Cortical thickness analysis in temporal lobe epilepsy using fully Bayesian spectral method in magnetic resonance...

Area of Science:

• Neuroimaging
• Epilepsy Research
• Statistical Modeling

Background:

• Temporal lobe epilepsy (TLE) is the most prevalent epilepsy type, characterized by cerebral cortex alterations.
• Magnetic resonance imaging (MRI) is crucial for detecting anomalies but generates spatially correlated data unsuitable for standard statistical models.
• This study addresses the challenge of analyzing spatially dependent MRI data to compare cortical thickness in TLE patients and controls.

Purpose of the Study:

• To compare cortical thicknesses in various cerebral cortex regions between TLE patients and healthy controls.
• To account for spatial dependencies inherent in MRI data during analysis.
• To evaluate the efficacy of a novel statistical method for identifying epilepsy-related cortical changes.

Main Methods:

• T1-weighted MRI scans were acquired from 33 TLE patients (19 left, 14 right) and 20 healthy controls.
• Cortical thickness was measured across 68 brain regions using a fully Bayesian spectral method to handle spatial correlations.
• Neural networks were employed to classify TLE patients based on identified cortical thickness alterations.

Main Results:

• Left TLE patients exhibited cortical thinning in regions including the anterior cingulate, orbitofrontal cortex, frontal pole, and temporal pole.
• Right TLE patients showed cortical thinning primarily in the entorhinal area.
• Neural networks achieved high classification accuracy: AUC of 0.939 for left TLE and 1.000 for right TLE versus controls.

Conclusions:

• The study confirms that altered cortical gray matter thickness is a common effect of epileptogenicity in TLE.
• The fully Bayesian spectral method effectively analyzes complex, spatially dependent MRI data.
• These findings highlight the potential of advanced statistical methods in understanding epilepsy-related neuroanatomical changes.