A feasibility study on multimodal CT-MRI registration using segmentation aid and CoLlAGe feature extraction approach

Area of Science:

• Medical Imaging
• Image Registration
• Computational Anatomy

Background:

• Multimodal image registration is crucial for integrating information from different imaging modalities.
• Accurate registration of Magnetic Resonance Imaging (MRI) and Computed Tomography (CT) is essential for liver interventions.
• Existing methods may face challenges in achieving high precision for liver-specific applications.

Purpose of the Study:

• To propose a novel framework for multimodal MRI-to-CT image registration.
• To focus on liver-specific clinical applications, enhancing registration accuracy.
• To integrate feature-based registration with segmentation for improved performance.

Main Methods:

• Liver segmentation using pretrained nnU-Net models for both CT and MRI.
• Extraction of liver-specific CoLlAGe features within segmented regions.
• Image registration guided by extracted CoLlAGe features.

Main Results:

• Achieved high accuracy in MRI-to-CT registration across 24 pairs.
• Dice coefficients consistently above 0.8 (mean ± SD: 0.921 ± 0.038).
• Average Symmetric Surface Distance (ASSD) close to 0 mm (mean ± SD: 0.086 ± 0.213 mm).

Conclusions:

• The proposed framework shows significant potential for multimodal MR-to-CT registration.
• The achieved precision is suitable for liver surgical applications.
• This approach advances the field of medical image registration for clinical use.