Correlation between MRI-derived water content and conductivity in tumour and healthy tissue: how much cell water is active?
Ana-Maria Oros-Peusquens1, Yupeng Liao1, and N. Jon Shah1

1INM-4, Research Centre Juelich, Juelich, Germany


About 80% of brain water is found inside the cells and a large fraction of it is interfacial water with properties substantially different from those of bulk water. Evidence for a large osmotically unresponsive compartment, available from literature, is substantiated by the finding that a very large fraction of brain water does not contribute to its electrical conductivity. This is determined by investigating the correlation between conductivity and water content in tumour patients in vivo. More than 80% of brain water is found to be unresponsive, with variations reflecting tissue and tumour type. This work describes a noninvasive method for the characterisation of a deeply microscopic parameter of the living tissue.


Nearly 80% of the brain consists of MR-visible water [1, 2] and about 80% of brain water is found inside the cells [3]. One of the most important properties of intracellular water is its ability to act as solvent for cellular constituents. An important fraction of cell water is consequently found at the interface between solutes and bulk water. It has very different properties from bulk water, e.g. greatly reduced freezing point [3] and is frequently referred to as ‘bound water’. Its distribution in frozen tissue can be visualized by MRI [5] and amounts to approximately 10% of tissue water.

Cells are known to swell and shrink when exposed to hypo or hyperosmotic aqueous solutions of cellular solutes, due to movement of water along the concentration gradient. This reflects the osmotical responsiveness of cell water. There is evidence, however, that not all of cell water is osmotically responsive [4]. Surprisingly, a very large fraction of cell water (25 to 92%), much larger than that of unfreezable water, was found to be osmotically unresponsive, depending on the cell type [4].

Electrical conductivity of soft and tumour tissue was found to correlate very well with tissue water content [6]. However, the correlation can be described within a model of protein solutions only when assuming that a large fraction of cell water is not involved in electrical conductivity [7]. This was interpreted as due to water of hydration.

The electrical conductivity of tissue can be measured by MRI [8]. We investigate in the following the correlation between water content and electrical conductivity, measured simultaneously by MRI, as an indicative of the percentage of cell water which does not participate in conduction. The precise nature of this water remains to be clarified. Its magnitude and dependence of tissue type (WM, GM, tumour) is illustrated with results obtained from two tumour patients.

Materials and methods

Two brain tumour patients (1 glioblastoma, 1 meningioma, female, aged 33 and 64) were investigated in a hybrid MR-PET 3T scanner as detailed in [9]. Briefly, water content and electrical conductivity were determined based on the magnitude and phase information of a 2D multi-echo GRE scan with TR=10s and nominal flip angle of 90deg. Measurement parameters included FOV=200×162mm2; slice thickness=1.5mm; 0.75mm gap; TE1=3.87ms;echo separation Δ(TE)=4.08ms, monopolar readout, 12 echoes; acceleration factor=2.The acquisition time (TA) was 7:21 min. Reconstruction of the electrical conductivity is based on electrical properties tomography [8]: σ=(Δφ)/(2μω), where Δ represents the Laplacian operator, μ the magnetic permeability, and ω the Larmor frequency. The transceiver phase was calculated from the 12 multi-echo GRE phase profiles by interpolating them to TE=0 after unwrapping. Water content maps were produced from the signal intensity extrapolated to TE=0 by monoexponential fitting of the echo train and corrected for transmit and receive field inhomogeneities using the bias field correction algorithm implemented in SPM12, which also defines masks for tissue classes. The water content map was obtained by calibrating tissue values to the saturation-corrected CSF signal intensity [7]. Active tumour tissue was identified by dynamic FET-PET uptake and automatically segmented [7]. For the glioblastoma patient, oedema surrounding tumour was identified from MRI information and manually delineated. The correlation between water content and conductivity was described in a linear model and the best fit parameters were determined for pairs consisting of CSF as reference value and one of WM, GM, tumour and oedema regions.

Results and Discussion

Fig.1 summarises the results for the two patients. For the glioblastoma patient, the region affected by tumour was subdivided into oedema (pink), enhancing tumour (magenta) and nonenhancing tumour (red) and the regions were fit separately. The fit parameters are given in Table 1. The slope reflects the average ion concentration for each region and the intercept the non-responsive water.

All tissue classes show a large fraction of non-responsive water. Normal WM and GM show similar properties in the two patients. The fraction of water involved in conductivity is higher in the region with oedema, which is probably of vasogenic nature involving more extracellular water than healthy tissue.

The main finding of this preliminary study is that the water fraction which does not participate to electrical conduction in the brain is very large (>80% of tissue water) and depends on the tissue and probably also tumour type. The feasibility of voxel-by-voxel mapping of conduction-unresponsive water is being investigated and more patients added to the study. This work describes a noninvasive method for the characterisation of a deeply microscopic parameter of the living tissue which opens new avenues to the investigation of tumours and healthy brain.


The authors gratefully acknowledge the contributions of Dr. C. Weiss, Prof. K-J Langen, Dr. G. Stoffels and Dr. C. Filss regarding patient recruiting, patient handling and PET measurements.


[1] P. Tofts editor. Quantitative MRI of the Brain: Measuring Changes Caused by Disease. ISBN: 978-0-470-01429-5

[2] N.J. Shah. V. Ermer, A.M. Oros-Peusquens, Meth Mol Biol 2011, 711

[3] R. Cooke and I.D. Kuntz. Annu Rev Biophys Bioeng 1974

[4] I. Cameron and G.D. Fullerton Cell Biol Int 2014, 38: 610-614

[5] A.M. Oros, J. Kaffanke, N.J. Shah, Proc. ISMRM 2005, p. 2480

[6] J.L. Schepps and K.R. Foster, Phys Med Biol 1980, 25: 1149-1159

[7] H.P. Schwan, Ann. NY Acad Sci 1965, 125: 344-354

[8] T. Voigt, U. Katscher, O. Doessel, Magn Reson Med 2011, 66: 456-466

[9] A.M. Oros-Peusquens et al, Nucl Instr Meth A, 2014


Table 1. Water content (percent units) and conductivity (S/m) properties of different brain regions. Linear fits of the correlation between the two quantitites are described by slope and intercept. The percentage of unresponsive water is defined as tintercept he ratio between intercept and mean water content characterising a region.

Figure 1: Conductivity and water content in different brain regions for a) glioblastoma patient; and b) meningioma patient. CSF voxels are depicted in black, WM voxels in blue, GM in green, active tumour tissue (based on FET uptake) in red. For patient a) the oedema region is shown in pink.

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