Diffusion Tensor Image Analysis ALong the Perivascular Space (DTI-ALPS): Revisiting the Meaning and Significance of the Method

Abstract

More than 5 years have passed since the Diffusion Tensor Image Analysis ALong the Perivascular Space (DTI-ALPS) method was proposed with the intention of evaluating the glymphatic system. This method is handy due to its noninvasiveness, provision of a simple index in a straightforward formula, and the possibility of retrospective analysis. Therefore, the ALPS method was adopted to evaluate the glymphatic system for many disorders in many studies. The purpose of this review is to look back and discuss the ALPS method at this moment.

The ALPS-index was found to be an indicator of a number of conditions related to the glymphatic system. Thus, although this was expected in the original report, the results of the ALPS method are often interpreted as uniquely corresponding to the function of the glymphatic system. However, a number of subsequent studies have pointed out the problems on the data interpretation. As they rightly point out, a higher ALPS-index indicates predominant Brownian motion of water molecules in the radial direction at the lateral ventricular body level, no more and no less. Fortunately, the term “ALPS-index” has become common and is now known as a common term by many researchers. Therefore, the ALPS-index should simply be expressed as high or low, and whether it reflects a glymphatic system is better to be discussed carefully. In other words, when a decreased ALPS-index is observed, it should be expressed as “decreased ALPS-index” and not directly as “glymphatic dysfunction”. Recently, various methods have been proposed to evaluate the glymphatic system. It has become clear that these methods also do not seem to reflect the entirety of the extremely complex glymphatic system. This means that it would be desirable to use various methods in combination to evaluate the glymphatic system in a comprehensive manner.

Introduction

In 2012, Iliff et al. proposed the glymphatic system hypothesis, which suggests the involvement of glial cells, the interstitial space, and cerebrospinal fluid (CSF) and cerebral interstitial fluid (ISF) in the elimination of waste products in the brain.¹ The concept of the glymphatic system was proposed as a pathway of the waste excretion system in the brain. Various studies have revealed the involvement of various systems, including CSF, cerebral ISF, blood–brain barrier (BBB), and meningeal lymphatic system. In this review, the term “glymphatic system” will be used as a representative term for these whole systems and also cover the concept of neurofluid dynamics.

The first report of the lymphocyte system by Iliff et al. was derived by using two-photon excitation fluorescence microscopy with fluorescent tracers injected into the subarachnoid space (SAS). This method has the advantage of direct observation of the brain in vivo, but the depth of view is only about 500 µm. In 2014, a technique was introduced to observe tracer behavior throughout the brain by administering gadolinium-based contrast agent (GBCA) intrathecally in rats and observing it with MRI over time.² This enabled observation of ISF dynamics, including deep brain regions. In humans, intrathecal administration of GBCA as a tracer has been carefully studied in a limited number of centers.³ However, intrathecal administration of GBCA is an unapproved use and is not generally available.

In 2017, the Diffusion Tensor Image Analysis aLong the Perivascular Space (DTI-ALPS) method was introduced.⁴ This method uses diffusion tensor images to calculate the ratio of diffusion in the perivascular space (PVS) direction with respect to the diffusion of free water in the interstitium. Although there are various problems as described below, the non-invasive and rather simple nature of the method has expanded its application to a variety of diseases and conditions. This promoted academic interest in cerebral ISF dynamics and waste elimination, which led many researchers to take an interest in this area. In the past few years, not only the DTI-ALPS method but also several other approaches to noninvasively evaluate ISF dynamics have been introduced, and their application in various diseases and conditions has been reported. This article reviews trends in the study of the glymphatic system, with a focus on the ALPS method. It discusses the meaning and significance of this method and explores how it should be utilized in the field.

Concept of DTI-ALPS

The concept of the DTI-ALPS method is depicted in Fig. 1. Within the cerebral hemispheres, the elimination of waste products from the brain parenchyma to the exterior necessitates transport either toward the brain surface or the ventricular walls. At the lateral ventricular bodies, the most direct path to either the brain surface or the ventricular wall occurs in the radial direction, specifically the left-right or x-direction. In the human brain, at the level of the body of the lateral ventricles, the medullary arteries, medullary veins, and their PVS run in this x-direction. This structure of medullary vessels within the brain parenchyma shows a construction that is purposeful in supporting the mass transportation that is the movement of material from and to the outside of the brain parenchyma. The basic concept of the ALPS method is to evaluate water movement in this x-direction. Here, the diffusion of large nerve fibers dominates in diffusion images. It is necessary to avoid the strong influence on diffusion by white matter fibers for the evaluation of subtle diffusion factors in the direction of medullary vessels. At the lateral ventricular body level, the projection and association fibers are each orthogonal to the x-direction, which allows the geometric elimination of the vector component of these large white matter fibers in the x-direction.^5,6 In the extreme, it may be sufficient to evaluate the diffusivity in the x-direction itself at the level of the lateral ventricular body when assessing mass transfer in this brain region. However, the diffusivity calculated in a diffusion image is greatly affected by the imaging device and the imaging sequence, such as diffusion time. In order to compare values of coefficients in different cases, it is desirable to evaluate the coefficients in ratio to some internal reference. For this reason, the ALPS-index was designed to evaluate the ratio of diffusivity in the x-direction to diffusivity perpendicular to both the x-direction and the projection fibers as well as perpendicular to both the x-direction and the association fibers as an internal reference, at the level of the lateral ventricular body. ALPS-index is given by following formula:

$$\mathrm{A}\mathrm{L}\mathrm{P}\mathrm{S}-\mathrm{i}\mathrm{n}\mathrm{d}\mathrm{e}\mathrm{x}=\frac{\mathrm{m}\mathrm{e}\mathrm{a}\mathrm{n}(\mathrm{D}\mathrm{x}\mathrm{x}\mathrm{p}\mathrm{r}\mathrm{o}\mathrm{j},\mathrm{D}\mathrm{x}\mathrm{x}\mathrm{a}\mathrm{s}\mathrm{s}\mathrm{o}\mathrm{c})}{\mathrm{m}\mathrm{e}\mathrm{a}\mathrm{n}(\mathrm{D}\mathrm{y}\mathrm{y}\mathrm{p}\mathrm{r}\mathrm{o}\mathrm{j},\mathrm{D}\mathrm{z}\mathrm{z}\mathrm{a}\mathrm{s}\mathrm{s}\mathrm{o}\mathrm{c})}$$

Fig. 1.

Concept of DTI-ALPS method. The assumed direction of mass transport within the brain parenchyma is shown in a. For waste material to reach the superficial cerebral veins or the subependymal veins, it must travel in the x-direction, indicated by the red arrow. In actual brain tissue, the medullary vessels also run in this x-direction. The positional relationship between the running direction of major fibers in the white matter outside the lateral ventricles and the location of the medullary vessels and perivascular space is shown in b and c. At this location, projection fibers and medullary vessels and association fibers and medullary vessels are in an orthogonal position to each other. This makes it possible to evaluate the diffusion component in the direction of the medullary vessels by eliminating the influence of the strong diffusion component of the projection and association fibers. The formula for the ALPS-index is shown in d. The ratio of the diffusion component perpendicular to both the projection fiber and the medullary vessels and the diffusion component perpendicular to both the association fiber and the medullary vessels is employed as an internal reference to evaluate the diffusion component in the direction of the medullary vessels. The image of the diffusion tensor ellipsoid is shown in e. The major axes of the ellipsoid correspond to projection fibers (blue) and association fibers (green), respectively. The small oval in e is the image of the ALPS-index. The oval that is large in the x-direction corresponds to a large ALPS-index, as shown in f. DTI-ALPS, Diffusion Tensor Image Analysis aLong the Perivascular Space.

where Dxxproj is x-axis diffusivity in the area of projection fiber, Dxxassoc is the x-axis diffusivity in the area of association fibers, Dyyproj is the y-axis diffusivity in the area of projection fiber, and Dzzassoc is the z-axis diffusivity in the area of association fibers. A large ALPS-index indicates that the diffusion of free water in the x-direction, or Brownian motion, is dominant, while a small ALPS-index indicates that free water movement in the x-direction is not dominant.

The first report of ALPS method was published as a measurement in a case of Alzheimer’s disease (AD).⁴ In evaluations of normal, mild cognitive impairment (MCI), and AD cases, the ALPS-index was significantly correlated with Mini-Mental State Examination (MMSE) score and significantly inversely correlated with age. As glymphatic system disorders are known to be present in AD,^1,7 these results suggest that the ALPS-index reflects the dysfunction of the glymphatic system in AD.

In considering the concept of the ALPS method, there are several points to note. It is not possible to distinguish whether the movement of water is inside or outside the PVS, or peri-venous or peri-arterial space due to the relatively low resolution of the diffusion image, in units of millimeters. Only the direction of the diffusion is distinguishable. However, what the ALPS method is trying to evaluate is the total or macroscopic movement of water in the direction of the PVS direction, not the movement of water in the PVS itself. Thus, these limitations are not a hindrance to this concept. In addition, since the method uses diffusion images, it evaluates the “movement” of water, not the “flow” of water. Thus, water movement toward the brain surface and toward the ventricles is equally subject to evaluation. Other issues will be considered in “Controversy in DTI-ALPS method” section.

Application of DTI-ALPS

Application for Alzheimer’s disease

AD is an irreversible, progressive brain disease that gradually damages the healthy brain, affecting memory, cognitive abilities, emotions, behavior, and mood.^8–13 Impairment of cerebral ISF dynamics in AD was reported in an early paper by Iliff et al.¹ Fluorescent-tagged amyloid β (Aβ) was transported along this route, and deletion of the aquaporin-4 (AQP4) gene suppressed the clearance of soluble Aβ, suggesting that this pathway may remove Aβ from the central nervous system. Thus, AD is considered one of the CNS (central nervous system) interstitial fluidopathies.^11,12 And, the first report of the DTI-ALPS method was also studied in patients with AD, indicating decreased ALPS-index in the AD cases.⁴

Several follow-up studies of the DTI-ALPS method in AD have been reported.^1,14–21 Steward et al. examined DTI-ALPS parameters in a cohort of older adults consisting of cognitively normal, MCI, and AD. MMSE, a brief test used to assess cognitive function, and Alzheimer’s Disease Assessment Scale–Cognitive subscale (ADAS-Cog 11) measurements were correlated with ALPS-index. The results showed significant differences in the right DTI-ALPS-index between the cognitively normal, AD and MCI groups in a univariate general linear model adjusted for age, gender and apolipoprotein E ε4 (APOEε4).¹⁸ This report is an early follow-up of the DTI-ALPS method in AD, and although it is an evaluation of the ALPS-index alone, it is a report that demonstrates the utility of the ALPS method, despite its limitations.

An association with amyloid and tau protein deposition by positron emission tomography (PET) has also been reported. Ota et al. investigated the association between ALPS-index and Aβ PET with 11C-Pittsburgh Compound B (PiB) and tau protein PET findings with 18F-THK5351 in patients with AD. Significant negative correlations between the ALPS-index and the standard uptake value ratio (SUVR) of 11C-PiB were found in bilateral temporal and left parietal cortex and left posterior cingulate gyrus in all subjects. Furthermore, there was a significant negative correlation between ALPS-index and SUVR of 18F-THK5351 in all subjects.¹⁹ This report correlates amyloid and tau PET results with the ALPS-index and also correlates with areas showing deposition. It is interesting to note that the ALPS-index correlates with tau and amyloid deposition in regions other than the white matter surrounding the lateral ventricular body, which is evaluated by the ALPS method. On the other hand, Matsushita et al. reported that SUVR of amyloid PET was not significantly associated with ALPS-index.¹⁷ Also, a comparative study with amyloid PET for subjective cognitive decline by Park et al. showed no significant difference between the mean ALPS-index of the amyloid-positive group and the mean ALPS-index of the amyloid-negative group.²² Kamagata et al. examined changes in the following noninvasive MRI metrics related to the perivascular network in patients with MCI and AD: PVS volume fraction (PVSVF), fractional volume of free water in white matter (FW-WM) and ALPS-index. Their results showed that AD patients had significantly higher total PVSVF, WM, basal ganglia PVSVF, and FW-WM and lower ALPS-index than healthy controls. Meanwhile, only the total PVSVF and WM values were significantly higher in the MCI group. Lower ALPS-index was associated with lower Aβ42, FDG (fluorodeoxyglucose)-PET uptake in the CSF and worse impairment in several cognitive domains. High FW-WM was also associated with lower CSF Aβ42 and worse cognitive function (Fig. 2).²⁰

Fig. 2.

Application of ALPS method in AD. Violin and box plots of the mean ALPS-index, PVSVF-ALL, PVSVF-WM, PVSVF-BG, PVSVF-Hipp, and FW-WM among the HC participants, patients with MCI, and patients with AD are shown. Compared to HC, AD patients had significantly higher total PVSVF, WM, BG PVSVF (Cohen d = 1.15−1.48, P < 0.001) and FW-WM (Cohen d = 0.73, P < 0.05) and lower ALPS-index (Cohen d = 0.63, P < 0.05). The MCI group had significantly higher total PVSVF (Cohen d = 0.99, P < 0.05) and WM (Cohen d = 0.91, P < 0.05). Figure quoted from Ref. 20. AD, Alzheimer’s disease; PVSVF, perivascular space volume fraction; WM, white matter; BG, basal ganglia; FW, free water; HC, healthy control; MCI, mild cognitive impairment.

Application for Parkinson’s disease and other degenerative disease

Parkinson’s disease (PD) is an alpha-synuclein disorder characterized by deposition of alpha-synuclein in neurons and loss of dopaminergic neurons in the substantia nigra.^23–25 Similar to AD, this neurodegenerative disease is characterized by the accumulation of toxic proteins as a pathologic feature. In recent years, it has been shown that PD is also associated with abnormal ISF dynamics related to metabolic waste elimination, and studies using the DTI-ALPS method have been reported.^26–36

A report of ALPS method analysis in cases of PD and essential tremor showed a lower ALPS-index in PD than in essential tremor.²⁷ In a comparison of stages of PD and normal controls, patients with PD had a lower ALPS-index than normal controls, and the decline was particularly pronounced in the group in the later stages of PD. There was a significant positive correlation between the ALPS-index and MMSE scores in the early PD group, and a negative correlation between the ALPS-index and the degree of PVS enlargement (Fig. 3).²⁸

Fig. 3.

Application of ALPS method in PD. Differences in ALPS-index between the PD group and NCs and between two PD subgroups and NCs were compared. Relationships between ALPS-index and MMSE score in early PD group (A), relationships between ALPS-index and EPVS score in early PD group (B) and relationships between ALPS-index and age in late PD group (C) are shown. There was a significant positive correlation between ALPS-index and MMSE score (β = 0.021, P = 0.029) (A), and a negative correlation between ALPS-index and EPVS score (β = − 0.050, P = 0.034). ALPS-index negatively related with age (β = − 0.012, P = 0.004). Figure quoted from Ref. 184. PD, Parkinson’s disease; NC, normal control; MMSE, Mini-Mental State Examination; EPVS, enlarged perivascular space.

Chen et al. used the DTI-ALPS method to evaluate ISF dynamics and its relationship to systemic oxidative stress status in PD patients. The PD group showing MCI and the PD group showing dementia had significantly lower ALPS-index than the normal control group. ALPS-index was inversely correlated with plasma nuclear DNA level, mitochondrial DNA level and cognitive score. The correlation between increased plasma nuclear DNA levels and decreased ALPS-index supports the idea that PD patients may exhibit increased oxidative stress associated with altered ISF dynamics.²⁹ In another report, a negative correlation between ALPS-index and severity of motor symptoms in PD was found in a subgroup of patients aged 65 years and older than in younger patients.³⁰ In a longitudinal study in patients with PD, the group with a lower ALPS-index deteriorated faster on The Movement Disorder Society Unified Parkinson’s Disease Rating Scale (MDS-UPDRS) Part III and Part II, Symbol Digit Modalities Test, and Hopkins Verbal Learning Test. Path analysis showed that ALPS-index acted as a significant mediator between tTau/Aβ1–42 and cognitive change in the Symbol Digit Modalities Test score at years 4 and 5.³⁵

Idiopathic rapid eye movement sleep behavior disorder (iRBD) is a sleep disorder in which the body moves during REM sleep and is a form of parasomnia. It has an aspect of alpha-synucleinopathy similar to PD and can result in phenoconversion to PD, Lewy body dementia, and multiple system atrophy. The overall conversion rate from iRBD to an overt neurodegenerative syndrome was 6.3% per year.³⁷ In the DTI-ALPS study of iRBD, PD, and normal controls, both PD and iRBD patients showed a lower ALPS-index than normal controls. PD patients had lower ALPS-index than iRBD patients, and ALPS-index correlated with cognitive decline in PD patients.³⁸ In a longitudinal study of phenoconversion in patients with RBD, PD, and normal controls, the risk of phenoconversion decreased with increasing ALPS-index.³⁹ Evaluations of neurodegenerative diseases other than PD, such as cortico-basal degeneration and progressive supranuclear palsy, have also been published.^40,41 In a study evaluating ALPS-index and correlating it with motor and cognitive function in cerebral cortical degeneration with basal cortical syndrome (CBD-CBS), the ALPS-index of CBD-CBS was significantly lower than that of normal controls. In addition, ALPS-index showed a significant positive correlation with MMSE and a significant negative correlation with the MDS-UPDRS III score.⁴¹

Application for small vessel diseases

Cerebral small vessel disease (SVD) affects small blood vessels in the brain and is associated with stroke, cognitive dysfunction, and motor dysfunction, and has also attracted attention as a background pathology for neurodegenerative diseases such as AD.^12,42–45 Benveniste et al. proposed that failure of brain fluid transport-via the glymphatic system plays a key role in initiation and progression of SVD. Their major case for this concept is that PVS are utilized as waterways for influx of CSF. Stagnation of transport by ISF is thought to cause loss of fluid homeostasis in the brain and may lead to transient white matter edema, perivascular dilatation, and eventually demyelination.⁴⁵

In an analysis by Kikuta et al. comparing older hypertensive patients with age-matched controls, the ALPS-index in the hypertensive group was significantly lower than in the control group. In addition, ALPS-index of all subjects was significantly negatively correlated with blood pressure values and pulse pressure values. This suggested that SVD due to hypertension was associated with damage in cerebral ISF dynamics.⁴³ In a study of 133 patients with SVD, a lower ALPS-index was independently associated with impaired executive, attention, and memory function in patients with SVD.⁴⁶ In a study of a cohort of more than 2000 community residents, the lower the ALPS-index, the greater the presence and severity of cerebral SVD. Furthermore, a low ALPS-index was associated with poor neuroimaging markers such as abnormal white matter signal, enlarged PVS, and brain atrophy.⁴⁷

Cerebral amyloid angiopathy (CAA) is a type of cerebral SVD characterized by deposition of Aβ in the middle and outer membranes of arteries in the cerebral cortex and meninges. The underlying etiology of Aβ accumulation is still unknown. In addition to the conventional idea of Aβ overproduction, there is also the idea of impaired Aβ clearance pathways. In a comparison of CAA patients and age-matched controls, CAA patients had an overall lower ALPS-index compared to controls. In addition, lower ALPS-index was associated with enlarged PVS in the basal ganglia, increased lacunes, higher white matter high intensity Fazekas scores, lower MMSE, and lower Montreal Cognitive Assessment (MoCA-J). During a median follow-up of 4.1 years, there was less recurrence of disease with a higher ALPS-index.⁴⁸

Application for idiopathic normal pressure hydrocephalus

Idiopathic normal pressure hydrocephalus (iNPH) is a clinical syndrome with enlarged ventricles without increased CSF pressure. iNPH has three characteristic symptoms: gait disturbance, cognitive decline, and urinary incontinence. Imaging features include Disproportionately Enlarged Subarachnoid Space Hydrocephalus (DESH) and narrowing of the callosal angle.^{11,12,49–51} iNPH has been pointed out as a degenerative disease, and its co-morbidity with AD is attracting attention.^52,53 Eide et al. reported the evaluation of CSF dynamics and ISF dynamics in iNPH group using a technique with GBCA intrathecally injected and MRI observation over time.⁵⁴ In this invasive study, the shunt-responsive iNPH group showed more conspicuous contrast reflux into the ventricular system and a tendency toward prolonged parenchymal enhancement compared to the control group. These results suggest that abnormalities in the glymphatic system may be related to the pathogenesis of iNPH.⁵⁵

The study by Yokota et al. on iNPH cases is the first follow-up study of the ALPS method.⁵⁶ This study compares a group of iNPH cases diagnosed according to diagnostic criteria with a group of pseudo-iNPH (piNPH) cases that do not meet the diagnostic criteria. Their results showed that the ALPS-index was significantly lower in iNPH cases, and the area under the curve in ROC analysis was 0.92, 1.00, and 1.00 for control vs. piNPH, control vs. iNPH, and piNPH vs. iNPH, indicating very high diagnostic performance. Between piNPH and NPH, the ALPS-index showed higher diagnostic performance than the Evans index or the callosal angle. In a study by Georgiopoulos et al. the association between the clinical findings of Time Up and Go test (TUG) and MMSE associated with iNPH and the ALPS-index was analyzed. Patients with iNPH had significantly lower mean ALPS-index than healthy control. A moderate exponential correlation was found between the mean ALPS-index and motor function as measured by time required to complete TUG, number of steps to complete TUG, 10 m walking time, and 10 m walking steps. A positive linear correlation was found between mean ALPS-index score and MMSE score.⁵⁷ As a diagnostic tool for iNPH, tap-test is an important test for predicting shunt efficacy.^58–60 In a study evaluating the relationship between Tap-test findings and ALPS-index, ALPS-index was significantly lower in the NPH group that did not respond than in the group that showed symptom relief.⁶¹ Shunting is an effective treatment for iNPH. Studies have reported changes in ALPS-index before and after shunting. The mean ALPS-index after shunt surgery was higher than before shunt surgery. In particular, there was a significant increase in postoperative ALPS-index in the group with symptomatic improvement after shunt surgery, whereas there was no significant increase in postoperative ALPS-index in the group without symptomatic improvement (Fig. 4).⁶²

Fig. 4.

Application of ALPS method in iNPH. A study of changes in ALPS-index after LPS surgery in iNPH is presented, comparing patients who responded to LPS with those who did not. The mean ALPS-index of the postoperative group was significantly higher than that of the preoperative group (A). Furthermore, for responding subjects, the mean ALPS-index in the postoperative group was significantly higher than in the preoperative group (B). On the other hand, in the non-responder group, the mean ALPS-index of the postoperative group was not significantly higher than that of the preoperative group (C). The mean ALPS indices of the responder group were not significantly different compared to those of the non-responder group in both the pre-operation and post-operation groups (D). Figure quoted from Ref. 62. iNPH, idiopathic normal pressure hydrocephalus; LPS, lumboperitoneal shunt.

Application for traumatic brain injury

Traumatic brain injury (TBI) is an established risk factor for early onset of dementia, including AD, and the post-traumatic brain often shows neurofibrillary changes consisting of aggregates of the protein tau.⁶³ In animal studies in mice using fluorescent tracers, function of the glymphatic pathway was reduced by approximately 60% following TBI, and this impairment persisted for at least 1 month after injury.⁶⁴ In addition to direct brain injury, animal studies in non-human primates have shown that the formation of a CSF leak delays the entry of intrathecal injected GBCA into the brain.⁶⁵ This indicates that mechanical factors such as intracranial pressure environment affect cerebral ISF dynamics.

In humans, several studies using the noninvasive ALPS method have reported abnormal ISF dynamics after TBI.^66–68 It has been reported that ALPS-index is decreased in the brain after TBI.^66,68 On the other hand, there are reports that ALPS-index is elevated in young patients with mild TBI.⁶⁷ The possible causes include increased expression of AQP4 after trauma, hemodynamic changes after trauma, and the possibility that the glymphatic system is a compensatory mechanism to reduce secondary injury by eliminating endotoxin after trauma. As a situation that does not develop into a traumatic injury, some reports have evaluated the ALPS-index in people who have experienced contact sports. In this report, the left ALPS-index was significantly lower in the heavy contact (rugby, judo, karate, boxing, kendo, wrestling, and soccer) and semi-contact (baseball, basketball, and handball) groups than in the non-contact (tennis, table tennis, track and field, skiing, archery, and orienteering) group. The right ALPS-index tended to be lower in the semi-contact and heavy contact groups than in the non-contact group. Bilateral ALPS indices showed a significant positive correlation with MoCA-J scores.⁶⁹

Application for demyelinating disease

Multiple sclerosis is a chronic immune disorder of the CNS characterized by inflammation and demyelination caused by the interaction of BBB disruption, local leukocyte infiltration, microglial activation, and release of inflammatory cytokines. Considering that toxic wastes are the cause of tissue damage, it seems essential for the brain to have a system to remove these wastes.^70,71 A retrospective study using the DTI-ALPS technique in patients with multiple sclerosis is reported. In this study, multiple sclerosis patients had an overall lower ALPS-index compared to healthy controls. Both relapsing-remitting and progressive multiple sclerosis patients had a lower ALPS-index versus healthy controls. Patients with advanced multiple sclerosis had a lower ALPS-index compared to patients with relapsing-remitting multiple sclerosis. In multiple sclerosis patients, lower ALPS-index was associated with more severe clinical disability and longer disease duration.⁷⁰

Myelin oligodendrocyte glycoprotein antibody disorder (MOGAD) patients, another demyelinating disease, have also been reported to be studied using the ALPS method. In a prospective study of MOGAD patients in remission and age- and sex-matched healthy controls, the mean ALPS-index was significantly lower in MOGAD patients than in healthy controls. Lower mean ALPS-index was significantly associated with worse Expanded Disability Status Scale scores. However, cortical and deep gray matter volumes were not significantly different between the two groups, nor were they correlated with Expanded Disability Status Scale scores.

It is suggested that MOGAD patients may have impaired glymphatic function as measured by the ALPS-index.⁷²

Application for sleep

One of the reasons why the glymphatic system theory has attracted so much attention is its relationship to sleep. Drainage by the glymphatic system is considered to be suppressed during wakefulness and markedly enhanced during sleep. Xie et al. reported that under physiological conditions, the activity of the glymphatic system is associated with sleep, based on observations of fluorescent tracers injected into the CSF cavity of rats using two-photon microscopy. During sleep, there is a significant increase in glymphatic system activity compared to wakefulness. They attributed this to the fact that during sleep, the volume of glial cells is reduced and the interstitial space is more dilated than during wakefulness, facilitating mass transport in the tissues.⁷³

The ALPS method has been used to evaluate sleep and cerebral ISF dynamics in several published studies.^74–78 In a study evaluating the correlation between sleep-related indices and the ALPS-index in community-dwelling older adults aged 60 years and older, the ALPS-index correlated with N2 sleep duration and N3 sleep duration. On the other hand, it correlated inversely with the apnea-hypopnea index. In addition, it correlated with gray matter volume.⁷⁴ Obstructive sleep apnea (OSA) is associated with sleep fragmentation and altered sleep organization, which may impair cerebral ISF dynamics and increase the risk of AD. In a study of OSA and control cases, Dzz values, derived from projection fiber areas, Dyy and Dzz values from association fiber areas, as well as ALPS and Dyzmean values were significantly reduced in OSA over controls. Significant correlations emerged between disease severity, sleep symptoms, and Dxy, Dxx, and Dzz values in OSA subjects.⁷⁵ In a study evaluating the association between sleep quality and the ALPS-index as assessed by the Pittsburgh Sleep Quality Index (PSQI), the ALPS-index was significantly lower in the sleep disordered group than in the control group. In addition, the ALPS-index showed a significant negative correlation with all components of the PSQI score.⁷⁶ In the DTI-ALPS study of narcolepsy patients, there was no significant difference in ALPS-index between the patient and control groups. However, the ALPS-index was correlated with polysomnography parameters and positively correlated with wake after sleep onset and REM sleep length in patients with type 2 narcolepsy.⁷⁷ As mentioned above, animal experiments with tracers have shown that tracer dynamics are enhanced during sleep. In humans, evaluation by the dual compartment tensor model has also been reported in diffusion imaging. The dual compartment analysis revealed that the increase in diffusivity measures from PM to AM was driven by an increase in the volume fraction of CSF-like free-water.⁷⁹ On the other hand, in the ALPS method study, MRI measurements were repeated at five time points from 8:00 to 23:00 and no time-period dependence was observed in the awake state in an experiment to evaluate circadian rhythm dependence (Fig. 5).⁷⁸ This study is also discussed in “Other methods for evaluation of glymphatic system” and “Consideration for the methodology of DTI-ALPS” sections.

Fig. 5.

Time-of-day dependency of ALPS-index. Circadian rhythm dependence of ALPS-index was assessed by repeated MRI measurements at five time points from 8:00 to 23:00. Very long echo-time low-b diffusion tensor imaging (DTIlow-b), which measures SAS flow along the middle cerebral artery, and conventional DTI-ALPS were performed. A, B show the time course of the glymphatic system influx and efflux measured by DTIlow-b and DTI-ALPS, respectively. Neither ADlow-b of the MCA SAS fluid nor the ALPS-index differed significantly in the data acquired at the five time points. Figure quoted from Ref. 159. MRI, magnetic resonance imaging; DTI, Diffusion Tensor Image; SAS, subarachnoid space; DTI-ALPS, Diffusion Tensor Image Analysis aLong the Perivascular Space; AD, Alzheimer’s disease; MCA, middle cerebral artery.

Findings on the aging

A number of studies using the DTI-ALPS method have evaluated the association between age and ALPS-index.^{4,17,18,27–29,36,43,56,61,70,74,78,80–92} Among these reports, a number of them showed negative correlations between age and ALPS-index. McKnight et al. showed a negative correlation between the ALPS-index and age in a group of subjects with PD or essential tremor aged approximately 40−80 years, with a correlated Spearman’s rank correlation coefficient of −0.203.²⁷ Zhang et al. showed a negative correlation between age and ALPS-index in 142 healthy elderly subjects aged 50 years and older, with a correlation coefficient of −0.209.⁸⁵ Ma et al. showed a negative correlation between age and ALPS-index in PD cases, with a standard partial regression coefficient of −0.012.²⁸ Siow et al. also found a negative correlation between age and ALPS-index in community-dwelling elderly aged 60 years and older, with a standard partial regression coefficient of −0.035.⁷⁴ Toh and Siow examined glioma cases in a wide age range, from 18 to 91 years of age. Although the correlation coefficient is low at −0.147, the graph they presented showing the correlation between age and ALPS-index appears to have a slight peak around age 40.⁹³ The public dataset study by Dai et al. for ages 20−87 also appears to have a similar peak around the 40s in the graph provided in the paper showing the correlation between age and ALPS-index, although the authors do not mention it.⁹⁴ Furthermore, the DTI-ALPS method described below also shows that the ALPS-index peaks in the 40s.⁹⁵ However, there is an aspect of this that indicates that the ALPS-index may be influenced by the status of white matter fibers, which will be discussed later. While there have been many reports of an association with aging in older patients, there have also been reports of the DTI-ALPS method in pediatric cases.^96–98 Lee et al. showed a negative correlation between age and ALPS-index with a correlation coefficient of −0.375 in a group of young patients with myoclonic epilepsy aged 12−46 years.⁸⁷ However, there is a lack of validation of the applicability of this technique in children, in whom the construction of white matter fibers is incomplete. The rapid changes in the developmental state of white matter with growth require careful interpretation of the results, especially when comparing cases of different ages. Further experience is needed.

Other Methods for Evaluation of Glymphatic System

Because the ALPS-index is given by a simple formula and can be calculated from retrospective data, it has been reported on many diseases, pathologies, and conditions as described above. However, as discussed in Section “Controversy in DTI-ALPS method”, there are also problems that have been pointed out.^99–102 When evaluating cerebral ISF dynamics, it is not sufficient to use the ALPS method alone, instead it is desirable to use other methods of evaluating cerebral ISF dynamics together. The following is a review of studies that have evaluated the relationship between the ALPS method and other methods of assessing cerebral ISF dynamics.

Intrathecal GBCA

A number of attempts have been reported to use MRI to evaluate the state of the glymphatic system (Table 1). However, the first studies on the glymphatic system by Iliff et al. were done in animals with intrathecal administration of fluorescent tracers and two photon laser microscope.¹ This method has the advantage of allowing direct observation of the living brain, but only near the surface of the brain. As a method to observe the whole brain, a tracer study in animals using MR with intrathecal GBCA was introduced. In humans, tracer studies using MR with intrathecal GBCA are considered to be the gold standard. However, it should be noted that the molecular weight of GBCA is 0.6 kDa (Gadobtrol), whereas that of Aβ in the brain is 100−200 kDa.¹⁰³ In addition, in that it evaluates penetration into the brain from the brain surface, it is evaluating the pathway opposite to the drainage of waste material from the brain parenchyma. Furthermore, it is difficult to evaluate the state of the deep white matter, which occupies a significant proportion of the brain volume. In these respects, it should be noted that this method is too limited to be considered a gold standard for evaluating the glymphatic system.

Table 1.

MRI methods intending to evaluate glymphatic system for human subjects

Method	MR sequence	Target phenomenon	Target location	Assumed rolls in glymphatic system
Intrathecal GBCA	T1WI	Penetration via the brain surface (Tracer: GBCA)	Brain surface	CSF influx to brain parenchyma
Intravenous GBCA	T1WI	BBB permeability (Tracer: GBCA)	Brain parenchyma	ISF production
		Venous wall permeability (Tracer: GBCA)	Surface vein	Movement of waste products
		Tracking of GBCA	Meningeal lymphatic vessel/ Perivascular space	Movement of waste products
Intravenous H217O	Steady-state sequence	BBB permeability (Tracer: labeled water)	Brain parenchyma	ISF production
DTI-ALPS	DTI	Brownian motion of the water molecule in the brain tissue	Deep white matter adjacent the lateral ventricle	Movement of ISF within the tissue
Free water analysis	DTI etc.	Increase of free water fraction	Brain parenchyma	Widening of interstitial space
Perivascular space volume	3D-T2WI etc.	Dilatation of perivascular space	Perivascular space	Accumulation of metabolic waste
Choroid plexus volume	3D-T1WI etc.	Enlargement of choroid plexus	Choroid plexus	Removal of waste/ Immune reaction etc.
ASL with long TI	ASL	BBB permeability (Tracer: tagged water)	Brain parenchyma	ISF production
BOLD-CSF coupling	BOLD	Synchrony of the neural activity and the brain/CSF motion	Brain parenchyma	Tissue microenvironment
Elastography	Elastography	Viscoelastic properties of the brain	Brain parenchyma	Tissue microenvironment

ASL, arterial spin labeling; BBB, blood -brain barrier; BOLD, blood oxygen level-dependent; CSF, cerebrospinal fluid; DTI, diffusion tensor image; DTI-ALPS, Diffusion Tensor Image Analysis aLong the Perivascular Space; GBCA, gadolinium-based contrast agent; ISF, cerebral interstitial fluid; T1WI, T1-weighted image; T2WI, T2-weighted image; TI, inversion time.

Intrathecal injection of GBCA in humans is an unapproved use, and few centers perform it. This is because GBCA intrathecal injection is invasive and, depending on the dose, unsafe. Eide, Ringstadt, and colleagues have systematically conducted tracer studies of GBCA intrathecal injection in humans. The first report was of intrathecal GBCA in a patient with intracranial hypotension due to spontaneous CSF leakage, in which MRI showed distribution of GBCA throughout the brain at 1 and 4.5 hours after intrathecal injection.¹⁰⁴ Their experiments were initially only qualitative evaluations of tracers, but more recent studies have conducted sophisticated mathematical analyses to identify and quantify CSF tracer transport parameters. The results indicate that extracellular diffusion alone is not sufficient as a tracer transport mechanism for the whole brain, and that local clearance rates contribute to the tracer transport.¹⁰⁵ In a study by Dyke et al. in healthy adult volunteers, GBCA in the intracranial spinal fluid space peaked at 1−3 h and began to decrease at 7 h. In contrast, in some areas of the brain parenchyma, such as the cerebral cortex and white matter, potentiation increased after 11 hours. Thus, it was suggested that the enhancement of the whole brain MR signal by intrathecal GBCA administration may indicate CSF/ISF exchange and brain parenchyma penetration throughout the brain.¹⁰⁶

A study compared the results of the DTI-ALPS method with the evaluation of glymphatic system function using intrathecal GBCA. The ALPS-index was significantly associated with the rate of change in signal unit ratio from baseline to 39 h in six brain regions.⁸⁵

Intravenous GBCA

Intravenous GBCA infusion as a tracer is easier to perform than intrathecal injection. Experiments using high-dose intravenous GBCA in rats showed an increase in signal in the fourth ventricle immediately after intravenous GBCA injection, indicating that GBCA is rapidly transferred from blood to CSF in rats.¹⁰⁷ The signal curve of the cerebral cortex and deep cerebellar nuclei reached the peak signal intensity later than the fourth ventricle but earlier than that of the prepontine cistern. Thus the GBCA distribution to the cerebral cortex and deep cerebellar nuclei seemed to depend on both blood flow and CSF. In human evaluations, the transfer of IV-GBCA into the CSF is more gradual. In human evaluation after administration of a normal dose of GBCA, leakage from the cortical veins into the CSF was observed. An enhancement within the PVS in the basal ganglia and around the cortical veins was observed after 4 h of IV-GBCA administration.^108–112 Interestingly, the leakage from the cortical vein was not observed in the younger subjects, but only in the older subjects (37 years and older).¹⁰⁹ In a study examining the relationship between the number and size of cystic structures around the cortical veins near the superior sagittal sinus and leakage of GBCA around the cortical veins, the number and size of depressed structures near the superior sagittal sinus were reported to be larger in subjects with GBCA leakage into the SAS than in those without leakage.¹¹³ Intravenous GBCA can also delineate meningeal lymphatic system. In a study of the features of presumptive meningeal lymphatic vessels located in the posterior wall of the sigmoid sinus (PML-PSS) in human subjects imaged after intravenous administration of GBCA, PML-PSS was found in 23 of 42 patients 4 hours after IV-GBCA was performed. The study also simultaneously examined the enhancement of the basal ganglia perivascular space (PVS-BG), and PVS-BG was seen in 21 of the 42 cases.^114,115

There have been several trials using intravenous GBCA to assess the degree of leakage of the BBB. In a study that used a Patlak graphical approach to quantify BBB leakage rates and regional plasma volume on dynamic contrast images after intravenous GBCA injection, the volume fraction of brain tissue with GBCA leakage was significantly higher in AD.¹¹⁶ In a study examining changes in GBCA leakage during the diurnal cycle, T1 mapping of gray matter and white matter at 0.5, 1, 1.5, 2, and 12 h after intravenous contrast injection was used. The results showed that signal changes in cerebral gray matter, cerebellar gray matter, and putamen increased from diurnal to nocturnal cycles. This suggests that the clearance of GBCA is greater after sleep than during daytime wakefulness.¹¹⁷ This technique has the potential to evaluate BBB leakage. However, BBB leakage from GBCA in healthy brain tissue is so small that the measurement is always evaluated near the noise floor. In this respect, the measurements are expected to be greatly affected by noise contamination, imaging conditions, and analysis methods. It should also be noted that the evaluation of BBB leakage does not evaluate the entire glymphatic system, but only a part of the entire system.

Currently, there are no published reports comparing the assessment of the extent of BBB leakage by intravenous contrast with that by the ALPS method. However, a study has been reported that evaluated the time course of enhancement in the parasagittal sinus using dynamic contrast-enhanced MRI with intravenous GBCA for the purpose of assessing meningeal lymphatic dysfunction. Time to peak values after contrast was significantly delayed in the chronic migraine group compared to the episodic migraine group, suggesting meningeal lymphatic dysfunction. This study also reported significantly lower ALPS-index in the chronic migraine group compared to the episodic migraine group.¹¹⁸

Perivascular space volume

Needless to say, the PVS is an anatomical structure that plays an important role in the integrity of the glymphatic system. The PVS is usually unremarkable on MR images, but in dilated cases of PVS, it can be seen as a punctate or linear structure.^119,120 This abnormal dilation of the PVS is hypothesized to be associated with impaired glymphatic function.^121,122 One suggestion about the mechanism is that PVS volume may reflect decreased flow in the glymphatic system resulting in either increased accumulation of metabolic waste products or failure to deliver needed metabolic substrates.¹²³ Recently, there have been several published reports of attempts to actively apply PVS enlargement as a biomarker of glymphatic system damage.^123–127

Various methods have been proposed to evaluate the degree of PVS dilation on MRI. In general, the easiest to use is the visual evaluation of T2-weighted images. Potter et al. proposed a robust and easy-to-use PVS assessment scale to effectively examine the diagnostic and prognostic significance of PVS. They rated basal ganglia, centrum semiovale, and midbrain PVS. Basal ganglia and centrum semiovale PVS were rated 0 (none), 1 (1−10), 2 (11−20), 3 (21−40), and 4 (>40), and midbrain PVS were rated 0 (none visible) or 1 (visible). They report that the ratings based on this criterion showed good intra- and inter-rater consistency.¹²⁰ Recently, an automated human PVS evaluation system using MRI has been developed. Dubost et al. proposed a convolutional network regression method to quantify the degree of enlarged PVS in the basal ganglia from 3D brain MRI. This method achieved an intraclass correlation coefficient of 0.93 and an intra-rater agreement coefficient of 0.80.¹²⁸ It is known that long-term spaceflight can cause changes in the brain and CSF, resulting in visual impairment called spaceflight-related neuro-ophthalmopathy. Astronauts who spent 6 months on the International Space Station showed enlarged PVSs in the basal ganglia and white matter after their spaceflight. Furthermore, astronauts who developed spaceflight-related neuro-ophthalmopathy had a more enlarged PVS in the white matter than those who did not develop the syndrome, indicating that long-term changes in gravity may affect the function of the glymphatic system.¹²⁹ Examination of T2-weighted images of postmortem brains shows a correlation between the deposition of Aβ in cortical vessels and the degree of opening of the PVS in the underlying white matter, indicating that the degree of opening of the PVS may be an indicator of dysfunction of the glymphatic system.¹²⁶

Several studies comparing the evaluation by PVS volume with that by the ALPS method are presented. In a study of TBI cases and controls, the ALPS-index was significantly lower in subjects with TBI compared to controls when age was used as a covariate. The ALPS-index showed a significant negative correlation with the plasma concentration of neurofilament light chains, a biomarker of injury severity. On the other hand, PVS volume did not differ in TBI patients compared to controls, nor did it correlate with the plasma concentration of neurofilament light chains.¹³⁰

In a study of patients with AD, the perivascular volume fraction of the basal ganglia in patients with AD was significantly larger than the volume fraction in normal controls. The ALPS-index showed a significant negative correlation with the perivascular volume fraction of the basal ganglia in AD patients.²¹ As mentioned in section “Application for Alzheimer’s disease”, using the AD Neuroimaging Initiative database, an examination of measurements such as the PVS volume fractions, fractional volume of FW-WM, and ALPS-index has been reported. In this study, patients with AD had significantly higher PVS fraction of the basal ganglia, significantly higher fractional volume of FW-WM, and significantly lower ALPS-index compared to healthy controls (Fig. 2).²⁰ In addition, Zhang et al. also reported that the PVS fraction in the basal ganglia of AD patients was significantly larger than the PVS fraction in control subjects and negatively correlated with the DTI-ALPS-index.²¹ A report has also evaluated the relationship between ALPS-index and PVS in patients with PD. The left DTI-ALPS-index showed a non-significant negative trend with the number of PVSs in the left basal ganglia in the early PD group, and a significant negative correlation in the late PD group.³⁶

Choroid plexus volume

The choroid plexus is a highly vascularized secretory tissue located within the ventricles of the brain. Its main functions have traditionally been considered to be the production of CSF and the formation of the blood-CSF barrier. However, recently, functions such as removal of toxic waste products and metabolites from the CNS, immune surveillance of the brain, and regulation of CSF composition for homeostatic brain function have been identified.¹³¹ In a study of choroid plexus volume and dopamine transporter (DAT) scanning findings in early, untreated PD patients, choroid plexus volume was negatively correlated with DAT availability in many striatal subregions. In the Cox regression model, larger choroid plexus volume was associated with future onset of gait freezing, and in the linear mixed model, it was associated with more rapid increases in dopaminergic medications. This suggests that choroid plexus volume may serve as a biomarker for baseline and longitudinal motor deficits in PD.¹³² It is an attractive hypothesis that the PVS must be dilated by a disturbance of the glymphatic system. At present, however, direct proof is lacking. The PVS may also be dilated by other factors, such as a decrease in brain parenchymal volume. However, it is difficult to distinguish PVS dilatation due to other factors from PVS dilatation due to a glymphatic system disorder. In this regard, one should be cautious in linking PVS dilation directly to glymphatic system disorder.

Several studies have been presented that examined choroid plexus volume and its evaluation with the ALPS technique. In a study in which glymphatic clearance and choroid plexus volume were measured by 3D-T1-weighted imaging before and 39 h after intrathecal contrast administration in patients indicated for lumbar puncture, higher choroid plexus volume correlated with slower glymphatic clearance rates in all brain regions. The higher choroid plexus volume was correlated with slower glymphatic clearance in all brain regions. In this paper, longitudinal follow-up observations of ALPS-index, choroid plexus volume, and abnormal white matter signal volume are conducted in a different cohort. The results showed that baseline choroid plexus volume was positively correlated with abnormal white matter signal volume and its degree of increase. Furthermore, in a mediation analysis, the ALPS-index partially mediated the association between choroid plexus volume and abnormal white matter signal-like content and its augmentation.¹³³ Fibromyalgia is often associated with various degrees of sleep disturbance and cognitive impairment, and studies have evaluated the relationship between sleep disturbance and glymphatic system function in these patients. The results showed that fibromyalgia patients had significantly higher choroid plexus volume and lower ALPS-index than age- and intracranial volume-adjusted controls. Increased choroid plexus volume was correlated with lower ALPS-index and longer disease duration.¹³⁴ The complex geometry of the choroid plexus has made volume measurement cumbersome in the past, but recently, automated measurement using deep learning models has been introduced, showing comparable results in choroid plexus delineation on T1-weighted, T2-weighted and FLAIR non-contrast MRI images.¹³⁵ It is expected that the measurement of choroid plexus volume will become easier in the future by using these software.

Diffusion-based methods other than ALPS method

In addition to the ALPS method discussed so far, several other methods for evaluating cerebral ISF dynamics using diffusion images have been introduced. The brain contains a mixture of both free water and brain tissue, namely CSF and ISF.

The ALPS method focuses on the directionality of diffusion without considering these compartments of diffusion. On the other hand, there is a method that evaluates the free water compartment including CSF and ISF within the diffusion compartments. One of the methods is the dual compartment tensor model. Studies have used this method to examine the effect of time of day on diffusion tensor measurements using a within-subject longitudinal design. In this study, a dual-compartment tensor model, which allows direct assessment of the free water volume fraction, was used to evaluate changes related to time of day, in addition to the conventional monoexponential tensor model. The results indicate that the mean diffusivity measured using the conventional single-exponential tensor model tends to increase systematically at the gray matter/CSF interface from the morning to the afternoon scan. Furthermore, the dual-compartment analysis revealed that the increase in diffusivity measurements from the afternoon to the morning is caused by an increase in the volume fraction of CSF-like free water.⁷⁹ As mentioned in “Application for Alzheimer’s disease” section, free water analysis was also performed in AD, and the free water fraction of white matter was higher in the AD group than in the normal group (Fig. 2).²⁰ Alterations in free water fraction may be related to glymphatic system function. However, the information is static, and there is no direct evidence as to how the increase or decrease in free water fraction is dynamically related to the process of waste excretion in the brain. This approach also does not seem to provide an assessment of the entire glymphatic system.

One of the analyses of CSF dynamics using diffusion imaging is the evaluation of CSF pulsation. The early diffusion-weighted image (DWI) by LeBihan et al. was called IVIM (intravoxel incoherent motion) imaging. The b-value used at that time was low, about 100−200 s/mm², thus the signal from CSF remained.^136,137 Later, when motion probing gradients with high b-values became available, the CSF signal in DWI was almost completely suppressed and was not considered for evaluation. A report was published in 2019 that once again attempted to evaluate CSF by daring to use a low b-value DWI.¹³⁸ Subsequently, reports have been published on the evaluation of CSF dynamics using multiple b-values, as well as reports evaluating the relationship between CSF pulsation and arterial pulse rate, and more detailed evaluation of spinal fluid pulsation with the addition of tensor information.^139–141 Assessment by measuring flow in the SAS along the middle cerebral artery by ultra-long echo time and low-b diffusion tensor imaging (DTIlow-b) has also been attempted; DTIlow-b has been proposed to reflect flow characteristics in one of the major CSF inflow pathways to the brain parenchyma (Fig. 5). The influx rate of the glymphatic system evaluated by DTIlow-b was significantly higher in participants aged >45 years than in participants aged 21–38, while the ALPS-index was significantly lower in those aged >45 years.⁷⁸ Of course, since this method evaluates CSF outside of the brain, it does not evaluate the entire glymphatic system.

Arterial spin labeling

Arterial spin labeling (ASL) is one of the most commonly used noninvasive perfusion imaging methods in routine clinical practice. Recently, a technique to evaluate AQP4 water channels using this ASL has been introduced. A multi-echo time (multi-TE) ASL MRI technique was applied to the mouse brain to evaluate the water permeability of the BBB by calculating the exchange time, which is the time that magnetically labeled intravascular water is exchanged across the BBB. The results showed a 31% increase in exchange time (452 ± 90 ms) in AQP4-deficient (Aqp4−/−) mice (P = 0.01) compared to wild-type mice (343 ± 91 ms), indicating the sensitivity of this technique to the absence of AQP4 water channels.¹⁴²

Human evaluations have also been performed. Using 3D turbo gradient and spin echo pulsed arterial spin echo with long TI (inversion time), the mean arterial pulse-corrected signal versus TI was plotted for each TI, and the slope of the plot was used as the clearance rate. Comparison between healthy subjects and AD patients showed a statistically significant clearance rate in the AD group. In the AD group, brain clearance rate correlated with CSF clearance rate, indicating that a proportion of ultra-filtered labeled blood was retained in the brain. This provides indirect evidence that perivascular clearance is reduced in AD.¹⁴³ This method uses water molecules that can pass through the BBB under physiological conditions as tracers, rather than large molecules as in GBCA, and when combined with the GBCA evaluation, it should be possible to evaluate ISF dynamics separately for solute and solvent. Of course, this method is also a part of the glymphatic system.

Imaging for brain tissue kinetics

Kiviniemi et al. suggested that low-frequency (<0.1 Hz) resting-state fMRI blood oxygen level-dependent (BOLD) signals are associated with spinal fluid dynamics and thus glymphatic system function.¹⁴⁴ It has also been reported that the large global BOLD signal during sleep is coupled with strong CSF motion, an important component of the glymphatic system.¹⁴⁵ Taken together, the overall resting activity and associated physiological modulations may represent highly coordinated neural and physiological processes closely related to glymphatic system function. BOLD-CSF coupling assesses these synchrony and may serve as a marker for measuring glymphatic function in the brain.¹⁴⁶

In a study in PD, BOLD-CSF coupling was found to be significantly lower in PD patients with MCI compared to those without MCI or controls. The decrease in BOLD-CSF coupling was associated with the cognitive decline seen in patients with PD. Additionally, lower BOLD-CSF coupling in PD patients was associated with a thinner right entorhinal cortex.¹⁴⁶ Glymphatic system dysfunction may also be involved in the pathogenesis of frontotemporal dementia. In a study of behavioral variant frontotemporal dementia (bvFTD), patients with bvFTD exhibited significantly weaker BOLD-CSF coupling. The study also found a negative correlation between the ALPS-index and BOLD-CSF coupling, as well as a negative correlation between the ALPS-index and CP volume.¹⁴⁷

Magnetic resonance elastography (MRE) is an approach that can provide biomechanical information about the tissue microenvironment because it can evaluate the viscoelastic properties of tissues in vivo.¹⁴⁸ There have been reports of attempts to assess glymphatic system function using MRE. Experiments in a mouse model showed that shear wave speed, a measure of stiffness, varied by approximately 12% across the cortex between sleep and waking states.¹⁴⁹ Also in humans, MRE has been reported for the purpose of evaluating the glymphatic system, and multivariate analysis showed that the ALPS-index was independently associated with the complex shear modulus and positively related to it.¹⁵⁰

Other tracer study methods

Studies to date using large tracer molecules such as GBCA may have systematically underestimated the aqueous flow of subarachnoid spinal fluid into the brain. The stable ¹⁷O isotope is not radioactive but has a quadrupolar nucleus that can be imaged by MRI. Direct imaging of the ¹⁷O signal is difficult due to its low gyromagnetic ratio and low natural abundance (0.037%), but the presence of H₂¹⁷O as a tracer can be detected indirectly by the effect of the quadrupole ¹⁷O nucleus on the 1 H MRI signal. Glymphatic flow imaged with the H₂¹⁷O tracer has been demonstrated to be more rapid and extensive than when imaged with the GBCA tracer.¹⁵¹

Animal studies have been reported using the AQP-4 accelerator TGN-073 and H₂¹⁷O JJ vicinal coupling proton exchange MRI, which can track water molecules delivered into the blood circulation. The results showed that the AQP-4-promoting effect of TGN-073 increased ISF turnover through the AQP-4 system, resulting in a significant decrease in H₂¹⁷O content in the cerebral cortex.¹⁵² Also in humans, attempts have been made to use H₂¹⁷O as a tracer, and dynamic 3D-FLAIR has been shown to detect dose-dependent signal changes.¹⁵³

Dynamic PET has been attempted to be used for evaluation of the glymphatic system. A study using dynamic PET with 18 F-THK5117, a tracer of tau pathology, to calculate ventricular CSF clearance rate (vCSF) as a biomarker of waste excretion has been reported.

This study focused on the washout of tau tracer. The vCSF was inversely correlated with amyloid deposition in AD patients and normal controls. In all subjects, the highest extracranial accumulation of tracer was in the nasal turbinates. In patients with AD, vCSF was reduced by 23% and the supranasal CSF efflux site was reduced by 66%.¹⁵⁴ There is a report comparing vCSF and ALPS-index with amyloid deposition in the brain using this technique. vCSF showed a high correlation (r = −0.548) with mild Aβ deposition, while DTI-ALPS showed a high correlation (r = −0.451) with severe Aβ deposition. Regression models using both vCSF and DTI-ALPS are better associated with brain Aβ deposition. Thus, these two independent brain clearance indices may better explain the variability of Aβ deposition when both are used rather than either term alone. In other words, it suggests that vCSF and DTI-ALPS reflect complementary aspects of brain clearance function.¹⁵⁵

Consideration for the Methodology of DTI-ALPS

Placement of ROI

In the DTI-ALPS method, it is necessary to place ROIs in the projection fiber region and the association fiber region. This ROI setting is, in a sense, the Achilles’ heel of the DTI-ALPS method, because the reproducibility is greatly affected by the ROI setting.⁸⁶ The most important point in the placement of the ROI is to ensure that all of the subject groups of cases are measured in a uniform manner. It is also important to make sure that the projection and association fibers are perpendicular to the x-axis. In particular, when setting the ROI in the area of association fibers, care should be taken to avoid mixing of subcortical fibers whose main axis is in the x-direction. If subcortical fibers are included, the ALPS-index will be higher than it should be.

In the first paper on the DTI-ALPS method, a 5 mm diameter circular ROI was manually placed in the left projection and association fiber areas (Figs. 1 and 6). Many papers have followed this method of placing ROIs in the dominant brain.^{28,32,43,46,48,74,80,89,97,122,133,150,156–158} Since association fibers are often thicker on the dominant side (left side), placement of a ROI on the left side has the advantage that a stable ROI can be set, even if it is done manually. In many recent papers, the average of the ROI on both sides is used.^{18,26,30,34,57,62,67,68,72,78,82,86,159–164} This method is expected to have the potential to cancel out the effects of head tilt and other factors. There is also a paper that examines the ALPS-index calculated for the right hemisphere and the ALPS-index calculated for the left hemisphere separately.^{16,19,69,165,166}

Fig. 6.

Placement of ROI for ALPS method. The most important point in the placement of the ROI is to ensure that all of the subject groups of cases are measured in a uniform manner. It is also important to make sure that the projection and association fibers are perpendicular to the x-axis. In particular, when setting the ROI in the area of association fibers, care should be taken to avoid mixing of subcortical fibers whose main axis is in the x-direction. If subcortical fibers are included, the ALPS-index will be higher than it should be. After that, there are various ways to place ROIs: only on the left side of the dominant hemisphere (a),²⁸ bilateral placement (b),³⁰ Atlas-based placement (c),⁵⁶ and placement with conversion to the standard brain (d).¹⁶⁰

Some studies employ atlas based placement.^56,147 Yokota et al. coregistered diffusion maps to the DTI-81 atlas of the International Consortium for Brain Mapping (ICBM) and established ROIs for projection and association fiber regions. In their study, the atlas-based ALPS-index was able to distinguish iNPH from pseudo-iNPH more clearly than manual ROI placement⁵⁶ (Fig. 6).

An increasing number of recent studies employ automated ROI setup. Most of the studies use the Montreal Neurological Institute (MNI) 152 template or the digital WM atlas JHU ICBM-DTI-81 (http://cmrm.med.jhmi.edu/) for standardization purposes. After the images of each case are corrected to the standard brain, ROIs can be set for all cases under the same conditions to eliminate arbitrariness (Fig. 6).^{20,72,75,94,167–173}

Reproducibility of ALPS method

The DTI-ALPS method has been shown to be highly reproducible when imaging conditions such as imaging sequence, imaging plane, and head position during imaging are standardized (CHAMONIX study).⁸⁶ Scanner differences did not affect ALPS-index values when imaging conditions were strictly matched. In addition, a study evaluated the reproducibility between MRI system vendors and between analysis pipelines. In this study, two pipelines were developed using DSI studio and FSL software. They validated inter-vendor reliability, inter-rater reliability, and test-retest reliability. The ALPS-index showed good inter-scanner repeatability (intraclass correlation coefficient: ICC = 0.77−0.95, P < 0.001), inter-rater reliability (ICC = 0.96−1.00, P < 0.001), and test repeatability (ICC = 0.89−0.95, P < 0.001).¹⁷⁴

The transverse plane of MRI is usually taken at the anterior commissure-posterior commissure line,¹⁷⁵ which is also the standard for evaluation in the ALPS method. Since this cross-section is orthogonal to the pyramidal tract, it is important for setting the ROI for the ALPS-index calculation. Changes in imaging plane and head position reduced reproducibility.⁸⁶ To overcome this effect of head position on the ALPS-index, Tatekawa et al. developed a method to calculate the original and reoriented ALPS-index using a technique that registers DTI vector information in a different space and creates a reoriented diffusion map. This technique enables the reconstruction and creation of diffusivity maps along the x-, y-, and z-axes even when the head position and imaging plane are inappropriate, thus improving the reproducibility of the DTI-ALPS-index calculation.^176,177

A report has proposed the application of harmonization techniques in multi-center studies. One multi-center study used the COMBined Association Test (COMBAT), a harmonization technique using regression of covariates in an empirical Bayesian framework, to harmonize ALPS-index variation across scanners, sites, and protocols. COMBAT harmonized the scanner differences and improved Cohen’s d of the left-right ALPS-index between AD and normal control.¹⁷⁸ As another means of improving reproducibility, Saito et al. proposed ALPS-index maintaining tensor vector orientation information (vALPS-index). In their result, the absolute difference in ALPS-index values between scan and rescan was larger in the ALPS-index than in the vALPS-index by approximately 0.6% as the relative difference. Cohen’s d for the left and right ALPS indices between methods were 0.121 and 0.159, respectively. Thus, the vALPS-index based on DTI-ALPS maintaining tensor vector orientation information has higher reproducibility than the conventional ALPS-index.¹⁷⁹

Variation of ALPS method

In the CHAMONIX study described above, although the number of MPGs (motion-probing gradients) affected ALPS-index values, the ALPS-index values of 12-axis DTI and 3-axis DWI showed a good correlation. This suggests the possibility of using the DWI-ALPS method with triaxial MPG.⁸⁶ In fact, the DWI-ALPS method has been reported to calculate the ALPS-index using DWI, which is a simpler method than diffusion tensor images. The DWI-ALPS method has the advantage of obtaining the ALPS-index from DWI, which can be acquired in about 1 min, and it is possible to incorporate this method into routine clinical MRI examinations.⁹⁵ An examination of the ALPS by the DWI-ALPS method in patients who received whole brain radiotherapy for brain tumors and normal controls is reported. The ALPS-index was significantly lower in the post-irradiation group than in the control group in the comparison of all age groups, suggesting that the ISF dynamics were affected in the patients after whole-brain radiation therapy. Interestingly, no significant differences were found in the younger age group. There was a weak negative correlation between ALPS and irradiation dose.⁹¹ When performing the DWI-ALPS method, it is important to note that, unlike the DTI method, the positional relationship between white matter fibers and the ROI cannot be adjusted in post-processing, so the transverse section must be perpendicular to the projection fiber at the time of imaging.

Controversy in DTI-ALPS Method

It is pointed out that the ALPS method is theoretically deductive.¹²² The relationship between ALPS-index and human glymphatic function has not been substantively and rigorously validated by pathophysiological studies. Therefore, the relationship between ALPS-index and glymphatic clearance should be interpreted with caution.^99–102 However, a number of reports have recently been published comparing the ALPS method with other glymphatic system evaluation methods using different imaging modalities to examine the validity of the ALPS method. This review highlighted a number of such reports. Of particular importance is the report by Zhang et al. on the ALPS-index and intrathecal contrast administration for the evaluation of glymphatic system function, showing a strong correlation between the two methods.⁸⁵

However, it is inevitable that the ALPS method has several problems, and these problems are discussed below. As already mentioned, manual ROI placement is one of the main problems of the DTI-ALPS method. Manual ROI placement can be influenced by subjectivity and arbitrariness. It hinders uniformity in the evaluation of multiple cases. This has been pointed out in many papers.^{68,80,81,88,160} This is discussed in section “Placement of ROI” with respect to various methods of solution.

One of the major problems with DTI-ALPS is the mixing of different information. There are two aspects to consider. One is the problem of mixing of water movement at various speeds that is being detected in the diffusion image. Although b = 1000 s/mm² has been used in many studies, it has not been verified whether this is the optimal b value to evaluate the movement of free water such as CSF and cerebral ISF in the tissues. This has been pointed out in several publications.^86,167 The use of a single b-value instead of multiple shells makes it impossible to separate the velocity components of diffusion, and the information is considered to be a mixture of diffusion components of different velocities. This issue needs to be verified in the future.

Another mixing problem is with respect to the anatomical structures contained in the voxels. These problems are due to the large voxel size of diffusion images.^57,101,180 The spatial resolution achievable with normal DTI is far larger than that of tissue structures including the PVS. The ROIs for calculating the ALPS-index include not only the medullary vessels and PVS, but also the surrounding white matter. Therefore, it is not possible to evaluate only the diffusivity of the PVS along the medullary vessels. However, this is not a fatal disadvantage of the ALPS method. The ALPS method aims at a relative evaluation of the diffusion component “along” the PVS direction, not “within” the PVS. It is important to understand that the ALPS method is a direction-dependent analysis. In other words, note again that the purpose of the ALPS method is to evaluate the directional dependence of water movement that can be observed in a diffusion image with b = 1000 s/mm.²

However, even under the above conditions, various problems have been pointed out. Placed ROIs may be affected by partial volume effects of other fiber tracts traveling nearby, such as the corpus callosum and insular cortex.^20,47,62 In addition, it may not be possible to evaluate only ISF dynamics due to mixing of pathological tissues. For example, multiple sclerosis cases may contain demyelinating lesions within the voxel. Diffuse axonal injury may be present in cases of head trauma.⁶⁶ Given the resolution of MRI and the scale of the tissue, reducing the voxel size, i.e., increasing the resolution, will not solve this problem. Rather, it can be evaluated by statistically adjusting for other possible factors. There is a study that statistically eliminated the effects of possible confounding factors delivered from images such as fractional anisotropy (FA) or mean diffusivity (MD). In a study aims to assess the relationships between metabolic syndrome (MetS) by Andica et al., individuals with MetS had a significantly (P = 0.031; Cohen’s d = 0.86) lower ALPS-index than the healthy controls, with age, sex, years of education, total Fazekas scale, PSQI score, intracranial volume, fractional anisotropy, and mean diffusivity included as confounding factors. The changes in the ALPS-index were observed after adjusting for FA and MD in general linear model univariate analysis, suggesting that the lower ALPS-index found in this study was primarily contributed by the changes in water diffusivity in the perivenous space and less likely impacted by the changes in white matter integrity.¹⁸¹

In disorders with brain deformities, such as iNPH, the DTI-ALPS method should be interpreted with caution. This is because the angles between white matter fibers and medullary vessels also change when the brain is deformed, such as in hydrocephalus. In the placement and measurement of ROIs, care should be taken to ensure that they are orthogonal to each other. However, Georgiopoulos et al. report that although ventricular enlargement has a strong influence on the ALPS-index, it does not cancel out the influence of the diagnosis itself.⁵⁷ One other mixing issue is the influence of the white matter hyper-intensity burden.¹⁸² However, the median ALPS-index was reported to be similar in patients with and without white matter hyperintensities.¹⁸³

Another major problem with the DTI-ALPS method is that it can only evaluate the white matter outside the lateral ventricles in an image section that includes the body of the lateral ventricles. The ALPS method cannot evaluate perivascular diffusivity in other regions where the PVS does not run in the x-, y-, or z-axis directions. It is unlikely that evaluation by the ALPS method, which is an evaluation at a limited site, reflects the function of the entire glymphatic system. The waste removal system in the brain has been confirmed to have various functions at various sites, and it is obvious that the phenomena observable by the ALPS method are only a part of these functions. This is also a problem that has been pointed out in a number of papers.^{27,28,30,47,61,82,90,163,173,184,185} This problem is an inherent weakness of the ALPS method and cannot be solved. However, other methods discussed in this review also do not evaluate the whole system. Therefore, it is desirable to evaluate the glymphatic system, ISF dynamics, and the mechanism of excretion of cerebral wastes in combination with other various methods. The complementary use of multiple evaluation methods is expected to be the trend in the near future.

Conclusion

The evaluation methods of the glymphatic system were reviewed with the ALPS method as the axis. Although the glymphatic Hypothesis has been the subject of much debate, it is a groundbreaking theory in that it evaluates the CSF and ISF that fill the central nervous system as a common system and shows their involvement in mass transport, including the excretion of waste products. There seems to be no single definitive method for the evaluation of cerebral ISF dynamics, which is a much more complex system than that of perfusion by the vascular system. The ALPS method and the other methods discussed in this study are intended to evaluate ISF dynamics in limited aspects, and future research is expected to integrate these multiple methods to elucidate the overall ISF dynamics in the brain.

A high ALPS-index is an indicator of predominant left-right Brownian motion of water in the plane at the level of the lateral ventricles, neither more nor less. As an originator of this method, it is regrettable that the title of the first paper (Evaluation of glymphatic system activity with the diffusion MR technique: DTI-ALPS in Alzheimer’s disease cases) clearly related the glymphatic system to the ALPS method.^4,186,187 However, this is also true for other methods of assessing cerebral ISF dynamics other than the ALPS method. In other words, there is no single method that can cover the entire system. Fortunately, the term ALPS-index has become popular and is now known to many researchers as a common term in this research area. It would be correct to describe the ALPS-index only as high or low, increased or decreased, and whether it reflects the glymphatic system or not should be considered as a subject for discussion. In other words, if a decrease in ALPS-index is observed, the title of the paper should say “decrease in ALPS-index” and not “glymphatic dysfunction” directly at least in the title of the paper. With regard to ALPS method, Professor Iliff mentioned that “Changes in the measure are seen under conditions where glymphatic function are impaired – so while we don’t have a strong handle on the exact nature of the signal it seems to work at some level” (personal communication), and the author agree with his opinion.

It is expected that various methods will reveal the full extent of waste elimination and ISF dynamics in the brain in the future. The ALPS method is expected to continue to provide useful information as one of the various evaluation methods.