Robust Mosaicing of Endomicroscopic Videos via Context-Weighted Correlation Ratio

Area of Science:

• Medical Imaging
• Computational Pathology
• Biomedical Engineering

Background:

• Probe-based confocal laser endomicroscopy (pCLE) offers real-time, in situ, and in vivo optical imaging for pathological assessment.
• The limited field of view in pCLE hinders a comprehensive understanding of scanned tissues.
• Accurate reconstruction of full tissue views from pCLE sequences is crucial for clinical diagnosis.

Purpose of the Study:

• To develop a novel mosaicing framework for assembling pCLE frame sequences into a complete tissue image.
• To enhance the clinical utility of pCLE by overcoming field-of-view limitations.
• To improve the accuracy and robustness of image registration for pCLE data.

Main Methods:

• A hybrid rigid registration combining feature and template matching for global frame alignment.
• Implementation of a parametric free-form deformation (FFD) model with multiresolution architecture for non-rigid distortions.
• Devised a context-weighted correlation ratio (CWCR) metric incorporating spatial and geometric contexts for accurate registration.

Main Results:

• The proposed mosaicing framework significantly outperforms state-of-the-art methods in handling intensity fluctuations, insufficient overlap, and tissue distortions.
• Experiments conducted on both robotic and manual setups validated the framework's effectiveness.
• The CWCR metric demonstrated superior performance in quantifying frame differences compared to other metrics, confirming the value of context-weighted strategies.

Conclusions:

• The developed mosaicing framework generates more rational and precise mosaics from pCLE data.
• This improved image reconstruction is highly beneficial for guiding diagnosis and treatment during optical biopsies.
• The study highlights the potential of advanced image processing techniques to enhance the diagnostic capabilities of pCLE.