Joint Analysis for Faciobrachial Dystonic Seizures Identification in LGI1 Encephalitis Using FDG-PET

Leucine-rich Glioma-Inactivated-1 (LGI1) is one notable variant of Autoimmune encephalitis (AE). Some LGI1 AE patients show Faciobrachial Dystonic Seizures (FBDS) symptoms. Positron Emission Tomography (PET) is an essential tool for LGI1 AE with FBDS symptoms detection, which shows hypermetabolism in the basal ganglia (BG) and medial temporal lobe (MTL) regions. Previous studies mainly focus on these two localized brain regions in clinical diagnosis, which lacks a comprehensive evaluation of the entire brain. Moreover, most studies have not conducted detailed classification diagnoses for different FBDS patients. To address these problems, a joint analysis framework is proposed for FBDS identification in LGI1 AE patients. It conducts a joint analysis of machine learning and statistical analysis results with PET data. The radiomics image features from PET images and the parameters for measuring tracer uptake from standardized uptake value ratio (SUV) parameter maps are extracted and analyzed comprehensively. Three binary classification tasks are carried out to evaluate the performance of the proposed method, i.e. health control vs. LGI1 AE patients without FBDS, health control vs. LGI1 AE patients with FBDS and health control vs. LGI1 AE patients with unilateral FBDS. Experimental results show that the AUC values of the significant brain regions in three classification tasks are all above 95%, and the consistency ratio of the significant brain regions selected by the proposed method exceeds 80%. Except for BG, MTL and temporal gyrus are crucial regions in LGI1 AE diagnosis, some ROIs related to motor function such as the cerebellar hemisphere play an important role for LGI1 AE patients with FBDS. These results will be useful for further exploring the role of the cerebellar hemisphere in the identification of LGI1 AE patients with FBDS.