Patch-based field-of-view matching in multi-modal images for electroporation-based ablations

Area of Science:

• Medical Imaging
• Image Registration
• Interventional Radiology

Background:

• Multi-modal imaging (CBCT, CT, MR) provides complementary data in interventional workflows.
• Accurate spatial alignment of diverse imaging modalities is essential for merging functional and structural information.
• Existing deformable image registration (DIR) methods struggle with differing fields-of-view (FOV) across modalities.

Purpose of the Study:

• To develop a fast, patch-based method for aligning fields-of-view (FOV) in multi-modal 3D medical images.
• To improve deformable image registration (DIR) accuracy, especially when imaging modalities have significantly different FOVs.
• To assess the method's suitability for real-time, on-line interventional procedures.

Main Methods:

• A novel patch-based approach combined with a state-of-the-art multi-modal image similarity metric.
• Calculating patch shifts to determine the most frequent shift value for FOV adjustment.
• Evaluating performance on 30 pairs of pre-/per-operative irreversible electroporation (IRE) images (CT to CBCT, MRI to CBCT).

Main Results:

• The patch-based regional registration approach offers a balance between voxel-wise and global shift methods.
• The method demonstrated benefits for CT to CBCT and MRI to CBCT registration, particularly with disparate FOVs.
• Computational requirements are compatible with clinical on-line procedures using commodity hardware.

Conclusions:

• The proposed patch-based FOV alignment method effectively addresses challenges in multi-modal DIR with differing image FOVs.
• This technique is beneficial for improving image registration accuracy in interventional procedures like irreversible electroporation.
• The method's speed and efficiency make it a valuable tool for real-time image guidance in interventions.