The research team, led by Eryuan Gao, Xiaoyue Ma and colleagues working with Jingliang Cheng, enrolled 55 patients who had undergone multi-b-value diffusion-weighted imaging and whose tumors were confirmed by pathology: 30 with atypical high-grade glioma and 25 with primary central nervous system lymphoma. Because both tumors presented atypically, the cohort represents the diagnostic gray zone where MRI-based differentiation matters most. The investigators manually outlined three regions of interest in each patient, covering the tumor parenchyma, the surrounding peritumoral edema and the necrotic core. From the parenchyma and edema, they extracted two families of diffusion parameters. NODDI contributed the intracellular volume fraction, the isotropic volume fraction and the orientation dispersion index, while DTI contributed mean diffusivity, axial diffusivity, radial diffusivity and fractional anisotropy.

The comparison was designed as a direct head-to-head test of the two techniques on the same patients and the same imaging data. Parameter values from the two tumor groups were compared using independent t-tests, Welch’s t-tests or the Mann-Whitney U test according to the distributional properties of each variable, and diagnostic efficiencies were compared between NODDI and DTI within the tumor parenchyma and the peritumoral edema using DeLong’s test. A significance threshold of P less than 0.05 was applied throughout, and the optimal cutoff points for the diagnostic parameters were set at the maximum Youden index, the threshold that best balances sensitivity and specificity.

The results in the tumor parenchyma were decisive. Among all NODDI parameters, the orientation dispersion index achieved the highest area under the curve, an AUC of 0.961. Among DTI parameters, axial diffusivity performed best, but its AUC of 0.871 fell significantly short of the ODI figure when the two were compared with DeLong’s test. The advantage of NODDI over DTI in the tumor core was therefore not a marginal statistical nicety but a demonstrable gain in diagnostic power, driven by a parameter that quantifies how strongly cellular orientations disperse within the tumor tissue itself.

The picture in the peritumoral edema was more nuanced. There, the isotropic volume fraction was the strongest NODDI parameter with an AUC of 0.827, while axial diffusivity again led the DTI parameters at 0.756. This time, DeLong’s test found no significant difference between the two AUCs, indicating that NODDI’s edge over DTI, so pronounced in the tumor parenchyma, did not reach statistical significance in the edematous rim. The authors concluded that the tumor parenchyma is the superior region of interest option for this differential diagnosis, a practical point for radiologists designing imaging protocols, since where parameters are measured matters as much as which parameters are measured.

The necrotic core was outlined as a third region of interest in the study protocol, but the reported between-group comparisons and diagnostic efficiency analyses focused on the tumor parenchyma and the peritumoral edema, where the full parameter sets were extracted and statistically tested. This scope means the headline findings apply specifically to the two regions where the techniques were formally compared.

The study’s highlights state that NODDI can distinguish high-grade glioma from primary central nervous system lymphoma, that it exhibits excellent diagnostic performance in both the tumor parenchyma and the peritumoral edema, and that the orientation dispersion index outperformed every other parameter produced by either technique in the tumor core. Within the NODDI family, the intracellular volume fraction and the isotropic volume fraction provided additional quantitative measures, but neither matched the ODI figure in the parenchyma, and the same pattern held for the DTI family, where mean diffusivity, radial diffusivity and fractional anisotropy trailed axial diffusivity.

Even with the modest size of the 55-patient cohort in mind, the study offers a clear, quantified demonstration that model choice in diffusion MRI changes diagnostic outcomes. NODDI extracted a signature from multi-b-value diffusion-weighted imaging that DTI could not match in the tumor parenchyma, and that signature translated into a statistically significant improvement in separating atypical high-grade glioma from primary central nervous system lymphoma. The orientation dispersion index, with an AUC approaching 0.96 in this cohort, stands out as the single strongest parameter reported, and the confirmation that the tumor parenchyma outperforms the peritumoral edema as a measurement region gives radiologists a concrete protocol recommendation grounded in the data.

The findings arrive as an open-access, peer-reviewed contribution to a growing literature on quantitative diffusion MRI in neuro-oncology. For the patients represented in this study, all of whom presented with tumors whose imaging appearance defied easy classification, the reported performance of NODDI parameters offers a sharper MRI-based tool for a differential diagnosis that currently rests on pathological confirmation. Whether the ODI thresholds identified here generalize to larger and more diverse cohorts remains a question for future validation, but within this study’s scope the conclusion is unambiguous: in the tumor parenchyma, NODDI’s orientation dispersion index separated atypical high-grade glioma from primary central nervous system lymphoma significantly better than the best parameter DTI could provide.

Subject of Research: Distinguishing atypical high-grade glioma from primary central nervous system lymphoma with diffusion MRI: neurite orientation dispersion and density imaging versus diffusion tensor imaging

Article Title: Distinguishing atypical high-grade glioma from primary central nervous system lymphoma with diffusion MRI: neurite orientation dispersion and density imaging versus diffusion tensor imaging

Article References: Gao, E., Ma, X., Wang, P., Bai, J., He, M., Zhao, K., Zhang, H., Wang, M., Song, Y., Wang, W., Yang, G., Zhao, G., & Cheng, J. (2026). Distinguishing atypical high-grade glioma from primary central nervous system lymphoma with diffusion MRI: neurite orientation dispersion and density imaging versus diffusion tensor imaging. BMC Cancer. https://doi.org/10.1186/s12885-026-16720-w

Image Credits: AI Generated

DOI: 10.1186/s12885-026-16720-w

Keywords: Distinguishing, atypical, high-grade, glioma, primary, central, nervous, system, lymphoma, diffusion, neurite, orientation