Skull phantom-based methodology to validate MRI co-registration accuracy for Gamma Knife radiosurgery

Area of Science:

• Medical Imaging
• Radiosurgery
• Image Registration

Background:

• Stereotactic radiosurgery (SRS) with Gamma Knife relies on accurate co-registration between planning MRI and cone-beam CT (CBCT).
• Emerging functional and metabolic MRI sequences require validation for co-registration accuracy before clinical use.

Purpose of the Study:

• To develop a phantom-based methodology for quantitatively comparing co-registration accuracy.
• To evaluate standard clinical MRI protocols against novel sequences likely to have poor co-registration.
• To assess the clinical translatability of new MRI sequences for SRS target delineation.

Main Methods:

• A human skull phantom with fiducials was used with standard MRI sequences (T1-weighted Dixon, T1-weighted FFE, T2-weighted FLAIR).
• Sequences were 'optimized' and 'compromised' (simulating poor accuracy) and co-registered to CBCT.
• Image similarity metrics (SSIM, MSE, PSNR) were employed for quantitative comparison.

Main Results:

• Median geometric uncertainties were quantified for optimized sequences, with varying accuracy across modalities and axes.
• Mean Squared Error (MSE) and Structural Similarity Index (SSIM) were most predictive of co-registration accuracy.
• Clinically relevant thresholds for MSE, SSIM, and PSNR can guide the rejection of unsuitable MRI sequences.

Conclusions:

• A robust phantom-based methodology was established for quantitative assessment of MRI-CBCT co-registration.
• This method enables evaluation of novel MRI sequences against clinical standards for Gamma Knife SRS.
• The approach facilitates the clinical translation of advanced MRI techniques for improved target localization.