Abstract
Colonography is an important screening tool for colorectal lesions. This paper presents a method for establishing spatial correspondence between prone and supine inner colon surfaces reconstructed from CT colonography. The method is able to account for the large deformations and torsions of the colon occurring through movement between the two positions. Therefore, we parameterised the two surfaces in order to provide a 2D indexing system over the full length of the colon using the Ricci flow method. This provides the input to a non-rigid B-spline registration in 2D space which establishes a correspondence for each surface point of the colon in prone and supine position. The method was validated on twelve clinical cases and demonstrated promising registration results over the majority of the colon surface.
Introduction
In colorectal cancer screening, Computed Tomography Colonography (CTC) is a vital non-invasive tool. To distinguish actual polyps from residual fecal matter or fluid, radiologists typically perform two scans: one with the patient in a prone (face down) position and one in a supine (face up) position. Gravity causes loose material to shift, while polyps remain fixed to the colonic wall.
However, the colon is a highly flexible organ. Changing the patient's position causes significant deformations, stretching, and torsion. Manually identifying the same anatomical location on both scans is mentally taxing and error-prone. Previous automatic registration attempts using 3D intensity-based methods or simple centerline alignment often fail because they cannot account for the complex twisting of the colonic surface.
Methodology
The proposed method transforms the complex 3D non-rigid registration problem into a manageable 2D image registration task by "unfolding" the colon surface.
- Surface Extraction and Topology: Triangulated meshes of the inner colon surface are extracted from CT segmentations. To facilitate mapping, the tubular colon is converted into a genus-one surface (a torus) by computationally joining the cecum and rectum ends.
- Discrete Ricci Flow: This core innovation deforms the surface's metric to become "flat" (zero Gaussian curvature) while preserving local geometric structure through conformal mapping. This allows the 3D tube to be unfolded into a 2D rectangular domain.
- Feature Mapping (Shape Index): Every point on the 2D plane is assigned a "Shape Index" (SI) value derived from 3D principal curvatures. This effectively turns local 3D geometry into a 2D "texture" map.
- 2D B-spline Registration: A multi-resolution non-rigid B-spline registration is performed on these SI images. A cylindrical constraint ensures continuity across the "seam" where the colon was unfolded. The resulting 2D transformation is then mapped back to 3D coordinates.
Results and Validation
The method was evaluated on 12 clinical cases. A radiologist with extensive experience manually identified approximately 10 pairs of reference points (polyps, diverticula, or unique folds) per patient to serve as ground truth. To account for human error, the radiologist repeated the landmarking after seven days, establishing a mean observer error of 3.8mm.
Key performance metrics include:
- Error Reduction: Without B-spline registration (basic alignment), the mean error was 23 ± 15 mm. With the proposed method, the error dropped significantly to 9.0 ± 11 mm.
- Polyp Accuracy: For cases where polyps were successfully registered, accuracy reached 4.7 ± 4.1 mm, which is critical as clinically relevant adenomas are often 5mm or larger.
- Reliability: 91% of registration points fell below a 20mm error threshold, providing a reliable "neighborhood" for radiologists to confirm matches.
Discussion and Limitations
By mapping the 3D surface to a 2D manifold, the approach bypasses the computational cost of 3D volumetric registration. Using the Shape Index as a "fingerprint" is effective because local fold shapes remain relatively consistent despite global deformations.
However, limitations remain. Registration performance was lower near the rectum and cecum due to artifacts from the topological "cuts" required for Ricci flow. Additionally, the method relies on high-quality segmentations; collapsed or poorly distended colon segments can introduce topological errors that disrupt the algorithm. Future work may involve using physical phantoms to establish a more absolute gold standard for validation.
Summary updated