Real-world FDG PET study separates autoimmune encephalitis from mimics

Autoimmune encephalitis is one of the most perplexing conditions in modern neurology, a disease in which the body’s own immune system turns against the brain, producing seizures, memory loss, psychiatric disturbances and, in severe cases, life-threatening neurological collapse. Diagnosing it has always been a race against time, complicated by the fact that its symptoms overlap with a bewildering range of other conditions, from viral infections and neurodegenerative disorders to primary psychiatric illness. Now, a large real-world study published in the European Journal of Nuclear Medicine and Molecular Imaging has provided some of the strongest evidence yet that a well-established nuclear medicine technique, brain positron emission tomography with the glucose analogue [¹⁸F]-FDG, can reliably separate autoimmune encephalitis from its many clinical mimics, offering clinicians a powerful complementary biomarker at a moment when every day of delay matters.

The study, led by Hélène Rostand and Aurélie Kas of the Department of Nuclear Medicine at Pitié-Salpêtrière Hospital in Paris, together with collaborators from reference centres for paraneoplastic neurological syndromes across France, took a deliberately pragmatic approach. Rather than enrolling a carefully curated cohort of textbook cases, the team examined every consecutive patient referred for brain [¹⁸F]-FDG PET/CT with suspected autoimmune encephalitis between January 2020 and July 2023. This real-world design matters enormously, because the true value of any diagnostic test lies not in how it performs on idealised patients but in how it performs in the messy, ambiguous setting of a working hospital. A total of 141 patients, with a mean age of 48 years and a wide age range spanning young adults to the very elderly, were retrospectively included in the analysis.

Establishing a definitive diagnosis for each patient was a rigorous process. Final diagnoses were assigned using international diagnostic criteria for autoimmune encephalitis, combined with expert consensus and careful clinical follow-up, ensuring that the PET findings could be judged against a trustworthy gold standard rather than circular reasoning. The results of this adjudication were sobering in themselves: only 33 percent of the patients ultimately had autoimmune encephalitis, 43 percent turned out to have alternative differential diagnoses, and 24 percent remained undetermined even after comprehensive work-up. Among the encephalitis cases, 70 percent met criteria for definite disease, 63 percent were seropositive for a neural antibody, and only 35 percent had supportive findings on magnetic resonance imaging. These figures starkly illustrate the diagnostic gap that clinicians face: a substantial majority of patients scanned for suspected encephalitis do not have it, and even among those who do, both antibody testing and MRI frequently fail to confirm the diagnosis.

The biological rationale behind using [¹⁸F]-FDG PET in this context rests on decades of neuroscience. The radiotracer [¹⁸F]-fluoro-2-deoxy-D-glucose is taken up by metabolically active cells in proportion to their glucose consumption, and since the brain’s enormous energy demand is dominated by synaptic activity, the PET signal serves as a proxy for regional neuronal and glial function. Foundational work in the late 1970s validated the tomographic measurement of local cerebral glucose metabolic rate in humans, and more recent research has shown that the FDG signal is in fact strongly driven by astroglial glutamate transport, linking the image directly to synaptic chemistry. In inflammatory brain disease, this translates into a characteristic interplay of hypometabolism, reflecting neuronal dysfunction in damaged tissue, and hypermetabolism, which may reflect active inflammatory processes, seizure activity or compensatory network changes.

When the researchers compared the 46 patients with autoimmune encephalitis to 54 healthy controls using voxel-wise and regional analyses, a distinctive metabolic signature emerged with striking consistency. The encephalitis patients showed widespread cortical hypometabolism accompanied by hypermetabolism in a set of deep and medial structures: the mesiotemporal lobe, encompassing the hippocampus and adjacent limbic cortex, the basal ganglia, the cerebellum, the midbrain and the insula. This pattern of simultaneous cortical depression and limbic-plus-extralimbic activation fits neatly with what is known about the pathophysiology of antibody-mediated encephalitis, in which limbic structures are preferentially targeted by autoantibodies against synaptic proteins, while motor and subcortical circuits are frequently engaged. Earlier studies in anti-NMDA receptor and anti-LGI1 encephalitis had hinted at exactly this combination, but the present study confirms it in a consecutive, unselected cohort that includes both seropositive and seronegative disease.

The most clinically consequential finding, however, came from the head-to-head comparison between confirmed autoimmune encephalitis and the patients with differential diagnoses. The two groups differed profoundly: encephalitis patients exhibited lower metabolism in occipital and prefrontal cortical regions but higher metabolism in the mesiotemporal lobe, the insula and the midbrain, with differences reaching high statistical significance. When these regional metabolic measures were tested for their ability to discriminate between the two conditions, the combination yielded an area under the receiver operating characteristic curve of 0.82 to 0.83, indicating good discriminative power in a setting where misdiagnosis carries severe consequences. Mesiotemporal hypermetabolism, whether observed alone or in combination with cortical hypometabolism, proved to be the single most specific feature of the disease. By contrast, several other regional metrics and metabolic ratios that had been proposed in earlier literature, including cortex-to-striatum ratios, performed less reliably in this cohort, with areas under the curve at or below 0.73, a finding that will temper expectations for simpler quantitative shortcuts.

Importantly, the team also evaluated the simplest possible use of the technology: visual reading of the PET images by an experienced nuclear physician, without quantitative post-processing. Here the results revealed a crucial dependence on age. In patients under 40 years, visual analysis achieved positive and negative predictive values of 82 percent and 81 percent respectively, meaning that a positive scan was very likely to indicate true encephalitis and a negative scan reasonably reassuring. In the 40-to-65-year age band, predictive values slipped to 81 percent and 69 percent, and in patients over 65 they fell further to 79 percent and 64 percent. The explanation is almost certainly the age-related confounding of brain glucose metabolism, since normal ageing, prodromal neurodegenerative disease and cerebrovascular pathology all alter the metabolic landscape of older brains, blurring the contrast between diseased and healthy tissue. The practical implication is that the test is most decisive in younger patients, precisely the population in which autoimmune encephalitis is most common and most frequently mistaken for a primary psychiatric disorder.

The stakes of this diagnostic uncertainty are difficult to overstate. Autoimmune encephalitis responds to immunotherapy, and early treatment is strongly associated with better neurological outcomes, whereas the conditions that mimic it, including viral encephalitis requiring antivirals, tumours requiring oncological management, and degenerative dementias requiring supportive care, demand entirely different and sometimes mutually incompatible interventions. Studies of misdiagnosis in adults have shown that a substantial fraction of patients labelled with autoimmune encephalitis actually suffer from other diseases, exposing them to unnecessary immunosuppression while the true illness progresses. Conversely, patients with genuine encephalitis who are seronegative or antibody-negative, a situation that applies to more than a third of cases in this cohort, may be denied timely immunotherapy because the laboratory confirmation never arrives. A metabolic biomarker that works regardless of antibody status, and that can flag disease even when MRI is normal, addresses precisely this vulnerable gap.

The study also strengthens the growing role of FDG PET as a longitudinal tool in this disease. Prior work has demonstrated that serial FDG PET can track the response to immunotherapy and that changes in brain metabolism over time carry prognostic weight, with recovery of cortical metabolism heralding clinical improvement. By establishing robust discriminative thresholds in a real-world population, the new findings give clinicians a firmer foundation not only for initial diagnosis but also for interpreting follow-up scans. The authors’ commitment to reproducibility deserves note: the entire image-processing pipeline, including the voxel-wise and regional analyses stratified by age and treatment status, has been released as a unified, publicly accessible workflow on GitHub, allowing other centres to adopt the methodology directly rather than reconstructing it from the methods section alone. The underlying datasets are available from the corresponding author on reasonable request.

Taken together, the study positions brain [¹⁸F]-FDG PET/CT as a genuinely complementary diagnostic biomarker for autoimmune encephalitis, one that neither replaces antibody testing and MRI nor competes with clinical judgement, but that adds an independent, physiologically grounded layer of evidence in the diagnostically critical early window. Its characteristic signature, cortical hypometabolism paired with mesiotemporal, insular and midbrain hypermetabolism, is now validated against the full spectrum of real-world mimics rather than healthy volunteers alone, which is what makes the high discriminative performance meaningful for daily practice. With diagnostic delays in autoimmune encephalitis still measured in weeks and treatment response closely tied to how quickly immunotherapy begins, the message from this Parisian cohort is clear: in the right patient, particularly the young patient with an unexplained encephalopathic or psychiatric syndrome, a metabolic scan of the brain may be the fastest route to the right diagnosis and, ultimately, the right treatment.