Co-Sparse Analysis Model Based Image Registration to Compensate Brain Shift by Using Intra-Operative Ultrasound...

Area of Science:

• Medical Imaging
• Neurosurgery
• Image Registration

Background:

• Image-guided neurosurgery systems are widely used but limited by brain shift, a deformation of brain tissue during surgery.
• Intra-operative ultrasound (iUS) can compensate for brain shift by updating patient coordinates through registration with pre-operative MRI data, though this is challenging.

Purpose of the Study:

• To propose a non-rigid multimodal image registration technique to accurately register intra-operative ultrasound and pre-operative MRI data.
• To address the challenge of compensating for brain shift in image-guided neurosurgery.

Main Methods:

• A novel non-rigid multimodal image registration technique based on a co-sparse analysis model was developed.
• This model captures interdependencies between MRI (intensity image) and iUS (depth image) modalities.
• A bimodal pair of analysis operators was learned by optimizing a joint co-sparsity function using conjugate gradient to minimize transformation between modalities.

Main Results:

• The proposed co-sparse analysis model achieved a mean registration error of 1.83 mm across a seven-patient dataset.
• Quantitative validation demonstrated superior performance compared to several state-of-the-art registration methods.
• The accuracy of non-rigid multimodal medical image registration was improved by 15.37% compared to curvelet-based residual complexity.

Conclusions:

• The developed intensity-based co-sparse analysis model effectively addresses the challenge of brain shift in neurosurgery.
• The technique offers improved accuracy in non-rigid multimodal image registration with clinically compatible computation times.
• This approach enhances the precision of image-guided neurosurgery by accurately registering diverse imaging modalities.