Personalized microstructural evaluation using a Mahalanobis-distance based outlier detection strategy on epilepsy...

Area of Science:

• Neuroimaging
• Radiology
• Medical Statistics

Background:

• Quantitative MRI methods are developing rapidly but face statistical challenges in single-patient comparisons.
• Detecting malformations of cortical development (MCDs) in drug-resistant epilepsies using conventional MRI is difficult.
• Existing methods struggle with the subtle nature of MCDs and statistical complexities.

Purpose of the Study:

• To develop a novel, straightforward voxel-wise statistical method for detecting MCDs.
• To combine quantitative MRI data into a multidimensional space for outlier detection.
• To improve the diagnostic accuracy for MCDs in individual patients.

Main Methods:

• A Mahalanobis-distance-based voxel-wise evaluation was developed.
• Simulations and DTI-eigenvalue data from 45 controls were used to optimize parameters.
• An automatic classification method, fine-tuned with leave-one-out strategy, addressed registration and gyrification artifacts.

Main Results:

• The method successfully identified outlier diffusion profiles in 12 of 13 MCD patients.
• Simulations confirmed the method's sensitivity in detecting abnormal tissue microstructure.
• The approach showed concordance with neuroradiological evaluations and independent calculations.

Conclusions:

• The proposed multidimensional statistical approach is sensitive for pinpointing abnormal tissue microstructure using DTI data.
• This method can aid in the everyday examination of individual subjects with potential MCDs.
• Future extensions incorporating other quantitative MRI modalities may further enhance diagnostic specificity.