Fully automated image updating for brain shift compensation after dural opening

Area of Science:

• Neurosurgery
• Medical Imaging
• Computational Anatomy

Background:

• Intraoperative brain shift during open cranial procedures compromises surgical navigation accuracy.
• Preoperative MR (pMR) images become less reliable once the cortical surface is exposed due to brain shift.

Purpose of the Study:

• To develop a fully automated image updating system to address brain shift at the start of open cranial surgery.
• To evaluate the accuracy and efficiency of this automated image updating system.

Main Methods:

• Integrated intraoperative stereovision (iSV) images with a computational model to compensate for volumetric brain shift.
• Utilized a Fast Segment Anything Model for automatic cortical surface segmentation and a two-step registration for feature matching.
• Employed a finite element model to estimate whole-brain deformation and generate updated MR (uMR) images.

Main Results:

• The automated system achieved an overall mean misfit, target registration error (TRE), and point-to-surface distance (PSD) of 2.1 ± 1.2 mm, 1.9 ± 1.0 mm, and 1.6 ± 1.0 mm, respectively.
• This represents a significant improvement compared to pMR images (6.5 ± 1.3 mm, 6.2 ± 1.2 mm, 4.5 ± 1.2 mm).
• Image updating was completed automatically in an average of 3.9 ± 0.6 minutes without user intervention.

Conclusions:

• The automated image updating system effectively compensated for brain shift after dural opening.
• The system demonstrated clinically acceptable accuracy and efficiency with minimal disruption to surgical workflow.
• The technology shows potential for integration into routine open cranial procedures due to its autonomous nature and ease of use.