Although the existence of perivascular spaces has been confirmed by pathologists for more than 150 years and MR has made it possible to clearly visualize enlarged perivascular spaces (EPVS), the function and clinical significance of perivascular spaces continued to be debated until 2012, when Iliff et al. (5) uncovered a new mechanism of brain metabolism-a unique para-vascular pathway of brain metabolism, thus gradually unraveling the mystery of perivascular spaces. Their study revealed the presence of interstitial solute clearance structures dependent on astrocyte podocyte aquaporins 4 (AQP4) in the perivascular space of the rodent brain and termed it the “glymphatic system” (5). Iliff et al. also found that soluble amyloid β (Aβ) protein injected into the striatum of mice was cleared from this para-vascular pathway (5). In their subsequent study, the exchange of CSF with ISF in the live rat brain was observed by dynamic contrast-enhanced MRI through intrathecal injection of two different molecular weights of MRI contrast agents. Both contrast agents passed largely similarly through the para-arterial space in the rats, but the smaller molecular weight contrast agent entered the brain parenchyma more extensively, while the larger one was sparsely distributed in the brain parenchyma (6). This finding points to the molecular size-selective nature of the glymphatic system for clearing substances, which may explain the tendency of large molecular Aβ protein to be deposited around blood vessels. The discovery of meningeal lymphatic vessels in 2015 was even stronger evidence for the existence of brain glymphatic system (7).

The precise drainage pathways and mechanisms of the glymphatic system are not known. It is hypothesized that the system consists of three principal segments (2, 8): (1) a para-arterial CSF influx, (2) an astrocyte end-foot water channel protein AQP4 mediated convection of CSF and ISF, and (3) a para-venous ISF efflux (Figure 2). The possible circulation pathways of the system are as follows: neuronal metabolites are carried to the perivascular space via convection of CSF and ISF facilitated by AQP4, from where they are transported directed into lymphatic vessels located in meninges and in the soft tissue surrounding the skull and eventually return to the general circulation for clearance by the kidneys and liver via cervical and nasal lymph nodes (Figure 2) (2, 7, 8).

The possible factors driving ISF through perivascular spaces are continuous production of CSF (8), vascular pulsation (9, 10), respiratory movement (11), the sleep–wake cycle (12–14), and AQP4 water channel protein (15).

The power source for this fluid drainage system is generated by the continuous CSF flow. The production of CSF from the choroid plexus drives the fluid from the ventricular system into the subarachnoid space and along the perivascular space into the cerebral interstitium. The driving force generated by CSF flow is important for the exchange between CSF and ISF and helps to flush out metabolites within perivascular spaces (8).

Cerebrovascular pulsatility drives fluid transport along this pathway. Rodent studies have shown that increased intracranial vascular pulsation is associated with increased PVS, and that vascular pulsation, particularly in the small cerebral arteries, is directly involved in the production of enlargement of perivascular spaces in the brain (9, 10). The elevation of blood pressure and the increase in the amplitude of pulsation of the arterial wall, although the arterial diameter remains constant, causes a decrease in the net flow of PVS and a decrease in debris clearance (10). Spontaneous low-frequency oscillations were observed in the arteries of awake mice which drives the drainage of solutes from the PVS (16). In mice with cerebral amyloid angiopathy (CAA), impaired vascular reactivity resulted in reduced solute clearance (16).

The similar effect of vascular pulsation on the glymphatic function of the human brain is derived from phase contrast MR. Lower cerebrovascular reactivity (CVR) was associated with higher WMH volume and increased PVS visibility in the basal ganglia (17). Furthermore, increased intracranial vascular pulsatility was associated with lower CSF stroke volume at the foramen magnum and, in turn, with increased PVS burden at the basal ganglia (17).

Respiratory movements affect the fluid drainage within the PVS by influencing CSF transport. Forced inspiration causes the CSF to move cephalad, while the venous blood draws away caudally (11). The movement of CSF from breathing might also flush perivascular spaces.

The glymphatic system is more effective during sleep. Using real-time assessment of tetramethylammonium diffusion and two-photon imaging, a recent study evaluated tracer clearance in live mice during wakefulness, natural sleep, and anesthesia. They found that the rate of clearance was highest during sleep, with the same performance under anesthesia, and a 95% reduction in tracer flux into the perivascular spaces of the mouse brain during wakefulness (12). Similarly, the clearance of Aβ by mice doubled during sleep (12). The study also shows a 60% increase in interstitial space under natural sleep or anesthesia, thus explaining the significant increase in the convection between CSF and ISF. Sleep may be involved in perivascular clearance in ischemic encephalopathy (13). CSF uptake and ISF drainage from the PVS increases during sleep, and poor sleep quality may affect the clearance of neurotoxins, interrupting ISF drainage and possibly leading to enlarged perivascular spaces (13). Hablitz et al. studied the relationship between the sleep timing and the glymphatic function in mice and found that the influx and clearance of glymphatic fluid had a circadian rhythm and peaked in efficiency during the mid-rest phase (14). The peak of perivascular AQP4 polarization occured simultaneously during the resting phase in mice, and the absence of AQP4 eliminated diurnal differences in glymphatic flow and drainage (14). Therefore, they concluded that the distribution of CSF is controlled by the diurnal rhythm, and AQP4 is involved.

AQP4 is a water channel protein on the feet of perivascular astrocytes and is a regulator of normal glymphatic function (15, 18), which facilitates the exchange of CSF and ISF in the VR spaces and the active outflow of ISF (5). Although AQP4 is the major water channel protein in mammals, it is only expressed in astrocytes and ventricular meningeal cells. The outer wall of the VR space is composed of astrocyte end-feet with abundant AQP4, which allows water in this interstitial space to move into the ISF, making the transport of small molecules and ions possible. In AQP4 knockout mice and rats, the transport of tracers in the CSF and ISF was significantly reduced, suggesting that the AQP4 gene is necessary for the fluid transport in the brain (5, 15). But the importance of AQP4 in the glymphatic system remains controversial. Smith et al. founded that uptake of CSF tracers and transport of fluorescent dextrans from subarachnoid space to brain parenchyma were unaffected by AQP4 gene knockout in rats or mices (19). They suggested that AQP4 does not appear to be necessary for fluid movement within the glymphatic system in rodent brains, and this movement applies more to diffusion than to convection. These conflicting findings may be due to the use of different techniques or anesthetics by different investigators (3, 20).

Glymphatic clearance is a molecular size selective mechanism (5, 6). Soluble Aβ (5), tau (3) and lactate (21) can be selectively removed through this system. The system also contributes to the delivery of apolipoprotein E (ApoE) (22), lipids (23) and signaling molecules (23). AQP4 plays a major role in the transport of ions and small molecules. Polarized AQP4 can fold toward the center and constitute highly selective water pores, allowing the smooth passage of water molecules (2, 5, 6, 8). The abundance of AQP4 on the astrocyte end-foot, which forms the outer wall of the perivascular space, provides for the transport of small molecules and ions. At the same time, the cleft between astrocyte end-feet allows the passage of macromolecular compounds, and most of the solutes with molecular weight <100 kDa will leave the perivascular spaces through this cleft (8). Today's research is increasingly focused on the glymphatic clearance mechanisms, hoping to develop new therapeutic strategies for associated neurological disorders in the future.

The PVS, as the main conduit for the drainage of ISF from brain tissues, is responsible for the exchange between CSF and ISF, the removal of intracerebral waste, and the maintenance of brain homeostasis (2). PVS also acts as a conduit for delivering various signaling molecules and metabolic factors (2, 18). The precise causes and mechanisms of PVS enlargement are not yet known. As the anatomical basis of the glymphatic system, the increased visibility and enlargement of PVS on imaging may be potentially linked to the dysfunction of the glymphatic system. Theoretically, PVS enlargement should be caused by fluid accumulation due to imbalance between CSF inflow and outflow, but direct evidence is still lacking (24). Altered blood flow and blood-brain barrier (BBB) dysfunction are also believed to be involved in PVS dysfunction (24, 25).

It is controversial whether the visible PVS on MRI refers only to the periarterial spaces or to both the periarterial and perivenous spaces. Most of the evidence suggests that visible PVS on MR are around arteries (3). However, the answer to the question of whether the absence of visibility on MR equates to the absence of enlargement of the perivenous spaces or the absence of venular dysfunction is no. Histopathological studies of autopsy specimens confirm the presence of many visible perivenous spaces in areas of collagenization of deep penetrating veins (3, 26). However, the extent to which this venous collagenization contributes to the visibility of PVS on MR, and the incidence and severity of venous collagenization in small vessel disease are unknown. Furthermore, the available studies do not find that the main pathway of solute efflux in the glymphatic system is along perivenous routes, but rather there is much evidence that solute inflow and efflux occur along periarterial routes (27). If the main pathway of solute efflux is not perivenous routes, could this explain to some extent why enlarged perivenous spaces are usually not observed on MR.

Patients with dilated VRS are generally asymptomatic, but multiple enlarged VRS or giant VRS may occasionally show the following symptoms or signs: headache, dizziness, cognitive impairment, inattention, visual impairment, epilepsy. PVS may be aggravated by aging, hypertension, inflammation, lacunar stroke, and dementia (3).

The enlarged perivascular space (EPVS) at different sites may have different pathophysiological basis and pathogenic mechanisms. EPVS often coexists with lacunar infarction. EPVS tends to be more pronounced in the basal ganglia region in patients with lacunar stroke than in those with non-lacunar stroke (28). Ischemic stroke patients with a higher EPVS load in the basal ganglia region have a higher risk of stroke recurrence (24). Small artery occlusion stroke subtypes have a relatively higher risk of high load EPVS in the hippocampus (29). Patients with CAA have a higher total cortical Aβ load and impaired Aβ clearance leading to enlarged perivascular spaces, especially in the centrum semiovale (30, 31). In patients with spontaneous cerebral hemorrhage, basal ganglia EPVS severity is associated with hypertensive arteriopathy (31), and high EPVS load in the centrum semiovale is associated with cerebral hemorrhage recurrence in patients with CAA (32). Post-stroke depression (PSD) is a common psychiatric symptom after stroke, occurring in approximately 1/3 of patients (33). EPVS is thought to be associated with the development of PSD. Researchers found that centrum semiovale rather than basal ganglia EPVS is more severe in patients with PSD and is associated with poor antidepressant response (34).

There are conflicting data on MR regarding the relationship between perivascular spaces and cognitive decline. Most studies have concluded that the burden of EPVS in both the basal ganglia and centrum semiovale is associated with cognitive decline (3). However, some studies did not find an association between EPVS in the basal ganglia region and cognitive dysfunction (3), whereas Arba et al. came to the opposite conclusion, suggesting that EPVS in the basal ganglia region, but not in the centrum semiovale, was associated with cognitive decline (35). Three other studies showed no independent association between EPVS and cognitive impairment (36, 37) and that EPVS was not a predictor of cognitive decline (38). These contradictory data may be related to the differences in the enrollment and assessment methods of each study. In addition, due to the wide range of cognitive functions included, EPVS may affect only some of these cognitive domains and not necessarily the overall cognitive function of the whole brain.

In addition, CADASIL 45, PD, cerebral trauma, idiopathic normal pressure hydrocephalus, mucopolysaccharidosis (39), neuropsychiatric retardation, autism (40), systemic lupus erythematosus (SLE) (41, 42), multiple sclerosis, primary angiitis of the central nervous system (43, 44), and cryptococcal infection (45, 46) are often accompanied with EPVS. For example, in patients with mucopolysaccharidosis, transient sieve-like PVS can appear in the white matter in the early stages of the disease and can increase, decrease or remain stable depending on the progression of the disease (39). In patients with cryptococcal infection, multiple EPVS can be found in the cerebral hemispheres, parietal ventricles, midbrain, bilateral cerebellar hemispheres, and especially in the basal ganglia, which is considered the most common early sign of cryptococcal infection and is associated with intracranial dissemination and proliferation of cryptococci (45, 46).

EPVS is not a specific disease, but should be considered as one of the imaging markers for the development of multiple diseases, and its mechanism of action and corresponding clinical relevance in neurological diseases are still unclear and need further investigation.

Based on STRIVE (1), PVS were referred to as a neuroimaging marker of CSVD. MRI instead of CT is recommended for detecting PVS. The recommended sequences include T1-weighted imaging (T1WI), T2-weighted imaging (T2WI), fluid attenuated inversion recovery sequence (FLAIR), diffusion-weighted imaging (DWI) and apparent diffusion coefficient map (ADC). VRS are best detected using T2WI. DWI is the most sensitive sequence for RSSI. FLAIR is used to identify VRS from WMHs and lacunes. A field strength of 1.5T is often of a similarly high resolution as 3.0T (47). It is generally believed that the VRS seen on conventional magnetic field strength (1.5T or 3.0T) MR is the peri-arteriolar space (48).

The signal intensity of perivascular spaces is consistent with CSF on all MRI sequences showing hypointense on T1WI and FLAIR, hyperintense on T2WI, without enhancement and generally no mass effect (1, 49). The diffusion of PVS is unrestricted. The signal of the surrounding brain is generally normal.

PVS appears striped when the penetrating artery is parallel to the imaging plane and rounded or ovoid when perpendicular to the imaging plane (Figures 3A–F) (1, 49). PVS is typically <3 mm in diameter (1). The distribution of PVS is usually bilaterally symmetrical. Sometimes, high-resolution MRI can reveal microscopic structures within the PVS: a central vessel can be observed in the cystic space and is defined as the “vessel sign”, which could be possibly helpful for differentiating the spaces from lacunes (Figure 3B) (1, 49). VR spaces can show a solitary round cyst, multiple strips and lines, unilateral predominantly dense cysts, or even a spider web-like appearance on MRI (Figure 3C).

PVS are commonly located in the basal ganglia, the centrum semiovale, the subcortical white matter of brain convexity, and the brainstem, however, the space is rarely seen in the cerebellum (1, 49). Large PVS is often first seen around the penetrating artery in the precribrum even in young people.

Perivascular spaces of 1–2 mm appear in all age groups (49, 50). Mildly enlarged PVS can also be detected in routine MR examination in healthy individuals of all ages, however, the burden of EPVS becomes increasingly apparent with advancing age (50). PVS of >3 mm can be observed in approximately one-third of elderly people (50). Increased visibility of VRS in the BG and white matter is significantly associated with age, and the VRS burden in the BG is heavier in men than in women (50). PVS with a diameter >3 mm must be distinguished from small lacunes. Lacunes tend to be 3–15 mm in diameter with a hyperintense rim around the vesicular cavity on FLAIR sequence (Figures 3D,E) (1). Sometimes, PVS can be especially enlarged with a diameter larger than 15 mm, even with mass effect, which is named as giant PVS or tumefactive PVS (Figure 3F) (51).

VR spaces are traditionally classified into three types based on their location. Type I-basal ganglia type: PVS courses with the lenticulostriate artery entering the basal ganglia through the precribrum (52); This is the most prevalent type. Type II-Hemispheric type: PVS follows the course of the medullary arteries into the white matter of brain convexity and extends into the subcortical white matter (52). Type III-midbrain type (Figure 3A): PVS lies in the midbrain following the path of penetrating artery from the posterior cerebral artery.

Similarly, giant VR spaces or tumefactive VR Spaces can be categorized into the three types described above, but type II and type III are more common (53). Type II VR spaces rarely cause clinical symptoms, but when they are diffuse, they may mimic PD or dementia (53). They are also seen in patients with neurofibromatosis and neuropsychiatric retardation (54, 55). Due to its special location, Type III VR spaces are more likely to compress the middle cerebral aqueduct or third ventricle, causing hydrocephalus. In some cases, tumefactive PVS can also cause pulsatile tinnitus (56), homonymous quadrantanopia (57), paralysis of the extremities (58), and reversible focal dystonia (59). Unlike the small PVS, the brain parenchyma surrounding the giant PVS has an abnormal signal intensity in 32.3% of type II cases and 3.8% of type III cases (53). They are often misdiagnosed as other pathologic processes, especially cystic neoplasms (49).

A recent study described an enlarged perivascular space located in the anterior temporal lobe with unique neuroimaging features (60). It occured in association with the vascular loop of the middle cerebral artery (MCA) branch and often compressed the adjacent cortex, with edema around the lesion. The clinical presentation was very similar to that of a cystic tumor and was often misdiagnosed and surgically removed as such. The author suggested that it should be considered as a separate subtype of perivascular spaces, namely type IV perivascular spaces, and proposed the term “opercular perivascular cyst” as naming. This type of PVS exhibits the same internal signal characteristics as CSF in all MRI sequences, without contrast enhancement or magnetic susceptibility artifacts, and the brain parenchyma surrounding the lesion shows variable high signal on T2WI and FLAIR (60).

Early studies of the cerebral glymphatic system were mostly carried out using two-photon imaging and fluorescence tracers. Iliff et al. used this technique to visualize the glymphatic pathway for the first time in rodents and showed that Aβ, the causative peptide of AD, was cleared from this pathway (5). The system is more effective during sleep (12, 14). A recent study explored the mechanisms of brain edema formation after stroke using radiolabeled tracers and multiphoton imaging coupled with DWI-MRI. According to the study, post-stroke cerebral edema is caused by an accelerated influx of CSF into the brain parenchyma via perivascular spaces within minutes of an ischemic insult (79). The finding overturns the traditional understanding of poststroke edema and provides a new concept for the development of alternative treatment strategies.

Researchers are continually innovating existing photo-microscopy techniques in order to obtain more accessible, minimally invasive, and higher definition images. Using a multiband filter cube, tunable LED light source, and CMOS camera, Sweeney et al. presented a transcranial fluorescence macroscopic imaging method to visualize CSF transport within perivascular spaces through the intact skull of living mice (80). In addition, a novel method for qualitative and quantitative estimation of glymphatic flux in rats using Evans blue-labeled albumin (EBA) was proposed that does not require radioactive labeling (81).

However, due to its drawbacks such as narrow field of view, shallow penetration depth, invasiveness to the study subject and the need for small cranial openings, this technique is not applicable to human brains. A real-time image of the entire human brain is not possible. Human brain researches are mainly focused on postmortem tissue assessment. Researchers evaluated the postmortem frontal cortex using Western blotting and immune-fluorescence from a total of 79 cognitively competent and histo-pathologically proven AD patients, and found a significant association between AQP4 expression and advancing age in all the participants (82). Individuals with AD had reduced AQP4 perivascular localization, and loss of perivascular localization slowed the clearance of Aβ and promoted its deposition (82). This may be one of the most successful applications of microscopy techniques in the human brain glymphatic system.

MRI's sensitive, non-invasive nature has made it possible to bring the study of the glymphatic system from the laboratory to the clinic. Researchers today are becoming more and more interested in the application of novel MR imaging techniques for assessing the human glymphatic system.