Comparison of tumor microstructure derived NODDI and DTI metrics to histopathology in different grades of brain tumor
Prasanna Parvathaneni1, Qiuting Wen2, Joanna J Phillips 3,4, Soonmee Cha1,4, Susan M Chang4, Sarah J Nelson1,5, and Janine M Lupo1

1Department of Radiology and Biomedical Imaging, University of California San Francisco (UCSF), San Fransisco, CA, United States, 2Indiana University, Indianapolis, IN, United States, 3Department of Pathology, University of California San Francisco (UCSF), San Fransisco, CA, United States, 4Department of Neurological Surgery, University of California San Francisco (UCSF), San Fransisco, CA, United States, 5Department of Bioengineering and Therapeutic Sciences, University of California San Francisco (UCSF), San Fransisco, CA, United States


New non-Gaussian measurement techniques like neurite orientation dispersion and density imaging (NODDI) that allow quantification of specific tissue microstructure features can provide meaningful biophysical indices to overcome the low specificity of DTI. In this study we applied three compartment model based NODDI and DTI to histopathology and explored the correlation with tumor cellularity between non-enhancing and contrast enhancing lesions. Unlike in normal brain where Vin represents the neurite density, it was positively correlated with tumor grade and tumor score in tissue samples from the tumor region, indicating the association of Vin with tumor cellularity. Although NODDI is not directly built on tumor, it brings parameters that were sensitive to tumor cellularity, which may complement the conventional DTI model and adds specificity. Thus NODDI, when combined with DTI, could add value in understanding the heterogeneity of tissue microstructure in brain tumors.


Diffusion tensor imaging (DTI) is a widely used technique in the diagnosis and monitoring of response to therapy in patients with brain tumors1-5. It has high sensitivity to underlying tissue structure but low specificity6. Recent work has shown the application of neurite orientation dispersion and density imaging (NODDI) to various neurological disorders including brain tumors7-11 to overcome the low specificity of DTI.


In this study, we have validated the association of NODDI and DTI with histopathology features in different tumor grades, as it will be highly valuable to the clinical community to confirm the diagnostic potential of these techniques7.


The association of diffusion parameters and tumor cellularity is explored within non-enhancing (NE) and contrast enhancing (CE) regions.

Patient Population: A total of 92 tissue samples from 34 patients with brain tumors (Figure 1), each with 1-4 tissue samples, were analyzed. Of these samples, 16 were from CE lesion (CEL), 68 were from NE lesion (NEL; T2 lesion - CEL), and 8 were at the border.

Histopathology parameters included tumor score (0 = neuropil without tumor, 1=infiltrating margin, 2=infiltrating tumor cells, 3=high percentage of tumor cells), angiogenesis (0=absent, 1=circumferential or simple, 2=complex), proliferation (% of all cells positive for MIB-1) and tumor grade.

Data Acquisition: Two-shell diffusion data including 24 gradient directions with b=1000 s/mm2 and 55 gradient directions with b=2000 s/mm2 were acquired with a standard SE-EPI sequence at a nominal voxel size of 2×2×2 mm3. Standard T1-weighted pre- and post-contrast and T2 FLAIR anatomical images were also acquired for region of interest (ROI) analysis.

Data Processing: The diffusion data were eddy-current corrected and DTI models were fit to the combined shell data with weighted least squares tensor fitting using FSL12. NODDI parameters were estimated as suggested in the original NODDI model13. Rigid alignment was performed between T2 FLAIR images and b0 images and the transformation was applied to the results for analysis of ROIs. The latter included the T2 hyper-intensity lesion (T2L) on T2 FLAIR images, the contrast-enhancing lesion (CEL) on T1 post-gad images and normal appearing white matter (NAWM) as segmented from T1-weighted pre-gad images. Image guided tissue targets were marked on the BrainLab surgical navigation system and the center coordinates of acquired tissue samples were used to define spherical ROIs of 5 mm diameter. Median intensity values of apparent diffusion coefficient (ADC), fractional anisotropy (FA), and radial diffusivity (RD) from DTI and intra neurite volume fraction (Vin), extra cellular volume fraction (Vec) and orientation dispersion index (ODI) from NODDI were analyzed within each ROI using automated software developed in our research group14.

Statistical analysis: Significant differences between tumor scores and tissue types were assessed using a Mann Whitney test. Categorical outcome variables were tested for association with each imaging parameter using ordinal logistic regression and continuous outcome variables were tested using Pearson correlation. P-values ≤ 0.05 were considered statistically significant in this exploratory analysis.

Results and Discussion

Median values of DTI and NODDI parameters within each ROI for different grades are shown in Figure 2. In all tumor grades, Vin was reduced compared to NAWM, indicating lower neurite density. Within grade 4 tumors, a higher Vin was observed in both the CEL and CE tissue samples compared to the NEL and NE tissue samples, consistent with the results found by Wen et al12.

While increasing and decreasing trends were observed for Vin and ADC with tumor score respectively (Figure 3), only a statistically significant difference in Vin (p<0.02) was found between CE tissue samples with a tumor score of 2 vs 3.

Across all grades, Vin values showed a positive correlation with tumor score (p<0.002) [Table 1], indicating that Vin could be a measure of tumor cellularity. Vec was negatively associated with tumor score (p<0.002) as was ADC (p<0.04) similar to previous studies15. Although FA and ODI did not correlate with tumor score, ODI was elevated in CE compared to NE tissue (Figure 2d). This could suggest less coherently organized structures16 with increased cellularity in CE tissue. Only FA showed a positive correlation with angiogenesis (p<0.05). All of the imaging parameters were associated with tumor grade.


Both ADC and NODDI parameters Vin and Vec were significantly correlated with brain tumor histopathology. Although NODDI is not directly built on tumor, it brings parameters that were sensitive to tumor cellularity, which may complement the conventional DTI model and adds specificity. In particular, Vin and ADC have distinct variations within CE and NE regions that when combined can offer additional insight into the heterogeneity of tissue microstructure among brain tumors.


This work was supported by the SPORE and P01 grants: P50-CA97257 and P01CA118816.


1. da Cruz, L.C.H., I.G. Vieira, and R.C. Domingues, Diffusion MR Imaging: An Important Tool in the Assessment of Brain Tumors. Neuroimaging Clinics of North America, 2011. 21(1): p. 27-+. 2. Ferda, J., et al., Diffusion tensor magnetic resonance imaging of glial brain tumors. European Journal of Radiology, 2010. 74(3): p. 428-436. 3. Kinoshita, M., et al., Fractional anisotropy and tumor cell density of the tumor core show positive correlation in diffusion tensor magnetic resonance imaging of malignant brain tumors. Neuroimage, 2008. 43(1): p. 29-35. 4. Kleiser, R., et al., Brain Tumors: Clinical Applications of Functional Magnetic Resonance Imaging and Diffusion Tensor Imaging. Tumors of the Central Nervous System: Brain Tumors (Pt 2), 2012. 4: p. 175-197. 5. Kralik, S.F., et al., Diffusion Imaging for Tumor Grading of Supratentorial Brain Tumors in the First Year of Life. American Journal of Neuroradiology, 2014. 35(4): p. 815-823. 6. Alexander, A.L., et al., Diffusion tensor imaging of the brain. Neurotherapeutics, 2007. 4(3): p. 316-329. 7. Colgan, N., et al., Application of neurite orientation dispersion and density imaging (NODDI) to a tau pathology model of Alzheimer's disease. Neuroimage, 2015. 8. Grussu, F., et al., Neurite orientation dispersion and density imaging of the healthy cervical spinal cord in vivo. Neuroimage, 2015. 111: p. 590-601. 9. Magnollay, L., et al., Application of neurite orientation dispersion and density imaging (NODDI) to relapsing remitting multiple sclerosis (RRMS). Multiple Sclerosis Journal, 2013. 19(11): p. 172-172. 10. Szczepankiewicz, F., et al., Quantification of microscopic diffusion anisotropy disentangles effects of orientation dispersion from microstructure: Applications in healthy volunteers and in brain tumors. Neuroimage, 2015. 104: p. 241-252. 11. Wen, Q., et al., Clinically feasible NODDI characterization of glioma using multiband EPI at 7 T. Neuroimage Clin, 2015. 9: p. 291-9. 12. Smith, S.M., et al., Advances in functional and structural MR image analysis and implementation as FSL. Neuroimage, 2004. 23: p. S208-S219. 13. Zhang, H., et al., NODDI: Practical in vivo neurite orientation dispersion and density imaging of the human brain. Neuroimage, 2012. 61(4): p. 1000-1016. 14. Barajas, R.F., et al., Regional variation in histopathologic features of tumor specimens from treatment-naive glioblastoma correlates with anatomic and physiologic MR Imaging. Neuro-Oncology, 2012. 14(7): p. 942-954. 15. Cha, S., Update on brain tumor imaging: From anatomy to physiology. American Journal of Neuroradiology, 2006. 27(3): p. 475-487. 16. Timmers, I., et al., White matter microstructure pathology in classic galactosemia revealed by neurite orientation dispersion and density imaging. Journal of Inherited Metabolic Disease, 2015. 38(2): p. 295-304.


Figure 1: Number of tissue samples (patients) within each tumor type.

Figure 2: Normalized median values of NODDI and DTI ROIs where NE(T): NE tissue samples and CE(T): CE tissue samples with tumor score 2 and 3

Figure 3: Normalized median values for ADC and Vin across all tumor scores. (a) nADC box-plot (b) nVin box-plot (c) Line plot of nADC and nVin median values

Table 1: Correlation between DTI and NODDI metrics and histopathology of tissue samples with tumor scores 2 and 3. Variables with p <0.005 highlighted in bold. Arrows indicate positive (up) or negative (down) correlation

Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)