Automatic and efficient MRI-US segmentations for improving intraoperative image fusion in image-guided neurosurgery

Area of Science:

• Neurosurgery
• Medical Imaging
• Image Processing

Background:

• Accurate neurosurgical navigation relies on precise tumor localization and identification of critical surrounding structures.
• Preoperative MRI (preMRI) based neuronavigation accuracy degrades due to intraoperative tissue deformation.
• Intraoperative ultrasound (iUS) offers real-time imaging but faces registration challenges with preMRI due to contrast variations.

Purpose of the Study:

• To develop an automatic and efficient segmentation method for B-mode ultrasound images.
• To enhance the registration process between preMRI and iUS for improved neuronavigation accuracy.
• To utilize segmentations of central cerebral structures (falx cerebri, tentorium cerebelli) as a guiding frame for multi-modal image registration.

Main Methods:

• Automatic segmentation of the falx cerebri and tentorium cerebelli in B-mode ultrasound images.
• Development of a multi-modal image registration approach combining intensity-based methods with structural segmentations.
• Validation using expert-identified landmarks to quantify Target Registration Error (TRE).

Main Results:

• Segmentation of falx and tentorium achieved an average Dice coefficient of 0.74 and Hausdorff distance of 12.2 mm.
• The combined segmentation and registration approach reduced mean TRE from 16.9 mm to 2.2 mm.
• Registration time was significantly reduced from 40.5 s (intensity-based) to 12.0 s (combined approach).

Conclusions:

• Automatic segmentation of cerebral structures in iUS is feasible and effective for improving preMRI-iUS registration.
• The proposed method enhances the accuracy, robustness, and speed of neuronavigation systems.
• This technique holds significant potential for improving surgical precision in neurosurgical interventions.