A framework for evaluating correspondence between brain images using anatomical fiducials.

Area of Science:

• Neuroimaging
• Medical Image Analysis
• Computational Anatomy

Background:

• Accurate spatial correspondence between brain images is vital for neuroimaging studies and clinical applications like stereotactic neurosurgery.
• Current methods often lack robust quantitative approaches for precise landmark placement.

Purpose of the Study:

• To propose and validate a set of 32 anatomical fiducials (AFIDs) for quick, accurate, and reliable placement on human brain MRI.
• To demonstrate the utility of AFIDs for evaluating image registration accuracy.

Main Methods:

• Utilized publicly available brain templates and individual datasets.
• Trained novice users to place 32 AFIDs with millimetric accuracy.
• Applied AFIDs to assess subject-to-template and template-to-template registration.

Main Results:

• Novice users achieved millimetric accuracy in placing AFIDs after training.
• AFID point-based measures proved more sensitive to focal misregistrations than standard voxel overlap metrics.
• Demonstrated the protocol's utility in evaluating registration accuracy.

Conclusions:

• The proposed AFIDs protocol offers a reliable and accurate method for spatial landmark identification in brain imaging.
• AFIDs provide a valuable quantitative tool for assessing image registration, outperforming traditional metrics.
• This open-source, transparent protocol benefits neuroanatomy education and various brain image alignment applications.