Role of Functional Neuroimaging in Primary Headache Disorders
Correspondence Address: Source of Support: None, Conflict of Interest: None DOI: 10.4103/0028-3886.315987
Source of Support: None, Conflict of Interest: None
Keywords: Cluster headache, functional imaging, magnetic resonance imaging, migraine, paroxysmal hemicrania, SUNA, SUNCT, trigeminal autonomic cephalalgias
Neuroimaging in primary headache disorders aims to identify unique pathophysiological correlates of each disorder with potential clinical implications for easier diagnosis and specific treatment options. Recent advancements in functional imaging significantly improved our understanding of the pathophysiology of migraine and trigeminal autonomic cephalalgias.
The first chapter of the international classification of headache disorders (ICHD3) represents the primary headache disorders, consisting of migraine, tension-type headache (TTH), trigeminal autonomic cephalalgias (TACs), and “other primary headache disorders”, which is a mixed bag of mostly rare and not well-studied diseases. In the absence of high-quality data regarding the rare headaches from group four and tension-type headache, this review will mainly focus on migraine and the TACs. The TACs have been registered in different diagnostic groups of the ICHD3, but have common clinical and pathophysiological background.
Except for hemicrania continua, which is by definition, a chronic headache disorder, all other headaches discussed in this review may present with episodic or chronic courses. Periodic worsening and circadian rhythm are characteristic for TACs but similar periodicity is also frequently seen in migraine.,
Furthermore, migraine and TACs share many other clinical features, but differ in others, and even overlapping headache syndromes are well described., All this suggests a common pathway. However, in each of these disorders and for every single patient these symptoms are expressed to a different extent and the overlap may lead to difficulties in distinguishing the different disorders purely on clinical grounds. In order to close this gap, functional imaging has not only been used to unravel the underlying pathophysiology of headache disorders but also, to develop it as a diagnostic tool that could reliably and objectively allow the differentiation of the different headache disorders from each other. From a pathophysiological point of view , neuroimaging was able to identify the hypothalamus, the brainstem and eventually their communicating network as potential key brain regions.,,,, This article gives an overview on the current status of functional imaging in primary headache disorders.
Clinical features of migraine may vary, and even from a genetic point of view migraine is a heterogeneous disorder. This may cause variable responses to different treatments even of newer specific treatment approaches targeting CGRP. Respecting that clinical presentation, genetics, and treatment response variations must be accounted for by the proposed pathophysiology of the disorder , meaning that different mechanisms may lead to the same endpoint - a migraine attack. Even more so as not every attack equals the previous. Early pathophysiological ideas were very blurry and mainly resulted from animal models and clinical observations. Once the vasogenic theory was dismissed, the assumption of altered neural function in peripheral and central structures processing nociception was put forward. Functional brain imaging in migraine may solve some of these difficulties, as it is the only method measuring changes in neural function in vivo and may thereby enrich the insights to pathophysiology in general and possibly even in an individual patient, where different aspects may be present to a different extent.
Positron emission tomography in migraine
An early PET-study identified dorsal parts of the brainstem to be, what was then long called “the migraine generator”. Cortical areas also showed activation during the spontaneous attacks, but only brainstem activation was persistent after the termination of the migraine attack using sumatriptan SC. This led to the conclusion that activation of the brainstem is rather the cause than the consequence of the migraine attack, whereas activation of cortical structures rather reflect the perception of pain.
Later studies supported and further substantiated these findings. Barah et al. identified the brainstem region as the dorsal pons using H215O-PET in a glyceryl-trinitrate (GTN) triggered migraine attack. In the following years additional studies reproduced the findings both in GTN- triggered and spontaneous migraine attacks.,, Further studies not only demonstrated brainstem activations but also activations in the midbrain and hypothalamus, which persisted even after the headache relief with sumatriptan, thus pointing towards migraine generating capabilities of these structures.
The important role of subcortical structures was underscored in another water-PET study addressing the premonitory phase of migraine. To interpret these results correctly, however, one has to take into account that a) these attacks were provoked using GTN and b) not only the brainstem activation was shown, but also other subcortical (including hypothalamus) and even other cortical structures were activated during the early premonitory phase.
Functional magnetic resonance imaging in migraine
More recent functional magnetic resonance imaging (fMRI) was applied to further study the interrelationship of different activated brain areas in patients with primary headache disorders. Functional MRI is superior to PET for most purposes, as it has better resolution in time and space and no radioactive tracers are needed. Although skepticism arose that the dorsal pons is the one and only migraine generating structure, it was undisputed that the brainstem is of high importance for migraine pathophysiology. Stankewitz and May started to not only study the migraine attack but to focus on the migraine cycle and cycling behavior of the trigeminal nucleus. They also provided fMRI gathered evidence for a stronger functional coupling between brainstem structures to the hypothalamus following trigeminal nociception. This opened the view for a more network-based understanding for migraine generation, which had already been proposed for cluster headache (CH) at that time. And indeed, in a longitudinal study one single migraineur was investigated every day for one month and hypothalamic activity following trigeminal nociceptive stimulation was already found to be increased prior to the occurrence of migraine attacks. The previously seen alterations in functional coupling of the brainstem-hypothalamus network were shown to be relevant in migraine generation, as this was found to already be altered at the preictal day and during the pain phase of the migraine attacks. Very recently the same authors studied nine patients (seven included in the analysis) with the same protocol. Activation of the hypothalamus was detectable up to 48 hours before headache onset, which was interpreted as a potential marker for the premonitory phase. [Figure 1] and [Figure 2] show exemplary network-wide fMRI activations in patients with migraine at both supratentorial [Figure 1] and brainstem [Figure 2] level.
Resting-state fMRI in migraine
Low-frequency fluctuations can be evaluated using resting state fMRI (rs-fMRI). This technique allows the evaluation of functionally coupled brain areas (also known as resting-state networks), based on synchronized variations of the BOLD (blood-oxygen-level dependent).,,, The method is highly flexible and allows different approaches, of which the independent component analysis (ICA) is the generally considered to be the least biased one by a priori hypothesis, whereas seed-based analyses only appear to be valid if strong hypotheses are available. Network-based approaches may investigate different amounts of networks, of which the three most relevant networks related to pain are (1) the default mode network (DMN), (2) the bilateral central executive network (CEN), and (3) the salience network (SN).
Several rs-fMRI studies were performed in migraine. Xue and colleagues studied 23 migraineurs and controls using dual-regression ICA to investigate the most relevant networks, and found an aberrant intrinsic connectivity within the CEN and SN, and stronger connectivity of the insula cortex in migraine patients. Authors interpreted these findings rather as a consequence of frequent migraine attacks than the particular pathophysiological correlate of migraine origin.
In contrast, a different study published in the same year showed reduced executive resting-state network functional connectivity (FN). This shows that the results of these studies are difficult to interpret and possibly prone to error.
A study investigated 34 subjects (17 vs. 17) compared FN between PAG and other brain areas relevant in nociceptive processing and found stronger connectivity in migraineurs versus controls. Migraine frequency was associated with stronger connectivity with some areas, whereas a significant decrease between the PAG and brain regions of pain modulation was seen.
Other approaches try to measure the regional homogeneity (ReHo). One study found decreased in ReHo in several cortical structures in patients with migraine without aura. Another study focused on the amplitude of fluctuations and found increased higher amplitudes in patients with aura compared to those without.
Other studies investigated patients longitudinally outside and during a migraine attack. A study from Denmark assessed FN in four networks using a seed-based approach and found increased as well as decreased FN between the right thalamus and different contralateral brain regions, but not in the pontine or cerebellar networks.
Recently, longitudinal rs-MRI was performed to explore migraine attack generation in seven episodic migraine patients using a ROI-to-ROI/-voxel approach. Comparing preictal and interictal phases, the nucleus accumbens was found to be highly functionally connected with the dorsal rostral pons, the amygdala, the hippocampus, and the gyrus parahippocampalis. Ictally, the dorsal pons had stronger connection to the hypothalamus than interictally. The authors suggested that changes in dopaminergic centers and within the pontine-hypothalamic network are relevant for attack generation and maintenance.
A promising study used machine-learning to establish rs-fMRI as a biomarker (migraine versus control) and found correct machine-learning-classificationof patients diagnosed with migraine to be more reliable for longer disease durations.
All resting state studies need to be interpreted with care, as the methods are highly flexible, results are not fully congruent and often difficult or not precise enough to interpret. Its role in headache disorders is not well specified yet.
Although it is the most disabling migraine variant, studies of functional imaging in CM are rare. An rs-fMRI study evaluated the intrinsic resting FN of several networks in 29 women with CM compared to controls applying multivariate linear regression.
The three major brain networks showed reduced coherence in CM, but no difference attributed to MOH was found. Allodynia was associated with better coupling in SN, while frequency of migraine days was correlated with reduced coupling in SN and CEN.
A second rs-fMRI study also investigated three selected networks in 20 CMs and 20 healthy controls. Authors found reduced connectivity between the DMN and CEN and between the dorsal attention system (DAS) with the CEN in CM, but stronger connectivity between DAS and DMN. The authors interpreted these findings as evidence for large-scale reorganization of functional cortical networks in CM.
Another fMRI study compared episodic migraineurs (n = 18), CMs (n = 17), and HCs (n = 19) using a task related protocol and reported increased activation of the anterior right hypothalamus in CMs compared to controls, while the more posterior hypothalamic portion activated bilaterally during headaches. In conclusion, the anterior hypothalamus maybe involved in migraine chronification.
Although it is the most frequent primary headache, the data on tension-type headache is very limited. This accounts for pathophysiology in general, as for functional imaging data in particular. Some structural MRI studies have shown changes in the gray-matter, but results were inconsistent., A recent study from China examined regional homogeneity abnormalities in a small cohort (ten patients vs. ten controls) using resting-state fMRI. TTH patients showed decreased regional homogeneity in several brain areas including the bilateral caudate nucleus. Till further studies are available to confirm these findings, these have to be interpreted with care.
CH is a rare headache disorder, but it is by far the most common of the TACs. It is impairing, as the strictly unilateral attacks are of highest intensity and accompanied by ipsilateral autonomic symptoms. Most functional imaging studies on TACs were performed in CH and aimed to confirm the hypothalamus as the driving force or generator of this disorder. This hypothesis is strongly supported by the circadian and circannual rhythmicity that is so characteristic for this primary headache disorder. Early findings involving the hypothalamus led to arguable clinical consequences extending as far as deep brain stimulation (DBS) for strongly affected patients.,,,
Positron emission tomography and functional magnetic resonance imaging in CH
Several investigations using PET have been performed in CH. The first-ever study was done in 1996 in seven episodic CH (eCH) patients. PET was performed during GTN-induced CH-attacks, and altered regional cerebral blood flow (rCBF) was observed in multiple cortical areas (increase and decrease). May et al. used H215O PET in GTN-triggered attacks and were able to show strong activation of the ipsilateral posterior hypothalamus in nine chronic CH (cCH) patients and reconfirmed this finding in 17 patients in 2000. Similar activation was observed in non triggered CH attacks in a single patient receiving DBS. Comparable to migraine, a more recent study that used fMRI, was able to show hypothalamic activation in four patients suffering episodic CH (eCH). As the spatial resolution is limited in fMRI, some authors discussed that the activation was located rather in the midbrain tegmentum, than in the hypothalamus itself. However, multiple studies were able to re-confirm hypothalamic activation, but it turned out that this was not the only site of activation as other parts of human pain processing networks including the insula, cingulate, temporal, and frontal cortex are also activated quite frequently. [Table 1] summarizes the available attack related functional studies of CH.
Resting-state functional magnetic resonance imaging in CH
Rocca et al. studied resting state activity in 13 patients with eCH outside bout compared with healthy controls. The authors observed altered FC within the network from the hypothalamus to the thalamus bilaterally and the sensorimotor cortex as well as the primary visual cortex. Other studies showed an even wider network involved, including the hypothalamus but not restricted to it, in resting-state fMRI of episodic and chronic CH both on the headache and non headache side Episodic-CH patients presented hypothalamic FC changes with the medial frontal gyrus, as well as the occipital cuneus in- and outside of bout. Hypothalamic FC was decreased in patients out-of-bout in the medial frontal gyrus, precuneus, and cerebellum. Interestingly, the bout frequency correlated with the hypothalamic connectivity to the cerebellum. The authors concluded that CH pathophysiology may extend beyond the traditional pain processing networks.
Although it is the second most frequent TAC, paroxysmal hemicrania is rare. It affects 3%–6% of all TAC patients. It has some overlap with cluster headache, as patients have unilateral, severe attacks associated with cranial autonomic features recurring multiple times per day, and usually last between two and 30 min., In contrast to CH, some (10%) attacks may be precipitated mechanically. Alcohol is not a reliable trigger. It is of high importance to distinguish paroxysmal hemicrania from CH, as therapeutic approaches are different. PH shows a reliable response to indomethacin, which is a diagnostic criterion in ICHD3. Again it was possible to transfer clinical observations to functional imaging, and in PET imaging, indomethacin was able to end the previously detected activation in several pain-processing structures. Activation was found in the contralateral posterior hypothalamus, as well as activation in a widespread network (including the ipsilateral lentiform nucleus, anterior and posterior cingulate cortex, contralateral temporal cortex, postcentral gyrus, precuneus, cerebellum and ventral midbrain, bilateral insula, and frontal cortex) in seven patients.
SUNCT (short-lasting unilateral neuralgiform headache attacks with conjunctival injection and tearing) exactly describes the clinical picture of the disease. Similar to trigeminal neuralgia (TN), which is the main differential diagnosis, the attacks are usually very short and occur plenty of times a day. As some patients may have other autonomic features but not conjunctival injection nor tearing the IHS classification also introduced SUNA ( short-lasting unilateral neuralgiform headache attacks with cranial autonomic features). In contrast to TN, SUNCT/SUNA often affects the first branch of the trigeminal nerve and comes without a refractory period. As clinically both disesases have a lot in common , imaging studies also tried to find similarities between them, as well as to other TACs and migraine. As in CH and PH, SUNCT/SUNA patients have high comorbidity with migraine.,, This may either reflect a (genetic) general predisposition for primary headache disorders or, alternatively express a pathophysiological connection between these disorders. In regard to pathophysiology, a shared final stretch of migraine and TACs was previously discussed. It is well possible that one-sided headaches share a common ground for the development of pain and autonomic symptoms, including functional alterations in hypothalamic or brainstem circuits. Interestingly pain intensity positively correlates with the development of autonomic features, as the trigeminovascular reflex is not only expressed in patients but also healthy persons and migraine.
Several fMRI studies have been performed during SUNCT-attacks and showed activation of the posterior hypothalamus.,,, One of these was able to demonstrate additional activation in the cingulate cortex, the insula, temporal, and frontal cortex in a single patient that was classified probable SUNCT. A different investigation detected brainstem activation during three typical SUNCT attacks, as well as activations in the right precentral, superior frontal, inferior frontal, as well as the middle frontal cortex, and bilateral supplementary motor area. As in other headache disorders, the authors suggested a broader concept of network connections within pain processing structures.
The latest revision of the international headache classification (ICHD3) sorted hemicrania continua back to the TACs. As paroxysmal hemicrania, hemicrania continua is defined as being indomethacin-responsive, but it is characterized as a chronic headache with continuous, unilateral head pain that varies in intensity. The headache is frequently associated with autonomic features. A PET study investigated seven patients with HC, both while suffering pain and being relieved after indomethacin administration, and compared these to seven matched control subjects. In pain, activation was found in the posterior hypothalamus, the ipsilateral ventrolateral midbrain, and the dorsal rostral pons. This activation completely resolved after intramuscular indomethacin. At that time the authors concluded that activations were different from migraine, but comparable to other TACs and consequently have to be classified as TAC and hence needs to be clearly distinguished from migraine. Nevertheless, with the more recent studies on migraine, also showing activation of similar structures, this has to be discussed and interpreted more cautiously.
Functional neuroimaging has significantly influenced today's understanding of primary headaches especially in regard to pathophysiology. Results from modern studies clearly point toward a deficient complex neural network rather than a single structure of abnormality, even though it remains undisputed that the hypothalamus and the brainstem are key structures in the pathophysiology of these disorders. Imaging provided significant pathophysiological glimpses, but ultimately is not able to unravel this enigma on its own, as the exact mechanisms of neuronal crosstalk between the hypothalamus and other pain-processing brain regions remain unknown. More sophisticated research (particularly more longitudinal studies) is required to adequately address each aspect of these disorders. The similarities that most TACs share on functional neuroimaging justify their status as unique disease entities in a common subgroup of primary headache disorders, but gross overlap in imaging studies in migraine also demonstrate that probably several subtle differences are not understood yet. The distinction of each individual primary headache syndrome from one another will be the future challenge with most useful results in everyday clinical practice and potentially a way to support the clinical diagnosis using the MRI result as a biomarker.
Financial support and sponsorship
Conflicts of interest
There are no conflicts of interest.
[Figure 1], [Figure 2]