Wernicke's encephalopathy (WE) is an acute neurologic disorder resulting from thiamine (vitamin B1) deficiency. WE was first described by Carl Wernicke [1] in 1881 as “superior acute hemorrhagic poliencephalitis” in two men with alcoholism and in a woman affected by pyloric stenosis, whereas the association of WE with thiamine deficiency was first suspected in the 1940s [2]. The exact prevalence and incidence of WE are unknown, but necroscopy studies performed in adults have revealed incidence rates ranging from 0.5% over a 5-year period in Norway to 2.8% over a 9-year period in Australia [3–7]. A review of pediatric WE cases revealed that the frequency of WE in children appears to be broadly similar to that reported in adults [8]. Autopsy studies have consistently shown that the diagnosis of WE is often made only postmortem, particularly when patients present with atypical clinical manifestations [3–7, 9]. For instance, in patients with alcoholism and AIDS, WE diagnosis is missed in as many as 75–80% of all patients [10].

Traditionally, the clinical diagnosis of WE rests on the classical triad consisting of ocular signs, altered consciousness, and ataxia, already described by Wernicke [1] in his original article. Ocular signs associated with WE include nystagmus, bilateral lateral rectus palsies, and conjugate gaze palsies reflecting involvement of the oculomotor, abducens, and vestibular cranial nerves nuclei. However, subsequent studies have revealed that this triad occurs in only 16–38% of all patients with WE, which explains at least in part why WE is often clinically underdiagnosed [3, 11, 12]. In view of the poor diagnostic performance of the classical triad, new classification criteria have been proposed. These criteria require two of four items including dietary deficiencies, oculomotor abnormalities, cerebellar dysfunction, and an altered mental state or mild memory impairment [13].

The pathogenesis of WE is thought to be related to thiamine deficiency, and, conversely, the prognosis of WE critically depends on the time of onset of thiamine supplementation [3]. Thiamine is needed by the cell membranes to sustain osmotic gradients but is also involved in glucose metabolism and in neurotransmitter synthesis. For healthy individuals, the daily thiamine requirement, which depends on the carbohydrate intake, is in the range between 1 and 2 mg. Because the body's reserves of thiamine are only 30–50 mg, the reserves would be completely depleted in 4–6 weeks in the absence of thiamine intake.

Many clinical conditions can impair the correct absorption of an adequate amount of thiamine, including chronic alcohol abuse [5, 14], gastrointestinal surgery [15–17], prolonged vomiting, chemotherapy, systemic infectious and noninfectious diseases, and dietary unbalance [12]. Alcoholism does not directly cause thiamine deficiency, although it may induce such deficiency because of its frequent association with malnourishment. More specifically, the low thiamine absorption rate at the mucosal level, the impaired hepatic function, and the alcohol-related raised thiamine metabolism together may lead to the development of chronic thiamine deficiency [14]. Thiamine-deficient membranes are unable to maintain osmotic gradients, which results in the swelling of intra- and extracellular spaces. Pathologic features are represented by edema, spongy degeneration of the neuropil, neuron sparing, swelling of capillary endothelial cells, and extravasation of RBCs [18].

In WE, the blood–brain barrier is defective in the periventricular regions, in which there is a high rate of thiamine-related glucose and oxidative metabolism [19]. MRI usually shows symmetric signal intensity alterations in the thalami, mamillary bodies, tectal plate, and periaqueductal area [12]. Signal intensity alterations in the cerebellum, cerebellar vermis, cranial nerve nuclei, red nuclei, dentate nuclei, caudate nuclei, splenium, and cerebral cortex represent atypical MRI findings [12, 18, 20–34] (Fig. 1 and Table 1). Atypical MRI findings are always found in association with the classical neuroradiological presentation. In the acute setting of WE, the cytotoxic edema can appear on both CT and MR images as symmetric hypodensity [35] and signal intensity alterations [12], respectively.

In the acute setting of WE, the sensitivity of the brain CT (Figs. 2A and 2B) is low compared with MRI. Antunez et al. [35] reported low-density alterations along the third ventricle walls in only two of 15 (13%) patients affected by WE during the acute phase of the disease. Those authors did not find any alteration in the periaqueductal area. The study protocol included 8-mm-thick slices parallel to the orbitomeatal plane. Antunez et al. concluded that CT is not useful in the diagnosis of WE. However, although no studies have formally investigated the role of CT perfusion protocols in comparison with MRI, we hypothesize that the application of such protocols might improve the diagnostic accuracy of CT in detecting WE. Recently, a patient affected by WE showing CT hypodensity of the fornices was described as an atypical case [36]. To our knowledge, there are no reports on the capacity of CT to show one of the most distinctive neuroradiological findings of WE—that is, cytotoxic edema of the mamillary bodies.

Reversible cytotoxic edema is considered the most distinctive lesion of WE, and it is easily shown on MR images. Typical findings are represented by symmetric alterations in the thalami, mamillary bodies, tectal plate, and periaqueductal area. High-signal-intensity alterations on long-TR spin-echo images are the most frequent pathologic findings seen on MRI compared with low-signal-intensity alterations on short-TR spin-echo images (88% vs 31%, respectively) [12]. Among seven patients affected by WE with evidence of pathologic MRI findings, only two had low-signal-intensity alterations on short-TR images. On the contrary, four patients showed high-signal-intensity alterations on long-TR images and four showed contrast enhancement [37]. Brain images of alcoholic patients with WE during the acute phase of the disease may differ from those of nonalcoholic patients. In fact, in patients with alcoholism, atrophy of the mamillary bodies, infratentorial regions, supratentorial cortex, and corpus callosum may be found in association with the alterations typical of WE [35, 38]. In contrast, signal intensity alterations in nonalcoholic patients likely represent the first thiamine-related metabolic breakdown. For these reasons, no atrophy is found in nonalcoholic patients during the acute phase of the disease or at follow-up [12, 26].

There are few data regarding the appearance of WE lesions on diffusion-weighted imaging (DWI) and apparent diffusion coefficient (ADC) transformation during the acute phase of the disease. DWI can detect changes in the diffusion of water molecules and may also show early ischemic changes in brain tissue [58, 59]. WE alterations typically show restricted diffusion, suggestive of cytotoxic edema [60]. However, DWI may show only slightly increased signal intensities within the thalami and periaqueductal area bilaterally but no signal intensity alterations in the other brain regions or no signal intensity abnormalities of the affected brain areas (Figs. 4C and 4D, and Fig. 5B). Thus, DWI may not be sensitive enough to reveal WE lesions [26]. The role of ADC in the diagnosis of WE is unclear because various patterns of decreased, normal, or increased ADC values have been described in WE [22, 60–63].

We know of only two reports on MR spectroscopy in humans [24, 62]. In one case, the authors showed low levels of N-acetylaspartate/creatine in the thalamus and cerebellum and a lactate peak in the cerebellum that did not resolve completely, suggesting tissue necrosis. In the other case, a remarkable lactate increase in the thalami represented the only alteration. This increase was attributed to the presence of anaerobe oxidation and thiamine deficiency worsened by an excessive parenteral dextrose load. To date, MR functional images do not have a clinical prognostic impact.

The clinical diagnosis of WE traditionally rests on the presence of the classical triad (ocular signs, altered consciousness, and ataxia) described by Wernicke [1] in his original article. However, this classical clinical triad is found, in fact, in only a minority of WE patients [3, 11, 12]. As a result, WE is often clinically underdiagnosed, particularly when patients present with atypical clinical manifestations or have no history of alcohol intake. New criteria have been proposed in an attempt to capture more effectively the spectrum of WE clinical manifestations [13], but it is unclear how these criteria perform in settings different from that in which they were generated and whether they have found wide acceptance in clinical practice.

Neuroimaging studies are powerful tools in supporting the diagnosis of WE and can also help to distinguish WE from other neurologic disorders, especially in comatose patients. Long-TR MR images are the most sensitive sequences. Symmetric alterations in the thalami, mamillary bodies, tectal plate, and periaqueductal area represent typical lesions of WE, but atypical lesions may also been seen. As we have shown here, atypical changes of WE almost always have been described in nonalcoholic patients and only in association with the typical alterations with the exception of two alcoholic patients showing cerebellar involvement associated with the characteristic findings of the disease [6, 32]. The reasons why specific brain areas are affected by WE are poorly understood, but we speculate that the brain areas characterized by intense thiamine metabolism would be those that appear to be typically involved on MRI. On the other hand, atypical MRI findings of WE, which occur in nonalcoholic patients only, are very similar to those of metronidazole-induced encephalopathy, suggesting that WE and metronidazole-induced encephalopathy may share common metabolic pathways [56, 57]. Because atypical lesions have been reported only in nonalcoholic patients with WE, it would be tempting to surmise that alcohol may have a protective effect on the brain areas that show atypical lesions in WE, although at the present this hypothesis remains a matter of speculation.

Turning to other applications of MRI, the roles of DWI and ADC and of MR spectroscopy are still unclear. Contrast-enhanced MRI is usually not required; however, in patients in whom there is a clinical suspicion of WE but no lesions on unenhanced MR images, gadolinium-based contrast material should always be administered because contrast enhancement of the mamillary bodies may be the only sign of WE [12, 50].

In conclusion, a thorough knowledge of the neuroimaging findings of WE may assist in making an early diagnosis and thus in reducing the morbidity and mortality associated with this disease. Close cooperation between radiologists and physicians is required, particularly when patients present without clear-cut clinical manifestations or with no history of alcohol intake.

Address correspondence to G. Zuccoli ([email protected] or [email protected]).