Supine and prone colon registration using quasi-conformal mapping

Area of Science:

• Medical Imaging
• Computational Geometry
• Radiology

Background:

• Virtual colonoscopy (VC) utilizes CT scans acquired in supine and prone positions.
• Accurate registration of supine and prone scans is crucial for clarifying findings and improving polyp detection rates in VC.
• Existing registration methods face challenges due to significant colon shape distortions caused by patient repositioning.

Purpose of the Study:

• To develop an efficient and robust algorithm for supine-prone registration in virtual colonoscopy.
• To ensure the registration results in a diffeomorphism, guaranteeing a one-to-one and onto mapping.
• To enhance the accuracy of polyp detection and reduce false positives in colon cancer screening.

Main Methods:

• Utilized conformal geometry and holomorphic differentials for colon surface flattening.
• Automatically extracted anatomical landmarks (taeniae coli, flexures) for segmentation and registration.
• Employed feature point detection (graph cut, morphological operations, PCA) and quasi-conformal adjustments for alignment.

Main Results:

• Developed an efficient algorithm for supine-prone registration of virtual colonoscopy data.
• Demonstrated successful alignment of corresponding features between supine and prone colon surfaces.
• Validated the method's efficacy through 2D flattened and 3D endoluminal visualizations and analytical measurements.

Conclusions:

• The proposed conformal geometry-based registration method effectively addresses challenges in supine-prone alignment for virtual colonoscopy.
• The technique ensures diffeomorphism and improves the reliability of polyp detection by enabling precise localization across different patient positions.
• This approach offers a significant advancement for improving the diagnostic accuracy and clinical utility of virtual colonoscopy.