically relevant information related to tumor consistency and resectability in pituitary macroadenomas.