A coarse-to-fine registration method for three-dimensional MR images.

Area of Science:

• Medical Imaging
• Computer Vision
• Image Processing

Background:

• Multimodal magnetic resonance (MR) image registration is crucial for disease diagnosis and surgical planning.
• Challenges include inconsistent intensity correspondence and large initial displacements in 3D MR volumes.
• Accurate spatial alignment of different MR images is essential for reliable analysis.

Purpose of the Study:

• To develop a robust and accurate method for pairwise 3D MR image rigid registration.
• To address the difficulties posed by large initial displacements and intensity variations.
• To improve the success rate and precision of multimodal MR image alignment.

Main Methods:

• A coarse-to-fine registration strategy is proposed for 3D multimodal MR images.
• Image feature points are extracted for coarse registration using point set alignment to reduce initial displacements.
• Grey histogram calculation within an adaptive region of interest and maximization of mutual information refine registration accuracy.

Main Results:

• The proposed method demonstrates improved registration success rates and accuracy compared to conventional techniques.
• Performance was evaluated on three MR image datasets against representative registration methods.
• Significant improvements were observed, especially in cases with substantial initial displacements.

Conclusions:

• The developed coarse-to-fine registration method effectively handles large initial displacements in 3D multimodal MR images.
• This approach offers enhanced accuracy and a higher success rate for medical image registration tasks.
• The technique holds promise for improving auxiliary disease diagnosis and surgical treatment planning.