This brain 18F-FDG PET/CT voxel-based analysis demonstrates a large and predominantly associative cortical decreased metabolism compared to the striata in patients with DE in comparison with healthy subjects. This decrease in the metabolic ratio between the cortex and the striata accurately identifies patients, and could be used as a biomarker of DE.
In the present study, we retrospectively evaluated a cohort of 73 patients with a diagnosis of DE regardless of their autoantibody status who underwent a complete clinical and paraclinical evaluation including 18F-FDG PET/CT. To our knowledge, this is the largest study evaluating the cerebral metabolic profile using 18F-FDG PET/CT in this setting.
DE is a relatively rare but serious disease with the potential to cause disability. Since immunosuppressive therapeutics have been shown to have a major impact on prognosis, the early diagnosis of DE is essential. However, it is consensually acknowledged that such a diagnosis remains problematic since it is often missed or delayed in patients with heterogeneous and unspecific clinical presentations. Furthermore, standard additional tests may be within normal limits in a significant proportion of patients. Accordingly, in our population of 73 patients with a validated diagnosis of DE, the analysis of cerebrospinal fluid was normal in 21 patients (29%). Similarly, in a recently published study evaluating the prevalence of CSF markers associated with inflammation in patients with active DE, Hebert el al.  identified 26/95 (27%) patients with white blood cell counts and protein levels within normal limits, comparable to the values found in our study. Likewise, brain MRI showed a mesial temporal lobe T2-FLAIR hypersignal in only 23 of our patients (32%) as well as some non-specific findings. The suboptimal sensitivity of brain MRI has been described in several studies [36, 45]. Probasco et al.  found only 23 MRIs (40%) consistent with the diagnosis of DE in a retrospective study including 57 dysimmune encephalitis patients evaluated by brain MRI. A meta-analysis by these same authors identified 68/114 (60%) patients for whom MRI was supposedly consistent with DE without describing the image abnormalities justifying such a diagnosis. The rather variable delay before MRI, which can extend up to several years following symptom onset, may have contributed to the decrease in sensitivity. Finally, seronegative DE constitutes a major diagnostic issue, estimated at approximatively 50% of cases  and comparable to what is observed in our population (36/73 patients with seronegative DE). Thus, the identification of a sensitive biomarker would be relevant in establishing the diagnosis of DE.
The role of brain 18F-FDG PET/CT in the currently used diagnostic criteria  is restricted because it represents an alternative to brain MRI for the diagnosis of definite encephalitis only, and because it is limited to the demonstration of temporal lobe hypermetabolism. Graus and colleagues underlined two major drawbacks for the use of brain 18F-FDG PET/CT, namely, the lack of large studies with homogeneous patient cohorts and the lack of image standardization using more objective semi-quantitative measures.
Recent studies [39, 46] have shown a significant diagnostic contribution of semi-quantification in comparison to visual analysis alone in this context, leading to more comparable and accurate results. For example, Lv et al. identified striatal hypermetabolism upon semi-quantitative analysis in 73% of patients who did not present with a visual metabolic striatal abnormality. Indeed, visual analysis is a subjective process that depends on the physician’s experience, and has limitations in evaluating hypermetabolism. Visual analysis requires auto-windowing of the image according to a reference region with the greatest 18F-FDG uptake (e.g., the striatum). This point may partly account for the great heterogeneity of the metabolic profiles described in the literature [20, 24, 36], since most of these descriptions were based on qualitative analysis in which some hypometabolism probably corresponded to a relative decrease in cortical metabolism in comparison with the striatum or limbic hypermetabolism.
Result heterogeneity was also observed among studies that investigated DE by means of a semi-quantitative analysis. Such heterogeneity may be explained by methodologies not suited for the combined detection of hypometabolism and hypermetabolism in DE patients. More precisely, some investigators have carried out their analysis following proportional scaling [32, 34, 35], while others have used intensity normalization on a region of interest . However, it is now acknowledged that the intensity normalization method could lead to poor detection of hypometabolic areas or to the artefactual identification of hypermetabolism in the best-preserved areas, as demonstrated in the field of dementia.
Based on 3 cases of dysimmune encephalitis with anti-VGKC antibodies, Rey et al.  detected a common 18F-FDG PET/CT pattern of relative striatal hypermetabolism contrasting with diffuse cortical hypometabolism. Moloney et al.  also performed qualitative and semi-quantitative analysis of brain 18F-FDG PET/CT in one patient over the course of the symptomatic phase of VGKC encephalitis. They identified an increase in the metabolic gradient between the striatum and cortex in semi-quantitative analysis and hypothesized that such a pattern may be specific for encephalitis, at least in the setting of VGKC encephalitis. In a cohort of 24 patients with anti-NMDAR, anti-GAD and anti-VGKC encephalitis, Tripathi et al.  identified heterogeneous metabolic profiles, with more often cortical parieto-occipital hypometabolism and striatal hypermetabolism, on semi-quantitative analysis.
In our study, we hypothesized that the metabolic gradient between striata and the cortex may have diagnostic value in DE. Therefore, we parameterized the intensity of all 18F-FDG PET/CT images on striatal activity. In comparison with a set of healthy control subjects, voxel-based statistical analysis identified a widespread cortical decreased metabolism compared to the striata, predominant on the associative cortex and essentially preserving the rolandic and temporal limbic regions. This aspect confirms a decrease in the metabolic gradient between the cortex and the striata in patients with DE, regardless of the type of antibody.
The ratio of cluster to striatal metabolic values (cluster metabolic ratio) was significantly lower in patients with DE than in controls. The cluster metabolic ratio effectively classified 97.4% of patients and controls in the discriminant analysis, with only three patients misclassified due to a ratio comparable to that of controls. All DE patients (12/73) with a negative standard paraclinical evaluation had a lowered cluster metabolic ratio. This metabolic gradient was more frequently affected than other paraclinical parameters evaluated. 18F-FDG PET/CT with semi-quantitative analysis therefore appears to be an effective biomarker in DE.
A lower cluster metabolic ratio was significantly correlated with longer-established disease as evidenced by the link with a longer delay between the onset of symptoms and the diagnostic. This association is consistent with previous studies in which the authors visually described cortical hypometabolism in advanced encephalitis . Similarly, the link with the absence of identification of antibodies could also involve delayed management, and worse prognosis. In this line, this cluster metabolic ratio also was associated witha higher risk of death.
Beyond its diagnostic role, brain 18F-FDG PET/CT likely has the potential for the assessment of therapeutic responses associated with an improvement in the cortex/striatum ratio [22, 23, 34] and for the identification of recurrence . Further studies are required to elucidate the potential of 18F-FDG PET/CT for these indications.
There are several limitations to the present study. First, this is a single-centre retrospective study performed in a tertiary referral hospital, and there is a risk of selection bias. A significant difference in disease duration prior to 18F-FDG PET/CT was noted, and some patients were evaluated years after the onset of symptoms. The observed metabolic patterns are probably variable depending on the stages of the disease. However, this variable mirrors clinical practice with a diagnosis that remains difficult and sometimes delayed.
Future prospective studies are now warranted to confirm the diagnostic performance of the decreased cortex/striatum metabolic ratio in DE patients. The currently used Graus criteria lack sensitivity, and this justifies the development of new diagnostic criteria integrating 18F-FDG PET/CT.