Evaluating the effective segmentation of human lateral geniculate nucleus.

Area of Science:

• Neuroimaging
• Neuroanatomy

Background:

• Subcortical visual pathway structures like the Lateral Geniculate Nucleus (LGN) are challenging to segment with standard MRI.
• Current LGN segmentation relies on numerous proton density (PD) scans (up to 40) and manual delineation, which is time-consuming.

Purpose of the Study:

• To determine the optimal number of PD scan repeats for accurate LGN segmentation.
• To assess the impact of scan repetition on segmentation efficiency and reliability.

Main Methods:

• Manual segmentation of the LGN by four raters across varying PD scan averages (1-40 repeats) in five participants.
• Comparison of rater segmentations against an expert's manual segmentation of 40PD averages.
• Evaluation of an automated segmentation method against manual segmentation results.

Main Results:

• LGN segmentation accuracy, measured by mask overlap, improved with PD scan averages up to 16 repeats.
• No significant improvement in segmentation accuracy was observed beyond 16 PD repeats.
• Automated segmentation performance was comparable to manual segmentation reliability.

Conclusions:

• Optimizing MRI protocols by using 16 PD scan repeats can enhance efficiency without compromising LGN segmentation accuracy.
• Reducing scan time through protocol optimization allows for more functional MRI or cost savings.
• Automated segmentation offers a reliable alternative to manual delineation for LGN structures.