A nonrigid registration method for correcting brain deformation induced by tumor resection

Area of Science:

• Medical Imaging
• Computational Anatomy
• Neurosurgery

Background:

• Brain deformation during tumor resection necessitates accurate image registration.
• Existing methods may struggle with incomplete data or complex biomechanical simulations.

Purpose of the Study:

• To develop a nonrigid registration method for aligning preoperative and intraoperative MRI.
• To address brain deformation during tumor resection by incorporating incomplete data and biomechanical modeling.

Main Methods:

• Formulated registration as a three-variable minimization problem (point correspondence, deformation field, resection region).
• Employed a fuzzy assignment matrix for point correspondence and a piecewise linear function for deformation, regularized by a heterogeneous biomechanical model.
• Utilized a nested expectation-maximization framework for simultaneous variable resolution.

Main Results:

• Synthetic data experiments confirmed model simplification improves accuracy.
• Clinical trials on 25 patients demonstrated superior performance over ITK's physics-based method.
• Achieved average accuracy improvements of 2.88 mm (Hausdorff distance) and 1.56 mm (anatomical points).

Conclusions:

• The proposed nonrigid registration method effectively corrects brain deformation during tumor resection.
• The method's ability to handle incomplete data and biomechanical simulation enhances its clinical applicability.