Relationship between choline and apparent diffusion coefficient in patients with gliomas

Abstract

Purpose: To examine the relationship between apparent diffusion coefficients (ADC) from diffusion weighted imaging (DWI) and choline levels from proton magnetic resonance spectroscopic imaging (MRSI) in newly diagnosed Grade II and IV gliomas within distinct anatomic regions.

Materials and methods: A total of 37 patients with Grade II and 28 patients with Grade IV glioma were scanned on a 1.5T system with 3D MRSI and DWI. Region level analysis included Spearman rank correlation between median normalized ADC and choline for each patient per grade within each distinct abnormal anatomical region. Voxel level analysis calculated a Spearman rank correlation per region, per patient.

Results: Grade II lesions showed no evidence of a correlation between normalized ADC and choline using either the region or voxel level analysis. Region level analysis of Grade IV lesions did not appear to correlate in the contrast enhancement or necrotic core, but did suggest a significant negative correlation in the more heterogeneous nonenhancing and combined regions.

Conclusion: There appears to be differences in the relationship between ADC and choline levels in Grade II and Grade IV gliomas. Correlation within these regions in Grade IV lesions was strongest when all regions were included, suggesting heterogeneity may be driving the relationship.

Figure 1

Anatomical images, (a) SPGR and (b) FSE used to localize (c) spectra and align (d) ADC images. Choline heights were used to generate (e) choline maps, while (f) ADC maps were generated from ADC images resampled to the spectral resolution.

Figure 2

Example of segmentation of the T2All from the FSE image, the CEL and NEC from the SPGR image and a calculated NEL as (T2All–CEL–NEC) for a patient with a Grade IV glioma. Patients with Grade II lesions included only T2All regions of the lesion.

Figure 3

Results from patient in Fig. 2 with a Grade IV glioma (a) spectral slice with distinct anatomically abnormal regions labeled per voxel, applied to the nCho and nADC maps to generate a (b) 2D plot of the regions.

Figure 4

Region level analysis: median (nCho, nADC) per patient for Grade IV regions: NEC, CEL, NEL, and Grade II subtypes: Oligo, Astro, Oligoastro.

Figure 5

Voxel level analysis: Spearman rank correlation co-efficients using all voxels for (a) patients with Grade IV lesions grouped by regions and (b) patients with Grade II lesions grouped by subtypes.

Figure 6

Plot of patients with Grade IV gliomas. (a) Median values in NEC and CEL and (b) median values in NEC and CEL regions, effect when combining the regions, effectively showing two clusters of patients, one group who have voxels with mostly active tumor and one group who have voxels of mixed origin with nCho and nADC values more similar to the values in NECs.

Figure 7

Box plots showing range of values. Example of a patient with a Grade IV glioma (a) showing higher nCho within NEL than CEL regions with large ranges, while another patient with a Grade IV lesion (b) showing smaller ranges of nCho.

Figure 8

Linear relationship between the Spearman rank correlation coefficients of the CEL region for the patients with Grade IV lesions and the range of nADC values within the CEL region. This indicates that the large correlation coefficients are likely to be driven by region heterogeneity.