Quantitative Sodium MR Imaging of the Human Brain at 9.4 Tesla provides Assessment of Tissue Sodium Concentration and Cell Volume Fraction during Normal Ageing

Introduction

The human brain shows increasing volume loss during normal ageing over the adult age range [1,2]. The cognitive decline in at least some cognitive domains with advanced normal age is also undisputed [3]. Initially, these two observations were incorrectly thought to be the consequence of extensive neuronal cell death [as reviewed briefly in 3,4]. This perception of cognitive decline due to tissue loss from extensive cell death during normal ageing was based on flawed methodology that has been refuted by modern neuropathology which now indicates that cell numbers and cell density remain essentially constant throughout normal adult ageing [3–10]. Volume loss in normal ageing is now attributed to reduction in cell size and complexity of dendritic arborization [3–10]. There may be some selective regions of cell loss in the prefrontal cortex and in the subiculum of the hippocampus during normal aging but this is not a global finding across the entire brain [3,11]. The cognitive decline may be related to other biophysical changes such as Ca²⁺ effects of increasing the afterhyperpolarization potential [3,12]. The neuropathology approaches have used ex vivo examinations of small tissue sections of the brain. This report now extends these microscopic observations to the in vivo macroscopic scale across the entire human brain. The method used is quantitative sodium magnetic resonance (MR) imaging in which the bioscales of tissue sodium concentration (TSC) and tissue cell volume fraction (CVF) are determined [13]. The methodology has been extended to the ultra-high field of 9.4 Tesla to provide higher spatial resolution than available on clinical 3 Tesla scanners as has been reviewed in detail elsewhere [14]. This cross-sectional study of a cohort of normal aging subjects confirms the modern neuropathological finding that the cell volume fraction is maintained in the normal aging human brain.

Experimental

Biological Model of the Tissue Sodium Concentration (TSC) and Cell Volume Fraction (CVF) Bioscales

A voxel placed in brain tissue can be represented in terms of a two-compartment model of the intracellular and interstitial volumes with sodium concentrations of 12 and 145mM, respectively. These concentrations are maintained within a small biological range in normal healthy tissue by multiple energy consuming sodium/potassium ion pumps referred to as the Na⁺/K⁺ ATPases. The TSC is the sum of the tissue cell volume fraction, CVF, weighted sodium concentrations in the intracellular, C_i, and interstitial, C_o, compartments, respectively, as:

$$\text{TSC}=\text{CVF}.{\mathrm{C}}_{\mathrm{i}}+(1-\text{CVF}).{\mathrm{C}}_{\mathrm{o}}$$ {1}

This equation can be rearranged to provide the tissue cell volume fraction as:

$$\text{CVF}=(\text{TSC}-{\mathrm{C}}_{\mathrm{o}})/({\mathrm{C}}_{\mathrm{i}}-{\mathrm{C}}_{\mathrm{o}})$$ {2}

Such models with the quantification of sodium MR signals have been given elsewhere and will not be elaborated on further [13,15–17]. As C_o and C_i show little biological variability in normal tissue due to multiple systemic, tissue-level and cellular homeostatic mechanisms, values can be estimated accurately. Thus, CVF can be derived in normal tissue when TSC is measured with quantitative sodium MR imaging.

The quantification of MR signal intensities as concentrations using calibration phantoms has been reviewed elsewhere [14]. Rather than using external calibration phantoms placed around the head of the patient, this study uses a large spherical calibration phantom with the same electrical loading and size as a human head in a separate imaging session with equivalent acquisition parameters to those used for the patient [13,17]. Although variations in electrical loading by heads of different sizes may occur, the power requirement for the 90° RF pulse is calculated for each subject and because human heads reduce the Q factor of the coil, the magnitude of this error is small. This use of larger calibration phantoms located at the center of the field of view (FOV) minimizes the magnitudes of the B1 and B0 corrections and partial volume effects that occur with small calibration phantoms restricted to the edge of the human head that is centrally placed in the FOV.

Human Subjects

All human subjects provided informed signed consent for this protocol that was approved by the Food and Drug Administration (FDA) and the Institutional Review Board.

The normal adults (N=49, age 48±19, range of 21–80 years, 47% male) had at least college education and were gainfully employed or, if retired, active in their communities. The subjects had no major medical, neurological or psychiatric disease disclosed by interview. All subjects over 60 years of age score in the normal range (28–30) on the mini mental status examination.

Imaging Instrumentation

All MR imaging was performed on a customized 80 cm clear diameter bore 9.4 Tesla superconducting magnet (GE Medical Systems Oxford Ltd, Abingdon, UK) fitted with a head sized asymmetric gradient set (clear bore diameter = 37.9cm, maximum amplitude = 98mT/m, maximum slew rate = 392mT/m/ms, GE Medical Systems Oxford Ltd, Abingdon, UK) [18]. The scanner electronics and software (Paravision 5.1, Bruker Biospin, Billerica, MA) were interfaced via customized transmit/receive switches to a customized quadrature birdcage RF coil. The long term field drift observed on this scanner has been less than 100Hz over many years with no evidence of short term drifts.

A spherical calibration phantom (16 cm diameter, Plexiglass) containing three 3 cm diameter cylindrical compartments (30, 70 and 110 mM sodium chloride in 3% agarose gel) was matched to the electrical loading of a human head in the RF coil using the appropriate surrounding solution (aqueous potassium chloride, 60 mM) [13].

Image Acquisition, Reconstruction and Quantification

Quantification of tissue sodium MR signals with short biexponential transverse relaxation times requires minimal TE values (0.26 ms measured from the center of the RF pulse of 0.5 ms duration and allowing 0.01 ms receiver and gradient switching time) to minimize loss of transverse signal. Saturation of longitudinal magnetization is avoided in tissue by using a long TR (160ms) value that is four times the T1 of tissue sodium (T1~35–40ms). Cerebrospinal fluid (CSF) is not fully relaxed (T1~55ms) and so cannot be used as a reliable internal calibration without correction for saturation. The flexible twisted projection imaging (flexTPI) sequence optimizes data acquisition efficiency (radial fraction = 0.305, K_max= 140.9 m⁻¹, projections = 3760, acquisition window = 10.04 ms, gradient = 5.47 mT/m, slew rate = 150 mT/m/ms, total acquisition time = 10 minutes for nominal isotropic resolution of 3.5×3.5×3.5 mm³) [13]. Frequency-segmented conjugate phase reconstruction was used to correct B₀ inhomogeneities using a B₀ map found from two sodium images of different TE values (0.26 and 1.26 ms) [19]. The B₁ variation across the FOV was corrected using a B₁ map determined from sodium images acquired at two different flip angles differing by a factor of two in power as described elsewhere [20]. The acquisition protocol and reconstruction algorithm for 9.4 Tesla is provided in Table I. The effects of eddy current effects and timing errors in the encoding gradients were corrected at 9.4 Tesla [21]. Head motion was minimal at less than a voxel dimension between sequential acquisitions in these cooperative subjects as required for accurate quantification [22] and did not require further spatial registration [23]. The image reconstruction and quantification procedure has been reported elsewhere [13,15,17].

Table I.

Algorithm for quantification of CVF in the human brain using the flexTPI pulse sequence at 9.4T [13]. Acquisitions of 10 minutes duration were obtained for each B₀ and B₁ correction of both head and calibration phantom.

Steps	23Na MRI Steps for Human Brain	Acquisition Parameters	Acquisition Time /Processing
1	Shimming on 23Na free induction decay	TR/TE=500/0.26 ms, flip angle<90°	5 minutes, no processing
2	23Na flexTPI acquisition (for quantification)	TR/TE=160/0.26 ms, flip angle=90°	step 2 with B0 & B1 maps with calibration give TSC & CVF
3	23Na flexTPI acquisition (for B0 map)	TR/TE=160/1.26 ms, flip angle=90°	steps 2 & 3 give B0 map
4	23Na flexTPI acquisition (for B1 map)	TR/TE=160/0.26 ms, flip angle=45°	steps 2 & 4 gives B1 map
	23Na MRI Steps for Phantom
5	Shimming on 23Na free induction decay	TR/TE= 500/0.26 ms, flip angle<90°	5 minutes, no processing
6	23Na flexTPI acquisition (for calibration)	TR/TE=160/0.26 ms, flip angle=90°	step 6 with B1 & B0 maps give calibration
7	23Na flexTPI acquisition (for B0 map)	TR/TE=160/1.26 ms, flip angle=90°	steps 6 & 7 give B0 map
8	23Na flexTPI acquisition (for B1 map)	TR/TE=160/0.26 ms, flip angle=45°	steps 6 & 8 give B1 map

Image Analysis

The TSC and CVF maps, obtained through the customized automated software pipeline (Matlab, Natick MA) were examined by regionally throughout the brain of each subject using regions of interest (ROI) analyses. The ROI were based on readily identifiable standard anatomical landmarks on the sodium images and included: frontal, parietal, temporal and occipital lobes, hippocampus, basal ganglia, thalamus, brainstem and cerebellum. The descriptive statistics (mean, median, standard deviation) were then used in statistical models to examine the age dependence of these bioscales for each ROI (Table II). Left and right-sided results were initially separated but combined for all ROI except the hippocampus when no statistically significant differences were detected.

Table II.

Tissue cell volume fraction, CVF, and tissue sodium concentration, TSC, values (mean±standard deviation SD) are tabulated for different brain regions in all adult subjects with normal cognitive function. The Pearson correlation coefficient, r, and corresponding p-value for the two-tailed t-value for each brain region show the absence of any significant dependence (p>0.05) of the CVF and TSC values on age despite the decreasing brain volume with advancing age. TSC is in units of mM.

ROI	N	CVF (Mean±SD)	TSC (mM) (Mean±SD)	r	p-value
Frontal	49	0.817±0.015	36.3±2.0	0.243	0.092
Parietal	49	0.819±0.013	36.1±1.8	0.244	0.0911
Temporal	49	0.811±0.013	37.1±1.7	-0.019	0.897
Occipital	49	0.827±0.014	35.0±1.9	0.061	0.677
Cerebellum	49	0.824±0.011	35.3±1.4	0.062	0.672
Brainstem	49	0.818±0.011	36.3±1.5	0.091	0.534
Thalamus	49	0.812±0.018	37.1±2.4	0.119	0.415
Basal Ganglia	49	0.809±0.017	37.5±2.2	0.091	0.534
Right Hippocampus	49	0.792±0.017 to 0.804±0.012	39.6±2.2 to 38.0±1.6	0.218 to 0.169	0.062 to 0.246
Left Hippocampus	49	0.786±0.019 to 0.799±0.014	40.4±2.6 to 38.7±1.8	0.269 to 0.183	0.069 to 0.208
Whole Brain	49	0.817±0.013	36.3±1.7	0.126	0.388

An estimate of percent tissue volume loss was made as the percent CSF space as measured by the ratio of the volume of CSF to the total intracranial volume encompassing the cerebral hemispheres and all surrounding CSF superior to the tentorium for each subject and plotted as a function of age. A linear correlation was performed to examine any age dependence of such estimated tissue volume loss (Figure 1b).

Figure 1.

a. (left) Tissue cell volume fraction CVF averaged over the whole brain as a function of subject age for cognitively normal adult subjects (N=49). The CVF shows a highly conserved mean value with a small variance (0.817±0.013, r=0.126, p=0.388). Specific brain regions with correlation coefficients indicating no age dependence are tabulated in Table II. All measurements were made at 9.4T.

a. (right) Tissue sodium concentration TSC (right), averaged over the whole brain as a function of subject age for cognitively normal adult subjects (N=49). The TSC shows a highly conserved mean value with a small variance (36.3±1.7, r=0.126, p=0.388). Specific brain regions with correlation coefficients indicating no age dependence are tabulated in Table II. All measurements were made at 9.4T.

b. CSF space (% of calvarial volume) is plotted as a function of increasing age showing significantly (r=0.8, p< 0.001) increasing volume loss with age. All measurements were made at 9.4T.

Results

TSC and CVF Bioscales in Cognitively Normal Brains as a Function of Age

As demonstrated in Figure 1a, TSC and CVF are conserved over the adult age range for cognitively normal individuals despite the volume loss of tissue that is well known to occur with aging and confirmed in this cohort by sodium imaging (Figure 1b). Although small variations in TSC and CVF exist across different regions of the brain (Table II), the correlation analysis indicates that no regions of the brain have TSC or CVF values that are significantly (p>0.05) dependent on age (Table II) despite the confirmation of the statistically significant (r = 0.8, p< 0.001) loss of tissue reflected in the increasing CSF volume with advancing age. Figure 2 provides a visual impression of this numerical result, presenting both CVF and TSC bioscales from representative young and old adults. The hippocampal CVF and TSC values demonstrated a significant difference (p < 0.02) between left and right sides and so were examined for age dependence without collapsing data from left and right structures. Neither right nor left hippocampal TSC and CVF show significant (p>0.05) age dependence, as indicated in Table II.

Figure 2.

a. Representative sodium bioscales for a young adult (24 years) with axial TSC maps (superior 4 rows) with TSC color scale from 0 to 150mM and corresponding CVF maps (lower 4 rows) with the CVF color scale from 0 to 1. All measurements were made at 9.4T.

b. Representative sodium bioscales for an elderly adult (73 years) with axial TSC maps (superior 4 rows) with TSC color scale from 0 to 150mM and corresponding CVF maps (lower 4 rows) with the CVF color scale from 0 to 1. The CSF space is greater in the elderly subject but the brain tissue remains green indicating similar tissue cell volume fraction. All measurements were made at 9.4T.

Discussion

The brain shows very little variance in the values of CVF and TSC with cognitively normal ageing (Figure 1a), despite the decreasing brain volume as reflected by increased CSF space depicted in Figure 1b. This result is the in vivo verification of the histological analyses reported from modern design-based stereology [4–7,10] and cell counting techniques [8,9]. This literature, reporting on selected human brain regions, indicates that cell number and cell density do not change significantly with advancing age in cognitively normal individuals and cannot explain the cognitive changes associated with aging [3]. This in vivo imaging result extends the scale of the histological findings across all large regions of the brain (Table II) for the nominal isotropic spatial resolution of 3.5×3.5×3.5 mm³. This result had been observed on lower spatial resolution sodium imaging at 3T (unpublished observations) and is now confirmed with the greater statistical power enabled by the higher spatial resolution at 9.4T and by the larger number of subjects. This finding of constant CVF across all regions of the brain suggests that a common mechanism may operate to preserve this structural parameter during normal ageing. Such a finding encourages speculation as to why CVF is conserved. The maintenance of ion concentration gradients across the cell membranes by the Na⁺/K⁺ ATPases is the major workload of the cell, both at rest and during signaling [25]. Changes in the ratio of the volumes of the interstitial and intracellular compartments may be posited to impact on the efficiency of these pumps to control ion homeostasis.

The hippocampal values of TSC and hence CVF in the left and right hemispheres shown in Table II showed a significant difference (p< 0.02). The histology literature has previously reported volume asymmetries [26,27] as well as differences in cell densities in specific subregions of the hippocampus in ageing as well as pathology [28,29].

Although our in vivo results are consistent with the ex vivo literature, the considerable CSF surrounding the hippocampus raises concern about partial volume effects (PVE). The point spread function (PSF) of projection imaging is about 60% greater than the nominal voxel dimensions [30,31]. This issue is even more relevant in neurodegenerative disease in which regional cell loss is increased. A non-quantitative sodium imaging study of patients with Alzheimer’s disease (AD) [32] reported on a sodium concentration index, based on the comparison of sodium imaging signal intensities from the hippocampus with other regions of the brain. Although consistent with cell loss in the hippocampus of AD patients, PVE of the CSF on values of TSC and CVF are also likely to have contributed to the changes, especially given the well-known loss of hippocampal volume in AD. Whether or not and how early CVF may detect hippocampal cell loss in the pathological process of AD remains to be investigated. If changes in CVF are detected early in the long latency period of AD, evaluation of early interventions within this prodromal period offers the greatest opportunity for halting clinical disease. CVF may be a quantitative parameter that is sufficiently sensitive to pathological progression prior to clinical symptoms to foster the development and evaluation of such early interventions during the latency period. A solution to PVE remains one of the most important challenges confronting this application of quantitative sodium imaging. Progress is being made on this matter primarily using approaches from PET [33–35].

The absolute values of TSC and CVF in brain tissue presented in Table II are similar to those reported using the same calibration and flexible twisted projection imaging method at 3T [13] and using external calibration references with twisted projection imaging at 3T [36]. Another group compared other quantitative projection pulse sequences [37–39] with the external calibration method at 9.4T to obtain similar tissue results [40]. Another different pulse sequence using a multi-planar 2D technique termed acquisition-weighted stack of spirals (AWSOS) produced similar tissue results at 9.4T [41].

The range of values for TSC in the brain reported in the literature has been summarized and discussed in detail elsewhere [14]. The numbers reported here are in the mid range of the reported values. The variability may relate to PVE. As the sodium signal from CSF is in close proximity with the cortical gray matter and cortical gray matter is not resolved, the values of CVF and TSC in Table II should be considered to be an estimate for a tissue mixture of white matter and gray matter.

This study has limitations beyond PVE that may reflect on the TSC and CVF values. The use of a separate calibration scan with a calibration phantom may produce systematic errors based on the differences in electrical loading of the RF coil by different sized subjects [13,14]. This is not expected to be a significant source of error as preliminary experiments demonstrated that the Q factor of the coil was considerable damped by the loading of a human head and coil tuning did not change significantly with the heads of different subjects. The B₁ mapping should also correct variations from different RF power if the loading changed from subject to subject.

Conclusions

Although sodium MR imaging at ultra-high field provides a spatial resolution that contains the major morphological features of the brain in clinically relevant acquisition times, such images may never compete with the exquisite anatomical detail obtained from the proton signal at lower fields. The relevance of sodium imaging is in the quantitative characterization of tissue ion homeostasis at spatial resolutions relevant to these anatomical dimensions.

The stability of the TSC and CVF bioscales in the normal human brain across the entire adult age range as measured by in vivo MR imaging reflects the modern histopathological findings of constant cell density with normal aging despite ongoing decreases in brain volume. This constancy may reflect a fundamental parameter of brain structure, possibly required to preserve normal cognitive function. The sodium bioscales are clearly sensitive to catastrophic cell death as in processes of ischemic stroke [36] or in tumors in response to successful therapy [42,43]. The sensitivity of TSC and TCF for detection of early neuronal cell death in neurodegenerative disease, possibly prior to clinical manifestations, remains to be defined.

Abbreviations

TSC

tissue sodium concentration

CVF

cell volume fraction

SNR

signal to noise ratio

ROI

regions of interest

CSF

cerebrospinal fluid

FDA

Food and Drug Administration

AWSOS

acquisition weighted stack of spirals

PVE

partial volume effects

Q factor

quality factor