Automated subcortical volume estimation from 2D MRI in epilepsy and implications for clinical trials

Area of Science:

• Neuroimaging
• Medical Image Analysis
• Epilepsy Research

Background:

• Most automatic subcortical segmentation techniques rely on 3D MRI, limiting analysis of 2D or non-isotropic images common in non-specialist settings.
• Quantitative analysis of subcortical brain regions is crucial for understanding neurological conditions like idiopathic generalized epilepsy (IGE).

Purpose of the Study:

• To evaluate a modified segmentation technique for 2D MRI scans in patients with IGE.
• To assess the feasibility and consistency of automated subcortical volume estimation using 2D versus 3D MRI data.

Main Methods:

• Compared volume estimates from 2D (0.4x0.4x3mm) and 3D (1x1x1mm) T1-weighted MRI in 31 IGE patients and 39 controls.
• Utilized a modified FSL FIRST pipeline for 2D image segmentation, including reorientation, cropping, interpolation, and brain extraction.
• Assessed segmentation consistency using Dice coefficients and compared volumes between patient and control groups, also examining the effect of increased slice thickness (6mm).

Main Results:

• Excellent agreement (Dice 0.86-0.96) was found between 2D and 3D segmentations for subcortical structures.
• 2D volumes were slightly lower than 3D volumes; increased slice thickness reduced agreement (Dice 0.55-0.83).
• Significant thalamic and putamen volume reduction was observed in IGE patients compared to controls on both 2D and 3D images.

Conclusions:

• Automated subcortical volume estimation from 2D MRI (0.4x0.4x3mm) using the modified FIRST pipeline is consistent with 3D MRI findings.
• Consistency decreases with increased slice thickness, highlighting the importance of in-plane resolution >1mm x 1mm.
• This 2D segmentation method is feasible for quantitative analysis in IGE studies, offering opportunities for clinical trials and broader research.