Introduction

Water content is highly regulated in the healthy brain and changes of a few percent units are already indicative of pathology. Surprisingly, water content was found to be remarkably constant also in notoriously heterogeneous active tumour tissue irrespective of malignancy and despite conventional MRI visualising varying levels of tumour-related oedema¹. Very constant, tissue-specific water content was also reported in fixed tissue², despite large variability in other MR parameters and the obvious absence of active regulatory mechanisms. Furthermore, a recent study of the correlation between water content and electrical conductivity (Na content) of tissue appeared to reveal a large pool of tissue water which does not participate in conduction (osmosis), perhaps because it is closely associated with surfaces³. These observations suggest that water content might be associated to the presence of macromolecules, ions, membranes, etc, around which it forms layers of hydration; this simple property of tissue is thus intimately related to its deepest microscopic structure. Another microscopic probe of tissue organisation is offered by the MR-measurable diffusion of water molecules. Diffusion properties of water are determined by the presence and structure of microscopic barriers – e.g. myelin sheaths and cell membranes - in the path of free diffusion. PET imaging of amino acid transport with O-(2-F-18-fluoroethyl)-L-tyrosine (FET) is a powerful tool for brain tumour diagnostics⁴, and is assumed to be so by reflecting the density of aminoacid receptors on membranes and, thus, to some extent, cell membrane distribution. These very different parameters, water content, diffusion measures - fractional anisotropy (FA), mean diffusivity (MD), radial diffusivity (RD), axial diffusivity (AD), mean kurtosis (MK), radial kurtosis (RK) and axial kurtosis (AK) - and aminoacid PET can thus be reasonably expected to be, to some extent, correlated and to reflect, to some extent, similar microscopic structures. We investigated in the present study whether this hypothesis is substantiated by the data and to what extent.

Purpose

This study focuses on investigating correlations between the mentioned parameters and verifying previous water content results in tumour tissue with increased statistics using data from 40 patients. A unique feature of our experimental set-up based on a hybrid 3T MR-PET system is the possibility to measure all these parameters simultaneously with the patient in the same physiological state.

Materials & Methods

A cohort of 40 brain tumour patients was investigated (25 females, mean±std age 45,7±14,25 years). Based on histological findings, eight were firmly diagnosed with low-grade gliomas (LGG) and 10 with high-grade gliomas (HGG); for all others histology is in progress. Data were acquired on a 3T Siemens MR scanner with a Brain-PET insert. Complementing the MR-PET diagnostic protocol, a long-TR GRE scan for water content mapping¹ and a DKI protocol corresponding to clinical standards were included. Imaging parameters are given in Table I. All images were coregistered into the ‘water space’. Brain images were segmented into “healthy” and “tumour” hemispheres and WM/GM classes derived from water content mapping¹. Tumour masks were defined from PET. Correlations between all parameter combinations were investigated and found with few exceptions poor. The results are summarised in Table II and Figs 1-3.

Results and Discussion

Radial diffusivity is regarded as a measure of myelin and MD as reflecting cellularity, especially in oncology. However, when these tissue parameters are estimated (histology, myelin water), the correlations are often weak^8,9. Water content perfectly reflects its complement “protein concentration” (sum of normalised values is by definition 1) and thus can be expected to correlate well with MT measures and radial measures of diffusion. Indeed, among all other parameters, its correlation with RD was highest. For all parameters, correlations with water content within a tissue class are, however, small and the overall correlation is mainly due to the bimodal structure of water content. The only strong correlations found were between model-derived AWF and diffusion/axonal indices (AD, AK) which might reflect fitting/modelling issues. For tumours the correlation was poor, but the shape of the distribution (either horizontal or vertical and distinct from healthy tissue) might be further useful, especially if it shows changes with treatment (for example approaching that of healthy tissue). Further investigations are required to clarify this aspect.

Conclusion

Despite reasonable expectations of finding some microscopic-structure-driven correlations between water, diffusion and FET parameters, compared here for the first time, low correlations were found in healthy tissue. Interestingly, the correlation between water content and diffusion indices -while still rather low - increased in tumour tissue. Importantly, this implies that each parameter reflects different aspects and thus their combination should be more powerful for oncology than using single parameters.

Acknowledgements

No acknowledgement found.