2D-3D deformable image registration of histology slide and micro-CT with DISA-based initialization

Area of Science:

• Medical imaging
• Pathology
• Biomedical engineering

Background:

• Accurate registration of histology and micro-computed tomography (µCT) is crucial for virtual histology in pathology.
• Challenges include low soft tissue CT image quality and sample deformation during histology preparation, hindering structure correlation.
• Existing methods struggle with the complex deformations and multi-modal nature of these datasets.

Purpose of the Study:

• To develop and evaluate a novel 2D-3D multi-modal deformable image registration method.
• To address the limitations of current techniques in correlating histology slides with µCT data.
• To enhance the accuracy of virtual histology for pathological applications.

Main Methods:

• A novel 2D-3D multi-modal deformable image registration approach was developed.
• Initial global registration employed a machine learning-based differentiable similarity measure.
• Final registration refinement utilized an analytical out-of-plane deformation model.
• The method was tested on tonsil and tumor tissues using both phase-contrast and conventional absorption µCT.

Main Results:

• The proposed method demonstrated superior performance in 2D-3D deformable image registration compared to intensity- and keypoint-based methods.
• Visual and fiducial-based evaluations confirmed the enhanced accuracy of the new registration technique.
• Successful registration was achieved across different tissue types and µCT modalities.

Conclusions:

• The novel 2D-3D multi-modal deformable registration method offers a significant improvement for correlating histology and µCT data.
• This advancement facilitates more accurate virtual histology and pathological analysis.
• The method shows promise for overcoming key challenges in soft tissue imaging registration.