Abstracts

Rationale: Psychosis is an important comorbidity in temporal lobe epilepsy (TLE), but its pathophysiology is still unclear. A recent machine-learning method can estimate individual’s “Brain-Age” from the T1-weighted MRI, and this Brain-Age gap estimation is expected as a novel individual biomarker in both epilepsy and psychiatric disorders. The aim of this study is to reveal characteristics of Brain-Age in TLE with and without psychosis. Methods: We recruited total 214 patients with TLE (21 with psychosis, 193 without psychosis), and 1196 healthy subjects. After normalization of the T1-weighted MRI using SPM12, the gray and white matter voxel intensities for each subject were considered the MRI features. Using the data of healthy subjects, the Brain-PAD (predicted age difference: predicted age—chronological age) was computed through 10-fold cross validation using a support vector regression model with linear kernel. Finally, the trained predictor model was applied to the patients’ MRI data. Results: In our predictor model, the mean absolute error (MAE) and the mean Brain-PAD in the healthy subjects were 5.28 and 0.12 years respectively. The Brain-PAD in TLE with psychosis (median: 11.22 years, IQR: 9.21) was significantly higher than that in TLE without psychosis (median: 5.48 years, IQR: 11.15) (p=0.008, Mann-Whitney U test). The Brain-PAD in TLE patients was significantly correlated with the onset age of TLE (Spearman’s rs: -0.383, p<0.001), while there’s no correlation with disease duration. Conclusions: The increased Brain-Age gap was associated with TLE, psychosis, and earlier onset age, possibly reflecting initial brain damages or some sort of predisposition. Further validation or longitudinal studies should examine the potential clinical values of the neuroimaging-based Brain-Age gap estimation especially in psychiatric comorbidity of epilepsy. Funding: None.