Semiautomatic registration of pre- and postbrain tumor resection laser range data: method and validation

Area of Science:

• Neurosurgery
• Medical Imaging
• Computer-Aided Surgery

Background:

• Accurate image registration is crucial for surgical navigation and understanding brain shift during tumor resections.
• Existing fully automatic methods struggle with the significant differences between pre- and post-resection surgical images.
• Brain shift, the deformation of brain tissue during surgery, complicates accurate image-guided interventions.

Purpose of the Study:

• To present a semiautomatic method for registering pre- and post-resection 3-D laser range scanner (LRS) images.
• To enable accurate computation of brain shift by establishing correspondences between pre- and post-operative data.
• To overcome the challenges posed by large inter-image differences in surgical settings.

Main Methods:

• A semiautomatic approach using vessel identification and registration.
• Optimal path finding algorithm utilizing image intensity features to identify vessel segments.
• Robust point-based nonrigid registration algorithm applied to identified vessels.
• Application of the computed transformation to the entire 3-D LRS data for complete image registration.

Main Results:

• The method demonstrates robustness against operator errors in point localization.
• Quantitative evaluation across ten surgical cases achieved submillimetric registration accuracy.
• Successful establishment of complete correspondence between pre- and post-resection 3-D LRS data.

Conclusions:

• The presented semiautomatic method provides accurate and robust image registration for surgical applications.
• This technique is a valuable component for systems designed to compute brain shift.
• The approach effectively addresses the difficulties associated with registering significantly different surgical images.