Robust thalamic nuclei segmentation from T1-weighted MRI using polynomial intensity transformation

Area of Science:

• Neuroimaging
• Medical Image Analysis
• Brain Anatomy

Background:

• Accurate segmentation of thalamic nuclei is vital for understanding cognition and neurological diseases.
• Standard T1-weighted MRI offers poor contrast for detailed thalamic segmentation.
• Existing advanced MRI sequences are not widely available in clinical settings.

Approach:

• Introduced histogram-based polynomial synthesis (HIPS) to create white-matter-nulled (WMn) MRI contrast from standard T1w MRI.
• Integrated HIPS into the THalamus Optimized Multi-Atlas Segmentation (THOMAS) pipeline (HIPS-THOMAS).
• Compared HIPS-THOMAS against a convolutional neural network (CNN) and a T1w-adapted THOMAS method across various MRI contrasts, manufacturers, and field strengths.

Key Points:

• HIPS significantly improved intrathalamic contrast and boundary definition on T1w MRI.
• HIPS-THOMAS demonstrated superior accuracy (higher Dice coefficients, lower volume errors) compared to CNN and T1w-THOMAS.
• HIPS exhibited the least inter-scanner variability in phantom studies, proving its robustness.

Conclusions:

• HIPS is an effective preprocessing technique for enhancing thalamic nuclei segmentation using standard T1w MRI.
• The HIPS-THOMAS method offers a robust and accurate solution for neuroimaging studies requiring detailed thalamic analysis.
• This approach expands the utility of readily available MRI data for critical brain structure segmentation.