Abstract
Febrile Infection-Related Epilepsy Syndrome (FIRES) is a rare and severe form of new-onset refractory status epilepticus (NORSE) that predominantly affects children and young adults. This case report details the clinical course of a 23-year-old female patient, emphasizing the diagnostic utility of the MRI "claustrum sign" and the application of a novel deep learning framework, Vox2Cortex-Flow, for longitudinal monitoring of cortical swelling. Early intervention with high-dose corticosteroids resulted in significant clinical and radiological improvement, highlighting the importance of prompt diagnosis and the potential of AI-driven neuroimaging in clinical practice.
Introduction
FIRES typically presents in previously healthy individuals following a mild, non-specific febrile illness. The condition is characterized by a catastrophic progression to refractory status epilepticus. Due to an unknown exact pathogenesis—hypothesized to be an infection-triggered inflammatory response—and a lack of specific biological markers, diagnosis is frequently delayed. Such delays are associated with poor long-term outcomes, including chronic epilepsy and cognitive impairment. This report addresses the challenges of early diagnosis and disease monitoring by identifying specific neuroimaging markers and demonstrating advanced image processing techniques.
Methods
The study follows a 23-year-old patient through longitudinal clinical observation, including serial EEG, CSF analysis, and MRI scans at 4 days, 1.5, 2.5, 3.5, and 10.5 weeks post-onset.
To quantify structural brain changes, the researchers utilized Vox2Cortex-Flow, a deep learning-based framework designed for accurate cortical surface estimation with spatiotemporal correspondence. Structural MRI data were registered to a standard MNI152 space, allowing for the generation of 3D cortical surface reconstructions and detailed cortical thickness maps. This approach enabled a precise, quantitative assessment of cortical edema and its subsequent resolution over time.
Results
Diagnostic Markers and Clinical Course
Initial neuroimaging at day 4 was unremarkable. However, by the 1.5-week follow-up, MRI revealed bilateral T2-hyperintensities in the claustrum—the "claustrum sign"—which served as a critical diagnostic beacon for FIRES. This finding prompted a shift in the treatment strategy toward immunotherapy.
Following a 5-day course of high-dose intravenous methylprednisolone (1 g/day), the patient showed rapid clinical improvement. By 10.5 weeks, she was seizure-free with a normal EEG and a MoCA score of 26/30.
Advanced Neuroimaging Evolution
The application of Vox2Cortex-Flow provided a granular view of the disease's structural impact. Cortical thickness maps identified significant swelling in the right frontal and temporal operculum, peaking between weeks 1.5 and 3.5.
By the final follow-up at 10.5 weeks, the quantitative models demonstrated a significant regression of cortical thickness, correlating closely with the patient's clinical recovery and the normalization of standard radiological findings.
Discussion
This case underscores the efficacy of early immunotherapy in FIRES, supporting the inflammatory hypothesis of the disease. The "claustrum sign" remains a vital early radiological marker that can guide clinical decision-making before autoimmune antibody results are available.
Furthermore, the study demonstrates that deep learning tools like Vox2Cortex-Flow can visualize and quantify subtle cortical changes that may be overlooked in standard radiological assessments. Such AI-driven 3D reconstructions offer a powerful means of monitoring disease evolution and treatment response in complex neurological conditions.
Limitations
As a single-patient case report, the generalizability of these findings is limited. The diagnosis lacks a confirmatory biological biomarker, and the patient's long-term prognosis is complicated by pre-existing psychiatric comorbidities and a relatively short follow-up period of five months. Future research should focus on validating these advanced neuroimaging tools across larger cohorts to standardize their clinical application.
Summary updated