Erfan Bashar
Back to Autoimmune Encephalitis

Autoimmune Encephalitis — Diagnosis

Updated:
~4 min read
Last medically reviewed:
On this pageTable of contents

Educational scope notice: This is a study note for medical students, not medical advice, diagnosis, or treatment guidance. Clinical management should follow local protocols and current guidelines.

Diagnosis uses three investigations together — brain MRI, cerebrospinal fluid (CSF) analysis, and EEG — plus antibody testing and tumour screening. No single result confirms or excludes the disease on its own, and a normal MRI or a negative early antibody panel does not end the workup when the clinical picture is suggestive.

MRI: the bilateral limbic pattern

The characteristic finding is bilateral medial temporal lobe hyperintensity on T2 and FLAIR sequences, involving hippocampi and amygdalae. This is the limbic encephalitis pattern. Symmetry matters for the differential: herpes simplex virus encephalitis favours the temporal lobes too but is often unilateral or asymmetric, while stroke follows a vascular territory and is almost always unilateral. Bilateral temporal changes therefore argue against stroke and toward encephalitis, with CSF testing separating autoimmune from viral causes.

CT is less sensitive than MRI for these inflammatory changes, and additional findings can include seizure-related diffusion restriction, thalamic signal change, or gyral enhancement. A normal MRI is common early and never excludes autoimmune encephalitis.

Fluorodeoxyglucose PET measures regional glucose metabolism. Affected temporal regions usually show hypometabolism, though hypermetabolism can appear early when active inflammation increases activity, so a hot temporal lobe should not dismiss the diagnosis. PET helps in two situations: when MRI is normal or equivocal but suspicion remains, and as a whole-body tool for finding an occult tumour driving a paraneoplastic process.

CSF: inflammation, infection, and intrathecal synthesis

Typical CSF shows a mononuclear pleocytosis with raised protein and normal glucose, improving with successful treatment. Normal cell counts never exclude the disease: most LGI1 cases lack pleocytosis. The decisive early test is viral PCR for pathogens including herpes simplex virus, varicella-zoster virus, Epstein-Barr virus, cytomegalovirus, human herpesvirus 6, JC virus, and enterovirus. A positive PCR redirects care toward infection; a negative panel in a suggestive picture supports immunotherapy.

Oligoclonal bands reveal whether antibody production is happening inside the nervous system. Simultaneous isoelectrofocusing of CSF and serum gives five patterns:

PatternCSF bandsSerum bandsInterpretation
Type 1NoneNoneNormal, no intrathecal inflammation
Type 22 or more isolated bandsNoneIntrathecal synthesis, CNS-specific production
Type 3Identical bands plus extra CSF bandsIdentical bandsIntrathecal synthesis with a systemic component
Type 4Identical bands (mirror pattern)Identical bandsSystemic inflammation leaking across a disrupted barrier, no proven local production
Type 5Monoclonal bandMonoclonal bandSystemic monoclonal gammopathy

Type 2 is the most specific for a CNS-limited process, while the Type 4 mirror pattern must not be mistaken for local synthesis. These patterns indicate immune activity inside the brain; they do not identify which antibody is responsible.

The Link index (IgG index) quantifies the same idea: (CSF IgG divided by serum IgG) divided by (CSF albumin divided by serum albumin). Albumin comes only from the liver, so normalising to albumin corrects for barrier leakage. A commonly used cutoff above 0.7 suggests intrathecal IgG synthesis, with exact thresholds depending on the local assay.

Antibody testing: serum plus CSF

Testing covers antibodies against surface antigens (NMDA receptor, LGI1, CASPR2, GABA-B receptor, AMPA receptor) and intracellular onconeural antigens (Hu, Ri, Yo, CV2, Ma, amphiphysin, GAD, among others), ideally in both serum and CSF. CSF testing is more sensitive for NMDAR antibodies, while serum testing catches LGI1 and CASPR2 disease more reliably. A surface-antibody result found only in CSF strongly supports the diagnosis, while a serum-only result carries less certain significance. The antibody refines the syndrome, the expected tumour association, and the prognosis, but treatment starts on clinical grounds without waiting for it, following the Graus criteria for possible and probable autoimmune encephalitis. Graus defines possible and probable disease on clinical and paraclinical grounds, so treatment can start before antibodies return. Definite disease usually requires the antibody result.

EEG: supportive, occasionally distinctive

Common findings are diffuse delta and theta slowing and temporal epileptiform discharges, both consistent with encephalitis without specifying its cause. One distinctive pattern is extreme delta brush — rhythmic delta activity with superimposed fast beta activity — which is suggestive of anti-NMDA receptor encephalitis when present but not specific to it. It appears in only a subset of those patients, so its absence means nothing.

Tumour screening and surveillance

Screening combines tumour markers, contrast CT of the body, and whole-body FDG-PET, guided by the antibody: anti-Hu points toward small-cell lung cancer, anti-Yo toward breast or ovarian cancer, anti-Ma2 toward testicular tumours. When initial screening is negative but a high-risk antibody or phenotype keeps suspicion alive, updated paraneoplastic criteria advise repeating screening every 4–6 months for 2 years, adapted to the presentation, because the tumour can declare itself later.

Evidence anchors

Suggest a correction