Diffusion Tensor Imaging in Neurofluids

Introduction

The glymphatic system is a brain-wide fluid exchange network that facilitates waste clearance¹. Cerebral spinal fluid (CSF) circulates from the subarachnoid space into the perivascular space of penetrating arterioles, where it interacts with the interstitial fluid (ISF) of the brain’s parenchyma through astrocyte aquaporin 4 (AQP4) water channels^1,2. This fluid then moves to the perivenous space and exits through the lymphatic system or along spinal nerves. The glymphatic system and neurofluid transport has attracted considerable interest for its role in brain waste management and the pathophysiology of neurodegenerative diseases. Research into the glymphatic system presents an exciting opportunity to better understand disease mechanisms, potentially prevent or delay disease onset, and develop new treatment strategies. However, there is a critical need for a noninvasive imaging method to understand fluid exchange and solute clearance in the human brain.

Magnetic resonance imaging is well suited to study solute and fluid exchange. The first dynamic in vivo imaging of the glymphatic system, which tracked the kinetics of a tracer injected into the cisterna magna, was achieved in the mouse brain using two-photon microscopy¹. This was soon followed by an MRI study in mice using a gadolinium-based contrast agent (GBCA), which demonstrated size-dependent solute exchange between the perivascular and interstitial spaces³. In humans, glymphatic solute transport can be visualized using similar MRI techniques with intrathecal (IT) GBCA, showing its movement from the CSF to the brain parenchyma^4–7. Although this method is currently considered the gold standard for assessing glymphatic function in the human brain, IT contrast is not approved for routine use and requires lengthy wait times for post-contrast imaging.

In addition to solute transport, neurofluid flow is a crucial component of the glymphatic system. Diffusion-weighted MRI, which is sensitive to both molecular diffusion and bulk flow of water, has emerged as a promising noninvasive technique for studying the glymphatic system. This comprehensive review article outlines the principles of diffusion imaging as they pertain to glymphatic function and provides an overview of current techniques.

Principles of Diffusion MRI

Diffusion refers to the displacement of water molecules resulting from Brownian motion. In an unrestricted compartment, diffusion is isotropic, occurring equally in all directions. However, the brain’s complex microstructure – comprised of various cells and structural elements – restricts diffusion resulting in a directional component. For example, water protons associated with ventricular cerebrospinal fluid (CSF) exhibit greater mobility compared to those confined within intracellular compartments. Water diffusion in the extracellular space of the brain parenchyma—known as interstitial fluid (ISF)—represents an intermediate level of restriction i.e. not as restricted as within the cell but still limited by cells, blood vessels, and other structural barriers. (see Figure 1).

Figure 1:

Diffusion of water in the brain occurs at different spatial and temporal scales. Within the white matter neuron fiber bundles, diffusion is dominant and fast along the fibers than perpendicular to the fibers. In the gray matter, diffusion is relatively uniform in all directions in the absence of closely and geometrically aligned structures. In the ventricles, CSF diffusion is relatively uniform in all directions and fast due to the lack of any barriers providing the greatest signal of the three tissue types. Created with BioRender.com.

Diffusion-weighted imaging (DWI) quantifies the magnitude of water diffusion using diffusion-sensitizing gradients. The strength of these gradients is represented by the b-value, or diffusion-weighting. As the b-value increases, the signal intensity decreases exponentially while altering the sensitivity to diffusion. Low b-values will attenuate the signal of water molecules with a large amount of motion while higher b-values are required to measure slow moving molecules likely restricted by physical barriers such as cells, organelles, and fluid composition. The selection of b-value can alter the sensitivity of the method to diffusion in different tissue types and brain regions (Figure 2). A general rule of thumb to be sensitive to the apparent diffusion of the tissue of interest is to pick a b-value where b × D = 1, where D is diffusion in mm²/s.

Figure 2:

Different physiological processes can be interrogated at different b-values in the brain. Created with BioRender.com.

At low b-values (< 500 s/mm²), the diffusion signal is sensitive to water movement in the tissue microstructure as well as CSF movement along perivascular spaces and the Aqueduct of Sylvius. Low to intermediate b-values can also be used to measure pseudorandom motion with a technique called intravoxel incoherent motion (IVIM), where the collective movement of water molecules with in a voxel is random and can be approximated by gaussian diffusion⁸, such as the incoherent flow of CSF or capillary blood flow. While blood flow can be measured with IVIM, other MRI approaches such as arterial spin labeling are more commonly used to study cerebral blood flow and hence it is not discussed in this review. CSF movement can also be described by pseudorandom diffusion through the trabeculae in the subarachnoid space using IVIM^9,10. All low b-value diffusion imaging methods are discussed in detail below in free water fraction imaging, low b-value imaging and spectral diffusion imaging.

At intermediate b-values (^~1000 s/mm2), diffusion properties of water within the tissue provide insights into tissue structure and integrity, particularly in white matter tracts by measuring the magnitude and directionality of diffusion (Figure 1). Diffusion tensor imaging (DTI) uses direction-encoding diffusion gradients to estimate a tensor. The principal eigen vector of the tensor represents the dominant or fastest direction of extracellular diffusion along axons. In the context of neurofluids, this approach is described further in single-shell DTI and DTI-ALPS. Sensitivity to tissue microstructure can be further increased by using b-values between 1000–3000 s/mm².

At very high b-values (>3000 s/mm²), the imaging becomes sensitive to restricted diffusion within cellular structures. This can provide insights into cellular density, the presence of barriers to diffusion (such as cell membranes), and intracellular versus extracellular water diffusion (including that relevant to the glymphatic transport of solutes). Especially, very high b-values (^~10,000 s/mm²) could be sensitive to intracellular and extracellular diffusion, and even molecular composition in some pathologies¹¹.

Finally, the diffusion-weighted signal arises from the random motion of water molecules and is therefore approximated by a Gaussian probability distribution. While the standard model is accurate in many cases, it does not fully capture the complexities of water molecule displacement in the extra- and intracellular compartments of the tissue, where diffusion deviates from Gaussian behavior¹². Diffusion kurtosis imaging (DKI) extends the capabilities of the standard diffusion model by accounting for non-Gaussian diffusion (known as excess kurtosis) in the complex microenvironment, especially at high b-values (>1500 s/mm²) and may reflect the ISF space¹³.

This review covers the currently published diffusion techniques used to investigate CSF and/or ISF neurofluids. Table 1 provides a detailed description about imaging protocols, and findings from studies as they relate to the CSF or ISF flow and glymphatic clearance. Figure 3 provides a visual of the different brain regions that can be interrogated by the diffusion imaging approaches listed below.

Table 1:

Summary of studies using the different diffusion imaging methods for investigating glymphatic or neurofluid transport.

Author/year	Technique/model	b-values (s/mm2)	Species	ROI
Jiang, C. et al. 2014	Single-shell DTI, single compartment	b=1000; 20-directions	Human	parenchyma
Elvsåshagen, T. et al. 2014	Single-shell DTI, single compartment	b=1000; 32-directions	Human	parenchyma
Tuura, R. O., et al. 2021	Single-shell DTI, single compartment	b=1000; 35-directions	Human	CSF spaces, parenchyma
Taoka, T. et al. 2017	Single-shell DTI-ALPS, single compartment	b=1000, 2000; 30-directions	Human	Medullary vein PVS
Thomas, C. et al. 2018	Multi-shell DTI, isotropic FWF	b= 0, 300, 1100; 10, 10, 60-directions	Human	parenchyma
Sepehrband, F. et al. 2019	Multi-shell DTI, FWF	b = 1000, 2000, 3000; 90-directions b = 500, 1000, 2000; 6, 48, 60-directions	Human	parenchyma
Sepehrband, F. et al. 2019	Multi-shell DTI, anisotropic FWF	b = 500, 1000, 2000; 6, 48, 60-directions	Human	parenchyma
Pieciak, T. et al. 2023	Multi-shell DTI, spherical means FWF	b = 1000, 2000; 60, 30 directions	Human	parenchyma
Jiaerken, Y. et al. 2021	Multi-shell DTI, NODDI	b = 1000, 2000; 30 directions	Human	parenchyma
Dumont, M. et al. 2019	Single-shell DTI, isotropic FWF	b = 1000; 41 directions	Human	parenchyma
Kamagata, K. et al. 2022	Single-shell DTI, isotropic FWF	b = 1000; 41 directions	Human	parenchyma
Li, H. et al. 2024	Single-shell DTI, isotropic FWF	b = 900; 61 directions	Human	parenchyma
Harrison, IF. et al. 2018	Low b-value DWI and DTI	b = 107; 3 or 6-directions	Rat	PVS MCA, CSF
Evans, P. G. et al. 2023	Low b-value	b = 43; 2-directions	Rat	PVS MCA
Hirschler L, et al. 2022	Low b-value	VENC = 5 mm/s; 6 directions	Human	PVS, CSF
Bito, Y, et al. 2021	Low b-value	b =100, 1000; 15 directions	Human	CSF
Bito, Y, et al. 2023	Low b-value	b =100, 1000; 13 directions, 3 diffusion times	Human	CSF
Wen, Q. et al. 2022 and 2024	Low b-value DWI	b = 150; 3-directions	Human	PVS
Han, G. et al. 2024	Low b-value DTI	b = 150; 6-directions	Human	PVS, parenchyma
Taokoa et al. 2019 and 2021	Low b-value DWI	b = 500; 3 directions	Human	PVS MCA
Taokoa et al. 2021	Multiple low b-value DWI	b = 50, 100, 200, 300, 500, 700, 1000; 3 directions	Human	CSF
Pierobon Mays, G et al. 2024	Multiple low b-value DWI	b = 0, 50, 100, 200, 300, 700, 1000;	Human	CSF
Levendovszky, SR et al. 2024	IVIM	b = 10, 20, 40, 80, 100, 150, 200, 300, 500, 700, 900, 1000; 6 directions	Human	CSF
Yamada et al. 2023 and 2024	IVIM	b = 0, 50, 100, 250, 500, 1000; 3 directions	Human	CSF
Wong SM et al. 2020 van der Thiel, MM 2021 and 2023	IVIM	b = 0, 5, 7, 10, 15, 20, 30, 40, 50, 60, 100, 200, 400, 700, 1000; 3 directions	Human	Parenchyma
Drenthen et al. 2024	IVIM	b = 0, 30, 90, 210, 280, 350, 580, 620, 660, 680, 720, 760, 980, 990, 1000; 1 direction	Human	Parenchyma
Örzsik, B. et al. 2023	DKI	b = 800, 2000; 32, 64 directions	Human	Parenchyma
Örzsik, B. et al. 2021	DKI	b = 20–2518 (57 total); 3 directions	Rat	Parenchyma
Lindhardt, T. B. et al. 2024	DKI	b = 800, 1800; 20 directions	Mouse	Parenchyma
Debaker, C., et al. 2020	DKI	b = 250, 1750; 6 directions	Mouse	Parenchyma

Figure 3:

Different diffusion techniques probe different facets of the glymphatic system. Low b-value imaging with or without multiple diffusion times, intravoxel incoherent motion (IVIM) are approaches used to study the pseudorandom motion in the subarachnoid CSF spaces, which is upstream of the perivascular CSF flow and CSF-ISF exchange. Low b-value approaches and DTI along the perivascular space (DTI-ALPS) using standard b-values ^~ 1000 s/mm² are used to understand perivenous CSF motion. Other methods spanning simple single shell DTI at b= 1000 s/mm², multi b-value free water fraction (FWF) imaging, spectral diffusion, and high b value diffusion imaging such as diffusion kurtosis imaging (DKI) are used to better understand ISF diffusion and to separate different types of water motion in the white matter. Created with BioRender.com.

Current Diffusion Imaging Techniques to Investigate the Glymphatic System

1. Single-compartment Diffusion Tensor Imaging (DTI)

Glymphatic function is related to sleep¹⁴ and circadian rhythm¹⁵, with a reported 60% increase in interstitial fluid (ISF) during slow wave sleep in rodents¹⁴. Consistent with the glymphatic hypothesis, time-of-day variations in diffusion measurements are observed in the human brain^16,17, including mean diffusivity (MD), fractional anisotropy (FA), and apparent diffusion coefficient (ADC), measured with single b-value (1000s/mm²) DTI. It was hypothesized that these time-of-day variations in diffusion metrics are related to changes in tissue ISF fluid content and glymphatic activity, later supported by multicompartment studies discussed in Section 2. Another study reported decreased MD in cerebrospinal fluid (CSF) spaces and increased MD in the brain parenchyma in morning compared to evening DTI measurements¹⁸. The MD changes were associated with rapid eye movement (REM) sleep duration but not with non-REM slow wave sleep, measured with overnight EEG. This finding contrasts other animal studies linking glymphatic influx with slow wave sleep¹⁹. It is likely that, in addition to increased ISF volume, changes in MD are associated with other physiological processes.

DTI along the perivascular space (DTI-ALPS) measures diffusivity of the CSF in the perivascular space at the level of the ventricular body²⁰. In this region, the orientation of the medullary veins and the associated perivascular space are perpendicular to the ventricular body (X direction). White matter fibers are oriented perpendicular to the medullary vein and its perivascular space, aligning along the Y and Z directions. Therefore, diffusion in the X direction is assumed to partially represent the perivascular space surrounding the medullary vein. The DTI-ALPS index is calculated as the ratio of mean diffusivity in the X direction to mean diffusivity in the Y and Z directions, with a higher index indicating increased diffusivity within the perivascular space. An increased ALPS index has been positively correlated with global cognitive function²⁰.

DTI-ALPS is computationally straightforward method that can be applied retrospectively to standard DTI imaging protocols (b=1000 s/mm²), contributing to its wide adoption and implementation. Significant associations of DTI-ALPS have been reported multiple diseases and conditions including AD^21,22, Parkinson’s disease²³, diabetes²⁴, obstructive sleep apnea²⁵, vascular disease²⁶. To date, DTI-ALPS is the most published diffusion technique used to investigate glymphatic function. However, several important limitations are also discussed in the literature^27–30, including the manual ROI placement, reliance on a single location in the subcortical white matter with no colocalization with the underlying vein, use of a region with low contrast enhancement on the gold standard intrathecal studies, and nonnegligible contributions of radial diffusivity of surrounding white matter tracts³¹. Despite these limitations, DTI-ALPS, like all diffusion techniques, has not been validated as a measure of glymphatic function and changes in the ALPS index should not be conflated with changes in glymphatic function²⁹. One study reported a correlation between ALPS index and intrathecal contrast³², however, a recent study found no such association³³, highlighting the need for further validation.

2. Free Water Fraction Diffusion

Free Water Fraction (FWF) diffusion imaging was initially proposed to disentangle the diffusion signal of extracellular free water from that of tissue compartments, allowing for more accurate DTI estimations^34,35. The time-of-day dependence in diffusion metrics is hypothesized to reflect glymphatic-related increase in ISF as well as increased fluid in the perivascular space (PVS)¹⁴. The diffusion signal can be fitted to a bi-exponential model that assumes two compartments: a tissue compartment and an isotropic free water compartment (reflecting primarily the ISF) with a diffusivity of water = 3 × 10⁻³ s/mm2 at 37°C – faster than the diffusivity typically calculated for gray or white matter tissue. Single-shell protocols using a dual compartment model can estimate parenchymal FWF, with increased FWF reported in Alzheimer’s disease (AD)^36,37 and (cerebral small vessel disease) CSVD³⁸.

Using the same two-compartment model but with a multi-shell acquisition, Thomas et al. demonstrated that increased diffusivity measures in the morning compared to the evening were driven by an increase in CSF/ISF-like FWF³⁹. Other techniques to model FWF have been presented including an anisotropic free water compartment to describe PVS⁴⁰, the isotropic diffusion component in neurite orientation dispersion and density imaging (NODDI)^41,42, and a spherical means techique⁴³. Multi-shell protocols have also shown increases in FWF in mild cognitive impairment⁴⁴ and with increasing age⁴⁵. These protocols use a low b-value shell (b^~500 s/mm²) to improve accuracy of FWF and are preferred over single-shell acquisitions which should be interpreted with caution⁴⁶.

3. Low b-value

In the brain, water molecular motion is governed not only by diffusion but also by bulk flow. One component of glymphatic solute transport is the CSF flow across the perivascular space and subarachnoid as well as bulk across the ISF. The velocity of cerebrospinal fluid (CSF) can be modeled as the sum of a turbulent (or incoherent) component and a stationary component⁴⁷. Turbulent flow has similar dynamics to those of molecular diffusion but occurs on a larger scale and faster rate. Therefore, low b-values (<200–500 s/mm²) are sensitive to a combination of incoherent flow and molecular diffusion, making them well suited for evaluating CSF flow⁴⁸ and glymphatic function.

Harrison et al, used low-b value (107 s/mm²) to assess fast CSF movement along the subarachnoid space and perivascular space of the middle cerebral artery MCA in the rat brain⁴⁹. To improve specificity to CSF, a long echo time was used to suppress the signal of the surrounding blood and tissue. Their results indicate that the low b-value diffusion coefficient can detect the directional dependence of fluid movement in the PVS parallel to the direction of arterial blood flow. Furthermore, using cardiac gating, they found that this diffusion was dependent on the cardiac cycle, with an increase of 300% measured during diastole. This study represents the first MRI-based measurement of pulsatile PVS flow, and the findings are consistent with prior invasive methods showing that cardiac pulsatility drives glymphatic flow^50,51. A follow up study showed that this method detects changes in perivascular fluid movement in a pharmacological model of hypertension⁵².

Cardiac cycle-dependent CSF mobility was extended to the human brain using a long echo time and high-resolution diffusion imaging (0.45 mm3) achievable with a 7T scanner^53,54. Like previous animal studies, CSF mobility was measured around large arteries and ventricles. Notably, the high resolution at 7T allowed measurement of CSF mobility in the perivascular space (PVS) of penetrating arteries in the cortex and within enlarged PVS in the basal ganglia⁵⁴. These findings support cardiac pulsation as a driver of PVS fluid flow in the human brain. Ran et al. introduced a diffusion-prepared (b = 100 s/mm2) 3D TSE sequence to measure PVS motility at 3T with 1 mm3 resolution. Their results indicate that the PVS around the MCA exhibited the highest motility related to the cardiac cycle, and this motility decreased with age⁵⁵. The high sensitivity was limited to PVS around the MCA due to the poorer resolution afforded by typical clinical scanners.

Although not yet conclusively proven, cardiac and respiratory pulsations are believed to be potential drivers of the CSF-ISF exchange and subsequent bulk flow in the ISF to transport solutes via the glymphatic system^50,51. The perivascular space of the penetrating pial arteries downstream of the MCA may have a different magnitude of CSF pulsatility. Measuring CSF dynamics in the penetrating arterioles is crucial for assessing glymphatic function; however, the resolution constraints of more widely available 3T scanners present a limitation.

Wen et al. proposed dynamic DWI, a method at 3T that overcomes the resolution limitations of prior low b-value diffusion studies⁵⁶. By leveraging the cardiac motility of CSF detected by DWI, they developed an iterative method to segment surface PVS spaces without requiring high spatial resolution or extended echo times. This approach is supported by co-registration with angiography, though it is still collected at a lower resolution than the PVS. Using retrospective gating and DWI with a b-value of 150 s/mm2 in three directions, their results demonstrate a strong cardiac cycle dependence of CSF dynamics consistent with prior findings. Dynamic DWI was also employed to measure the time delay between finger pulse oximeter readings and perivascular CSF arterial pulsation, exploring a novel method to assess intracranial pulse waves⁵⁷. The time delay was increased in the older participants compared to the younger group and this method is proposed as a potential biomarker of arterial stiffness.

Bito et al. present a mathematical framework suggesting that pseudorandom motion of CSF can be modeled as a convolution of linear flow and molecular diffusion⁵⁸. Diffusion components were observed at both b-values, with sensitivity to fast diffusion at b = 100 s/mm2 and slow diffusion at b = 1000 s/mm2. At low b-values, CSF compartments (third and fourth ventricles, subarachnoid space, foramen of Monro, Aqueduct of Sylvius) showed high anisotropic diffusivity compared to the ventricles. Flow in some regions (prepontine cistern, sylvan fissure) was pseudorandom, while in others (Aqueduct of Sylvius) it was directional. This revealed simultaneous fluid dynamics, though their relative contributions could not be determined. A follow-up study used multiple diffusion times at b = 100 s/mm2 to linearly fit MD with diffusion time, quantifying linear flow and pseudorandom diffusion contributions to obtain the velocity distribution (V) and diffusion coefficient (D)⁵⁹. Low V and D in the lateral ventricles suggest that CSF motion is primarily driven by pseudorandom diffusion, while high V and moderate D in the 3rd and 4th ventricles and sylvian fissure indicate more linear or directional flow. This study quantifies two distinct components of the diffusion signal across different CSF compartments.

The studies presented here used a single b-value protocol; however, a single b-value cannot resolve velocity or distinguish signal loss due to increased dispersion or compartment mixing. A range of b-values have been used (100–500 s/mm²)⁶⁰ and further investigation is needed to determine the optimal b-value for CSF and possibly ISF. Finally, the same b-value may not be applicable to all species or brain regions due to the differences in brain size, relative white matter volumes, and distribution of AQP4 channels⁵³.

4. Spectral / IVIM

In addition to a single low b-value, a range of low to intermediate b-values has the potential to provide increased specificity. Intravoxel incoherent motion (IVIM) is a diffusion model that incorporates additional sources of dephasing beyond molecular diffusion, including incoherent flow such as that found in CSF⁶¹. Introduced by Le Bihan et al, IVIM can evaluate CSF dynamics⁹ and it has been shown that CSF/ISF signal contributes to the IVIM signal and is greater than typical diffusion coefficient calculated at b = 1000 s/mm^{2 62–65}. A full range of b-values and biexponential fit allows for the separation of a fast pseudo-diffusion and slow molecular diffusion compartment. The signal from the fast compartment could have multiple sources including the microvasculature, CSF or the ISF^66,67. Of relevance to the glymphatic system, this includes CSF spaces such as the subarachnoid, extracellular ISF in the parenchyma or fluid within the PVS.

IVIM has been used to evaluate CSF spaces including estimates of CSF motion in ventricular and cisternal spaces. CSF motion in the third ventricle was found to be correlated with age⁶⁸. Following a night of sleep deprivation, D* in the subarachnoid space was decreased compared to a night of normal sleep⁶⁹, suggesting D* is sensitive to sleep and potentially glymphatic physiology. IVIM measures in the CSF have also shown sensitivity in studies of patients with hydrocephalus^70,71 and Parkinson’s Disease⁷².

Traditionally, IVIM analysis assumes two compartments⁸. However, since the signal from the fast compartment could have multiple origins, Wong et al used a spectral analysis and reported the presence of an intermediate diffusion compartment distinct from the microvasculature and brain tissue⁷³. This intermediate fluid fraction is a potential marker of vascular disease and neurodegeneration and could be indicative of ISF motion^74–76. While three-compartment modeling likely provides increased specificity for ISF, it may be more prone to fitting errors. Continued investigation of robust modeling methods is essential⁷⁷. Additionally, spectral IVIM analysis requires a high number of b-values, and optimizing the number and values can reduce scan time⁷⁸.

5. Diffusion Kurtosis Imaging (DKI):

DKI accounts for non-Gaussian diffusion (known as excess kurtosis) in the complex microenvironment by using with high b-values (>1500 s/mm²)¹³. DKI has shown increased sensitivity in disorders like stroke, where intracellular cytotoxic edema reduces extracellular space, increasing ISF tortuosity and kurtosis. Changes in kurtosis may reflect alterations in the intracellular and extracellular environment caused by glymphatic function.

Örzsik et al observed decreased measures of mean kurtosis (MK) and radial kurtosis (RK) during sleep compared to wakefulness, with no changes in conventional DWI measures or tissue volume⁷⁹. Higher-order analysis suggested that changes in kurtosis were more likely due to changes in relative volumes of intra- and extracellular water rather than intercompartment exchange and they conclude that the sleep-associated decrease in kurtosis is consistent with the expected increase in ISF volume.

Anesthesia is known to affect glymphatic function in the mouse brain with anesthetic specific effects on extracellular space volume¹⁹ and several studies have evaluated DKI under different anesthesia regimes^80,81. Isoflurane-induced decreases in mean kurtosis may represent increased extracellular space volume but the mechanism behind this change cannot be fully discerned due to the widespread effects of anesthetics⁸⁰. Additionally, animal studies have reported diffusion changes in AQP4 water channel knockouts or inhibition that impairs glymphatic flow^82,83. One such study found increased shifted ADC (where diffusion coefficient is calculated at 1750 s/mm² using b = 250 s/mm² for reference, as opposed to b = 0 s/mm² in order to increase sensitivity to tissue microstructure environment) and decreased S-index (the degree of hindrance based on the kurtosis model) in the mouse cerebral cortex and hippocampus following AQP4 inhibition⁸⁴. The decreased hinderance may be due to increased extracellular space caused by AQP4 inhibition⁸³, making diffusion kurtosis a potential method sensitive to astrocyte activity. However, this finding needs to be confirmed, and the exact implications of kurtosis changes remain under investigation.

Unmet Needs and Future directions

Many reviewed diffusion methods focus on larger CSF spaces rather than the perivascular spaces (PVS) of penetrating arteries or interstitial fluid (ISF) in the parenchyma. While CSF compartments are important to glymphatic function, they don’t provide the full picture. Additionally, many diffusion methods are not specific to glymphatic function and may capture other physiological processes. Future work should focus on expanding work on emerging methods to improve measurements in penetrating PVS⁵⁶ as well as ISF volume and motion⁷³. Comparison of these diffusion imaging techniques with CSF and/or plasma solute concentration is necessary for validating their use for understanding glymphatic biology.

Simultaneous MRI studies in animal models and validation in human studies are needed for this purpose. Since glymphatic transport is a sleep-active process, sleep quality or deprivation is often used to experimentally validate MRI methods. However, most human studies are of the awake brain or sleep following sleep deprivation instead of natural sleep. Scans of natural sleep could help validate findings of increased ISF during sleep from animal studies in people. Currently, there are studies of functional MRI in the sleeping human brain⁸⁵ but similar experiments with diffusion imaging are lacking. The high demand on the gradients, and the RF energy from continuous scanning of some the above-described diffusion imaging methods are a barrier for performing these studies. Other ways to validate human MR methods beyond sleep could include pharmacological interventions of the noradrenergic system or stimulation. Prazosin, a α-adrenergic antagonist significantly increases ISF volume and glymphatic transport of tracers in mice¹⁴. Similar increases have been observed in human studies⁶⁶.

Multiple studies have established the complex nature of CSF and ISF flow where multiple components of fluid motion co-exist. Continuing the evaluating of these components with multi-shell diffusion imaging paired with high spatial resolution scanning should offer more insights into the fluid exchange processes in the human brain. Alternatively, diffusion time dependence of ADC signals can provide complementary information about fluid flow through the cellular microarchitecture. Newer techniques such as the neurite exchange imaging (NEXI)⁸⁶ are emerging that may help understand the interaction between intra- and extracellular fluid exchange that may affect the convective interstitial fluid flow in the tissue.

Conclusion

Diffusion imaging methods offer valuable insights into the structure and function of the glymphatic system, providing researchers with non-invasive tools to investigate its role in brain health and disease. From DTI to the emerging advanced techniques presented here, these methods enable the visualization and quantification of fluid dynamics and tissue microstructure within the brain. By leveraging diffusion imaging, researchers can deepen our understanding of glymphatic physiology, identify biomarkers of glymphatic dysfunction, and understand neurological conditions where glymphatic function is hypothesized to change, like neurodegenerative diseases such as Alzheimer’s and Parkinson’s. Continued advancements in diffusion imaging technology, coupled with interdisciplinary collaborations, hold the potential to further elucidate the role of the glymphatic system in brain health and disease.

Key points:

• Diffusion-weighted and diffusion-tensor MRI is sensitive to the diffusion of water molecules and can detect slow, incoherent fluid flow, making it a valuable tool for studying the glymphatic system noninvasively.

• Diffusion-based methods can provide insights into the movement and exchange of fluids within the brain at many spatial and temporal scales.

• Although sensitive to factors altering neurofluids and exchange, no diffusion technique is fully validated yet.

Synopsis.

In this review article, we describe the development and application of diffusion-based MRI methods for studying glymphatic physiology. Fluid exchange and solute transport are two key components of the glymphatic system. While solute transport is difficult to visualize without contrast agents, diffusion MRI can be sensitive to different temporal and spatial scales of fluid flow that may be relevant to glymphatic exchange. Typical diffusion tensor imaging (DTI) with b=1000 s/mm2 is sensitive to changes in tissue water content as measured by diffusion metrics like mean diffusivity (MD), fractional anisotropy (FA), and apparent diffusion coefficient (ADC) but it is confounded by many factors such as edema, inflammation, and atrophy. Low b-value diffusion imaging, on the other hand, has been found to be sensitive to perivascular and subarachnoid CSF flow. Here we describe the use of low b-value imaging, free water fraction imaging, and diffusion time sensitization to leverage CSF as well as ISF motion in the parenchyma. We also describe multiple b-value (spectral or intravoxel incoherent motion, IVIM) diffusion imaging to better delineate diffusion components (directional vs pseudorandom, fast vs. slow) within the brain. Finally, we touch upon newer approaches that use advanced models of the diffusion signal, including high b-value imaging, such as kurtosis imaging or s-index derivation. These methods evaluate the contributions of intracellular and extracellular water to the diffusion signal for understanding the changes in ISF volume during the sleep-active glymphatic transport. We briefly enumerate the applications of these methods in different neurological diseases.

Clinics Care Points.

• Diffusion tensor imaging of cerebrospinal fluid and interstitial fluid flow changes is sensitive to many neurological disorders but is still diagnostically non-specific