# Towards an optimal diagnostic and prognostic model based on semi-quantitative assessment of 18F−FDG PET in children with autoimmune encephalitis

## TL;DR

This study develops a PET-based model to improve diagnosis and predict outcomes in children with autoimmune encephalitis.

## Contribution

A novel semi-quantitative PET model is proposed for accurate diagnosis and prognosis prediction in pediatric autoimmune encephalitis.

## Key findings

- The PET diagnostic model achieved 91.4% sensitivity, 85.1% specificity, and 88.8% accuracy.
- Male gender, high CASE score, memory dysfunction, and low SUVRmaxL/T ratio predict poor prognosis in AE.
- PET-based models outperform visual analysis in diagnosing autoimmune encephalitis.

## Abstract

The metabolic pattern in autoimmune encephalitis (AE) has been frequently reported. Through this semi-quantitative analysis, we aim to explore a practical diagnostic model based on positron emission tomography (PET) for timely diagnosis of pediatric AE with high accuracy. Moreover, we aim to identify factors that affect the prognosis of pediatric AE and explore the utility of PET as a prognostic biomarker.

Data were collected from 93 AE patients and 67 non-AE patients (age range: 1-18 years old). Semi-quantitative parameters of 18F-FDG PET imaging were evaluated, including the score of cortical lesion extent and the ratios of lesion-to-basal ganglia and thalamus. The Clinical Assessment Scale in Autoimmune Encephalitis (CASE) was used to rate the disease severity and long-term outcome. Multivariate statistical analysis was used to establish a diagnostic model and analyze predictors.

The diagnostic model includes three PET parameters. The sensitivity, specificity, and accuracy of the model are 91.4%, 85.1%, and 88.8%, respectively. Participants were followed up for a median of 34 months. Logistic regression analysis indicated that male, initial CASE score >4.5,memory dysfunction, and the ratio of the maximum SUV of the lesion to thalamus (SUVRmaxL/T) < 0.577 are independent factors associated with poor prognosis in AE. We established a prognostic model through these predictors.

18F-FDG PET plays a vital role in the diagnosis and prognosis of AE. The PET-based diagnostic model has higher specificity and accuracy than visual analysis. The prognostic model is a useful predictive tool for the long-term prognosis of children with AE.