Incremental diagnostic value of microstructural time-dependent diffusion MRI in differentiating PCNSL from glioblastoma over conventional MRI.

PURPOSE: To evaluate the incremental diagnostic value of microstructural parameters derived from time-dependent diffusion MRI (td-dMRI) over conventional MRI in differentiating primary central nervous system lymphomas (PCNSLs) from glioblastomas (GBMs).

MATERIALS AND

METHODS

This study included 63 patients with PCNSL and 98 with GBM who underwent preoperative conventional MRI and td-dMRI. Quantitative microstructural parameters, including intracellular volume fraction (fin), extracellular diffusivity (Dex), cell diameter (d) and cellularity, were derived using the imaging microstructural parameters using limited spectrally edited diffusion (IMPULSED) model. Three eXtreme Gradient Boosting (XGBoost) diagnostic models (conventional, td-dMRI, and combined) were constructed using selected features. Diagnostic performance and the incremental value of td-dMRI were evaluated using the AUC.

RESULTS

PCNSLs demonstrated significantly higher lesion multiplicity, fin, Dex, d, and cellularity, but lower ring-like enhancement, necrosis, T2-weighted imaging signal intensity ratio (T2-SIR), and minimum apparent diffusion coefficient (ADCmin) compared to GBMs (all P < 0.05). The combined model achieved a significantly higher AUC (0.994) than the conventional (0.957) and td-dMRI (0.972) models. Adding td-dMRI parameters provided a statistically significant incremental diagnostic benefit (ΔAUC = 0.031) over conventional MRI alone.

CONCLUSION

Integrating td-dMRI derived microstructural parameters provides significant incremental value to conventional MRI, markedly improving the preoperative differentiation of PCNSLs and GBMs.