Brain Scan Water Diffusion Measure Falls as Glioma Grade Rises in 33-Patient Study

In 33 patients with histopathologically confirmed gliomas, tumor fractional anisotropy derived from diffusion tensor imaging showed a significant inverse correlation with WHO tumor grade, while mean diffusivity, the companion diffusion measure, failed to show any significant grade-related association. The work, published in the Egyptian Journal of Radiology and Nuclear Medicine, was conducted in the Department of Radiodiagnosis at Dr. Ram Manohar Lohia Institute of Medical Sciences in Lucknow.

Glioma grading has traditionally depended on histopathological assessment of biopsy or surgical resection specimens, which means the definitive answer about tumor aggressiveness usually arrives only after tissue has been obtained. Conventional magnetic resonance imaging provides important but sometimes nonspecific information about tumor characteristics, such as enhancement, necrosis and surrounding edema. Diffusion tensor imaging extends standard diffusion-weighted imaging by quantifying the directionality and mobility of water molecules within tissue. From this data, two parameters are commonly derived: fractional anisotropy, or FA, which reflects how strongly water diffusion is constrained along organized white-matter tracts, and mean diffusivity, or MD, which captures overall water mobility regardless of direction.

The rationale for the study rests on tumor biology. As gliomas become more malignant, they tend to disrupt the ordered architecture of brain tissue through increasing cellularity, infiltration of white matter and growing intratumoral heterogeneity. That disruption should, in principle, reduce the directional organization of water diffusion, lowering FA, while effects on overall diffusivity are harder to predict because competing pathological processes pull MD in opposite directions. The authors set out to evaluate the association between DTI parameters and histopathological glioma grade, to assess relationships between DTI parameters and Ki-67 expression, a marker of cell proliferation, and to explore diagnostic performance in separating Grade 1–2 from Grade 3–4 tumors.

The cohort comprised 33 patients, 29 adults and 4 pediatric patients, ranging in age from 8 to 72 years, with a mean age of 43.97 years, and included 23 males and 10 females. All had histopathologically confirmed glioma and underwent preoperative MRI with DTI before any surgical intervention or pathological diagnosis. Patients with previous intracranial surgery and those whose final diagnosis was not glioma were excluded. The final grade distribution included 2 Grade 1, 13 Grade 2, 11 Grade 3 and 7 Grade 4 tumors, meaning 15 patients had lower-grade tumors and 18 had high-grade disease. Histopathological and available molecular diagnoses included 12 astrocytomas, IDH-mutant; 9 oligodendrogliomas, IDH-mutant and 1p/19q-codeleted; 6 glioblastomas, IDH-wildtype; 3 pleomorphic xanthoastrocytomas; 2 pilocytic astrocytomas; and 1 diffuse midline glioma, H3 K27-altered. Twenty-one patients were IDH-mutant and 12 were IDH-wildtype.

Imaging was performed on GE Sigma scanners at either 1.5 Tesla, used for 10 patients, or 3 Tesla, used for 23 patients, with a 30-channel head coil. DTI acquisition used diffusion gradients in 15 to 32 non-collinear directions with b-values of 0 and 1000 s/mm2, and FA and MD maps were generated on a GE Advantage Workstation. A single radiologist manually placed one region of interest, sized 4 to 10 mm2, in each of three predefined regions: the tumor core, the peritumoral region and contralateral normal-appearing brain parenchyma. ROIs were positioned within representative solid tumor tissue while avoiding necrosis, hemorrhage and cystic change, and for non-enhancing tumors placement was guided by abnormal regions on T2-weighted and FLAIR images rather than enhancement alone.

The headline result was a clear, graded decline in tumor FA. Mean tumor FA was 0.245 in Grade 1, 0.176 in Grade 2, 0.143 in Grade 3 and 0.077 in Grade 4, and a Kruskal-Wallis test found a statistically significant difference across grades with p equal to 0.001. Spearman correlation confirmed a significant inverse relationship between tumor FA and grade, with a coefficient of −0.694 and p below 0.001. Post-hoc Dunn-Bonferroni comparisons showed significant differences between Grade 2 and Grade 4 and between Grade 3 and Grade 4, while other pairwise contrasts did not reach significance. Tumor MD, by contrast, showed no significant difference across grades and no significant correlation with grade, with mean values of 1.400, 1.325, 1.309 and 1.490 across Grades 1 to 4. Notably, Grade 4 tumors had a higher mean MD than Grade 3 tumors, which the authors attribute to the necrotic, cystic and edematous components common in high-grade tumors rather than to any systematic increase of diffusivity with grade.

The peritumoral region, the tissue just beyond the visible tumor margin, told an even stronger story. Peritumoral FA showed a significant inverse correlation with grade, with a Spearman coefficient of −0.793, and peritumoral MD showed a positive correlation of 0.435. This region may contain infiltrating tumor cells and microstructural abnormalities extending past the visually defined margin, and previous DTI studies have documented altered diffusion metrics in tissue surrounding gliomas, with contributions from both edema and tumor infiltration. The authors caution, however, that this peritumoral analysis should be considered secondary and exploratory given the small sample size and the manual, single-reader ROI methodology.

The proliferation marker Ki-67 added biological context. Tumor FA correlated significantly and inversely with Ki-67 expression, with a coefficient of −0.632 and p below 0.001, while tumor MD showed no significant relationship with Ki-67. This association supports a link between reduced anisotropy and tumor proliferative characteristics within this cohort, though the authors emphasize it does not establish FA as a surrogate for molecular classification, because molecular and immunohistochemical assessment was not uniformly available for all contemporary WHO CNS5 parameters. On exploratory receiver operating characteristic analysis, tumor FA differentiated Grade 1–2 from Grade 3–4 tumors with an area under the curve of 0.833; at a cutoff of 0.120 or below, sensitivity was 61.1 percent and specificity 93.3 percent. Peritumoral FA performed better within this dataset, with an AUC of 0.941, sensitivity of 94.4 percent and specificity of 86.7 percent at a cutoff of 0.240, while tumor MD achieved an AUC of only 0.611.

Sensitivity analyses strengthened the central finding. The inverse FA-grade association persisted among patients scanned at 3 Tesla, among those scanned at 1.5 Tesla, among adults only, and after excluding the two Grade 1 tumors, with correlation coefficients ranging from −0.640 to −0.800 across these subsets. Nevertheless, the study’s limitations are substantial and the authors enumerate them candidly. The sample was small, particularly for Grade 1 tumors with only two cases. Adults and children were pooled. Two field strengths and varying numbers of diffusion directions were used. Regions of interest were placed manually by a single radiologist without independent repeat measurements, so interobserver reproducibility could not be assessed, and single-region sampling may not capture marked intratumoral heterogeneity. Complete molecular profiling required for integrated WHO CNS5 classification was not uniformly available.

Most importantly, the ROC-derived cutoffs were calculated from the same cohort using the Youden index and were never subjected to external or independent validation, so the reported sensitivity and specificity figures may be inflated by optimism bias and cannot be treated as clinically validated thresholds. The authors frame their work as a prospective, single-institution validation of an established relationship between diffusion metrics and glioma grade rather than the identification of a novel imaging biomarker, and they call for larger multicenter studies with standardized DTI acquisition, reproducible or automated ROI methods, comprehensive molecular classification and independent validation cohorts before any FA cutoff could be applied in clinical practice. Within those boundaries, the study offers prospective evidence that fractional anisotropy, measured non-invasively before surgery, declines in step with glioma aggressiveness and with proliferative activity, while mean diffusivity alone remains a less consistent reflection of grade.

Subject of Research: Diffusion tensor imaging parameters and their correlation with histopathological glioma grade

Article Title: Prospective observational study evaluating correlation between diffusion tensor imaging findings with glioma grade

Article References: Navik, K. K., Kumar, T., Raj, G., Chauhan, A., Narayan, S., Singh, N., Singh, D. K., & Kulshreshtha, D. (2026). Prospective observational study evaluating correlation between diffusion tensor imaging findings with glioma grade. Egyptian Journal of Radiology and Nuclear Medicine, 57(1), Article 212. https://doi.org/10.1186/s43055-026-01879-w

Image Credits: AI Generated

DOI: 10.1186/s43055-026-01879-w

Keywords: glioma, diffusion tensor imaging, fractional anisotropy, mean diffusivity, tumor grading, MRI, Ki-67, ROC analysis, brain tumor, neuroradiology, WHO CNS5, prospective study