The diagnostic value of an automated analysis of structural brain volumes in MRIn-MTLE patients

Abstract

Abstract Background To explore the alterations of brain region volumes in magnetic resonance-negative temporal lobe epilepsy patients, we constructed a classification model based on volume parameters and analyzed its classification efficacy. Methods T1-weighted images of magnetic resonance-negative temporal lobe epilepsy (MRIn-MTLE) patients and healthy controls were retrospectively analyzed, and the volumes of brain regions were segmented and calculated using FreeSurfer software to compare ipsilateral volumes among patients of different gender. Feature Explorer software based on Pyradiomics was used to construct a classification model based on volume parameters and analyze the classification efficacy. Results There were differences in the volumes of several brain regions on the left and right sides in both the heathy control (HC) group and MRIn-MTLE patients; these differences were significant (P < 0.05). In males, the estimated total intracranial volume(eTIV) and volumes of pars opercularis, rostral middle frontal gyrus, caudal middle frontal gyrus, superior frontal gyrus of left hemisphere in HC group were larger than in MRIn-MTLE group; in females, the eTIV and volumes of caudal middle frontal gyrus, precentral gyrus, post central gyrus of left hemisphere and caudal middle frontal gyrus, isthmus of cingulate gyrus, posterior cingulate gyrus, precentral gyrus, temporal pole, inferior temporal gyrus, cuneus, entorhinal cortex of right hemisphere were smaller in MRIn-MTLE patients compared with the HC group (P < 0.05). The model’s classification efficacy reached 0.780 AUC, and accuracy of 0.721. Conclusions MRIn-MTLE patients have volume reductions in multiple brain regions, and these differences differ in male and female, which indicates MRIn-MTLE might has different impact based on gender, our further studies should take gender differences in the volume of brain structures into account, while further investigating the physiological or anatomical basis of the differences. Volume parameters can be used as classification features to construct patient classification models.