Narrative Review of Neuroimaging in Migraine With Aura

Karissa N. Arca MD; Juliana H. VanderPluym MD; Rashmi B. Halker Singh MD


Headache. 2021;61(9):1324-1333. 

In This Article

Interictal Imaging

Noncontrast MRI of the Brain

Both nonspecific white matter abnormalities or hyperintensities (WMAs/WMHs) and "infarct-like lesions" (ILLs) have been identified in routine neuroimaging of patients with migraine with or without aura. A 2013 systematic review and meta-analysis by Bashir et al. investigated three types of MRI findings in MWA and migraine without aura (MWoA)—WMA, ILLs, and volumetric gray and white matter changes.[23] WMA were defined as small, punctuate hyperintensities found in deep, periventricular, and infratentorial areas. Silent ILLs were defined as lesions ≥2 mm and isointense to CSF (Virchow–Robin spaces excluded). Volumetric changes were characterized by voxel-based morphometry (VBM) and will be discussed in the volumetric imaging section below. In the meta-analysis portion of the study, WMAs were no more prevalent in MWA compared with MWoA. MWA had a relatively higher odds ratio of ILLs compared with MWoA; however, only two studies were included in the meta-analysis of ILLs. There was no significant heterogeneity between the studies used for the meta-analysis.[23]

In the Cerebral Abnormalities in Migraine, an Epidemiological Risk Analysis (CAMERA) study, which was included in the analysis by Bashir et al., women with migraine (either MWA or MWoA) were independently found to have increased risk of developing deep WMH on MRI as were those with higher attack frequency and longer disease duration. Migraine subtype (MWA or MWoA) did not play a role in risk stratification.[24] Two studies have found that in people with MWA and high volume of WMH, interictal cerebral blood flow is lower compared with MWoA and controls.[24,25] A co-twin (where one twin had migraine and the other did not) and control study of 166 patients did not find a statistically significant association between WMH and subclinical infarcts in patients with MWA compared with controls and co-twins.[26] The CAMERA study also identified an increased number of ILLs in the posterior circulation in people with migraine compared with controls; those with MWA had a significantly greater prevalence of these lesions (although the specific association with subclinical infarcts and MWA was not supported in the Northern Manhattan Study [NOMAS]).[27,28] In a subsequent analysis, investigators found that patients with posterior circulation ILLs tended to be older and had a higher incidence of deep white matter lesions.[29] ILLs in the posterior circulation, particularly at border zones, may be due to hypoperfusion and arterial narrowing resulting in buildup of embolic molecules and thrombus formation.[29] Collateral blood flow is less prominent in the posterior circulation, which may also predispose the posterior fossa to such lesions. The extent of intracranial atherosclerosis in participants from the CAMERA study is unknown, nor is the presence of atrial fibrillation, patent foramen ovale, or hypercoagulable states.[29]

The significance of WMA/WMH, and ILLs in migraine, specifically MWA, is not yet understood. Additional systematic review and meta-analysis would be beneficial to directly compare more recent studies. For patient care, clinicians should be aware of the typical distribution and incidence of these imaging findings to delineate migraine from a secondary disorder and to provide effective patient counseling. For instance, brain MRI should be performed in patients presenting with a strong family history of MWA and early cognitive impairment due to concern for Cerebral Autosomal Dominant Arteriopathy with Subcortical Infarcts and Leukoencephalopathy (CADASIL). Up to 45% of people with CADASIL have MWA, which is often the first presenting symptom. MRI findings in patients with CADASIL can be striking and often start with hyperintense subcortical lesions on T2/FLAIR sequences. Initially, this may have the presentation of chronic small vessel disease. However, as the disease progresses, lesions become prominent in the anterior temporal lobes and the external capsule. Many patients will also have microhemorrhages and brain atrophy.[30]

Volumetric Imaging

Structural differences in the brains of people with MWoA or MWA are not observable on routine imaging; rather, they can be appreciated in research studies examining brain volume. A recent meta-analysis analyzed 27 cross-sectional voxel-based morphometry (VBM) studies, which included 1086 people with migraine and 877 healthy controls; of those with migraine, 216 had MWA. Most subjects were scanned outside of a headache attack, and aura was not separately captured. Sixteen of the 34 data sets independently found no change in gray matter volume (GMV), and when all data sets were pooled, there was no consistent difference in GMV between people with migraine and controls. Metaregression analyses were performed but did not include MWA as a potential influential factor in the results.[31] A subsequent longitudinal study capturing preictal, ictal, and postictal migraine phases of seven subjects ("ictal" presumed to be referring to the headache phase, rather than the use of "ictal" in this paper referring to the aura phase) found no phasic gray matter alterations.[32]

Bashir et al. also evaluated gray and white matter changes in MWA, MWoA, and controls; however, meta-analysis was not completed for these imaging findings.[23] Nine studies that used either VBM or diffusor tensor imaging were identified, seven of which reported lower GMV in various areas of the brain in migraine (including both MWoA and MWA) compared with controls. Five studies reported reductions in gray matter that correlated with attack frequency and disease. Only one study in the meta-analysis identified increased GMV in the periaqueductal gray and dorsolateral pons in patients with MWA compared with MWoA.[23,33] To further support that the brainstem may be involved in MWA through activation of the trigeminovascular system, another study outside Bashir's analysis demonstrated that those with MWA had larger total brainstem volume, specifically in the midbrain and pons.[34,35]

The following studies were published after Bashir et al. A recent study of 28 patients with migraine using interictal VBM to compare migraine with and without visual aura found that MWA compared with MWoA had reduced GMV in the right cerebellum, left pre- and postcentral gyri, right inferior frontal gyrus, left Brodmann area 20–22 and left lingual gyrus. Conversely, MWA compared with controls had increased GMV in the right superior parietal gyrus and left thalamus. Both MWA and MWoA compared with controls had reduction of GMV in the cerebellum, frontal, and temporal lobes.[36] In the aforementioned co-twin study, subjects with MWA were also found to have a thicker cortex in the V2 region of the occipital cortex compared with patients without migraine.[37]

It is unclear whether the location of pain or aura during a migraine attack is related to GMV changes. A single study evaluating 20 patients with side-locked aura compared the brain structure on the typical headache side with the contralateral hemisphere. The cortical thickness of the inferior frontal gyrus on the typical headache side was reduced compared with the contralateral hemisphere.[38] However, it is uncertain which side should be considered "hypertrophied" or "atrophied." The inferior frontal gyrus is involved in pain modulation and inhibition, as are many of the aforementioned structures.[38]

Comparing volumetric differences in different types of aura, Hougaard and colleagues prospectively examined the MRI brain structures of 60 people with migraine with visual or combined visual and sensory aura and compared them with healthy controls.[39] They found that GMV was decreased in the anterior cingulate cortex in MWA compared with controls. However, there was no difference in GMV or cortical thickness in MWA with or without sensory aura.[39] Reduced GMV in the cingulate gyrus is not unique to migraine; it has also been documented in other neurological, psychiatric, and pain disorders. Although this may be a marker of disease, the prognostic or pathophysiologic significance is not fully understood.[39] In a small study, Petrusic and colleagues also investigated morphometric changes in the cerebral cortex in patients who experienced only visual aura and those who experienced visual, somatosensory, and language symptoms.[40] The cortical surface area of the left rostral middle frontal cortex was reduced, and the sulcal depth of the left temporal pole was increased in the patients with multiple aura symptoms compared with visual aura only. These changes somewhat correlate with the affected brain area producing aura symptoms; however, it remains unknown if alterations in the cortical structure represent an inherited or disease-reactive state.[40]

The systematic review by Bashir et al. identified that reduction in gray matter correlated negatively with migraine frequency and duration. However, changes in GMV, cortical thickness, or thalamic volume do not appear to be related to disease duration or attack frequency in any of the subsequent studies, although methodology and sample size were variable.[34,36,37,39–41] Overall, the volumetric changes appreciated in MWA do not directly correlate to the structures involved in symptomatically producing aura (i.e., occipital cortex for visual aura and somatosensory cortex for sensory aura). Further longitudinal studies are required to understand if the observed changes are inherited or reactive to disease.

Functional Imaging and Spectroscopy

As previously noted, fMRI is a useful tool for research studies but currently has a very limited role in direct patient care. Hougaard and colleagues investigated the functional interhemispheric differences interictally in response to visual stimulus in 20 patients with "side-fixed" visual aura attacks and found that the visual stimulus did not provoke aura or a migraine attack.[42] However, in response to visual stimulus, the BOLD responses were increased in the symptomatic hemisphere of those with MWA; specifically, the inferior and superior parietal lobes and the inferior frontal gyrus. Although the primary visual cortex was not involved, these areas of the brain are known to be involved in various visual functions, including eye movements and attention to movement of objects in the visual field.[42] Although additional investigations should be done to confirm these findings, these results may help clarify other vision symptoms experienced by people with migraine and highlight the intricate connectivity of cortical structures in addition to the primary visual cortex, which may be involved in the generation of MWA.

It is important to consider that the patients in the aforementioned study had "side-fixed" visual aura. Extrapolating these data to others with variable visual aura may not produce the same results; rather, one might expect that interhemispheric activity may be equal. However, one would still expect hyperactivity compared with controls.[42] This, however, was not the case in the study by Hadjikhani and colleagues in which the amplitude of the BOLD signal remained the same during visual stimulation; there was no difference in activation in those with or without migraine.[22]

There have been mixed findings of resting state functional connectivity in MWA compared with controls when no stimulus is provided. In a subsequent study by Hougaard et al., subjects with migraine and controls demonstrated no differences in functional connectivity when no visual stimulus was given.[43] In contrast, another resting state fMRI study demonstrated that patients with MWA had increased functional connectivity in the right lingual gyrus compared with controls and MWoA.[44] Taking both the resting state and stimulus-induced findings into account, perhaps the brain dysfunction in MWA is due to a combination of individual neuronal hyperexcitability in response to physical stimuli as well as abnormal network connectivity.[42,43] However, additional larger studies with unified methodology would be needed to confirm or refute this hypothesis.

Further investigating interictal resting state differences between MWA and MWoA, Farago et al. found that resting state BOLD fluctuation was higher in the cingulate cortex, superior parietal lobe, cerebellum, and bilateral frontal regions in MWA compared with MWoA. Additionally, resting state network frequency was higher in MWA compared with MWoA.[45] Increased resting state BOLD fluctuation has been demonstrated in other chronic pain conditions as well.[45] These resting state data are important for confirming the activity of higher-order brain structures in the pathogenesis of migraine. It furthers the understanding of migraine as a chronic disease with exacerbations rather than a diagnosis of individual discrete episodes. Furthermore, MWA and MWoA have demonstrated decreased activity in the right middle frontal gyrus and dorsal anterior cingulate gyrus compared with controls without corresponding structural changes.[46] These regions of the brain are important for executive functioning and may provide insight into daily functioning of people with migraine. Despite normal neuropsychometric faculties, patients with migraine may be more sensitive to the demands of certain activities of daily living and higher-level cognitive tasks.

MR spectroscopy, another modality only available in the research setting, has identified mitochondrial dysfunction as a potential mechanism for MWA. Multiple modest-sized studies have identified interictal mitochondrial dysfunction during photic stimulation demonstrated by elevated lactate/N-acetylaspartate (NAA) ratio.[47,48] The methodology of these two studies varied with Sándor et al. seeking to find a difference in brain metabolic activity between pure visual aura and other more complex aura syndromes, whereas Sarchielli et al. aimed to evaluate differences in brain metabolites between MWA, MWoA, and healthy controls. Sándor et al. found that migraine with pure visual aura had persistently elevated lactate levels before and during stimulation, whereas lactate only peaked during visual stimulation in migraine with complex aura. Because of the similarities of the pure visual aura pattern with various mitochondrial diseases, the authors speculate that lactate transporter systems may be overloaded in pure visual aura.[48] Sarchielli et al. found a consistent decrease in NAA during photic stimulation in MWA, which was lower than MWoA and controls and attributed to mitochondrial dysfunction.[47] Larger studies should be performed to confirm the findings from these investigations and to determine whether mitochondrial dysfunction is one of the underlying mechanisms of cortical hyperexcitability in MWA.

Hemiplegic Migraine

Some investigators distinguish HM as a separate entity from MWA, whereas others consider HM on the spectrum of MWA. Because both MWA and HM present with acute, progressive neurologic symptoms, we will first describe findings that may be apparent on standard imaging modalities to help distinguish HM from acute stroke, followed by other general imaging findings.

In contrast to diffusion restriction and focal T2/FLAIR hyperintensity seen in acute ischemic stroke, cortical edema has been reported in HM. Two case reports of HM have demonstrated cortical edema on T2-weighted imaging during an attack—one with frontal cortical edema 6 h after symptom onset and the other with temporoparietal and occipital cortical edema 2 days after symptom onset.[49,50] Roth et al. reported cortical edema in the affected hemisphere on T2/FLAIR MRI up to 11 days after symptom onset in patients with genetically confirmed familial hemiplegic migraine type 2.[51] Two individuals were also noted to have dilated intracranial vessels in the corresponding hemisphere 6–7 days into an attack.[51]

Three case reports demonstrated contrast enhancement in addition to cortical edema in the affected hemisphere during an HM attack, demonstrative of blood–brain barrier breakdown.[49,52,53] In one of the cases, the patient had not yet been diagnosed with HM and ultimately underwent cortical biopsy, which showed "advanced neuronal suffering, ballooned cells, neoangiogenesis with fibrohyalinosis suggestive of noninflammatory vasculopathy."[52] This pathologic finding may explain cortical atrophy that has been reported in HM.

The biphasic nature of perfusion in MWA attacks has been demonstrated more precisely in HM. Eleven HM attacks were captured with various imaging modalities in a case series of three adult patients with genetically confirmed familial hemiplegic migraine type 2 (p.H916L mutation in the ATP1A2 gene). Imaging was performed at multiple time points during a single attack. Hypoperfusion was demonstrated up to 19 h after symptom onset and transition to hypoperfusion was seen after 18 h.[20] Roth et al. also investigated perfusion during an HM attack using perfusion-CT, which showed two patients with hyperperfusion in the affected hemisphere at 3 and 6 days into the attack.[51] The earliest that imaging was obtained was 2 days into an attack, which is likely why hypoperfusion was not appreciated.[51]

Several cases of hemicerebral atrophy have been reported in patients with familial hemiplegic migraine type 1 with atrophy corresponding to the affected hemisphere.[54,55] Treatment with sodium valproate may stop progression of the disease and prevent further atrophy by preventing the previously described perfusion changes, which result in neuronal hypometabolism.[54–56] Few studies have investigated the brain metabolism in HM; however, available evidence shows that HM attacks are associated with both hypo- and hypermetabolism and may continue to have hypometabolism between attacks.[56]