Endoluminal surface registration for CT colonography using haustral fold matching

Abstract

Computed Tomographic (CT) colonography is a technique used for the detection of bowel cancer or potentially precancerous polyps. The procedure is performed routinely with the patient both prone and supine to differentiate fixed colonic pathology from mobile faecal residue. Matching corresponding locations is difficult and time consuming for radiologists due to colonic deformations that occur during patient repositioning. We propose a novel method to establish correspondence between the two acquisitions automatically. The problem is first simplified by detecting haustral folds using a graph cut method applied to a curvature-based metric applied to a surface mesh generated from segmentation of the colonic lumen. A virtual camera is used to create a set of images that provide a metric for matching pairs of folds between the prone and supine acquisitions. Image patches are generated at the fold positions using depth map renderings of the endoluminal surface and optimised by performing a virtual camera registration over a restricted set of degrees of freedom. The intensity difference between image pairs, along with additional neighbourhood information to enforce geometric constraints over a 2D parameterisation of the 3D space, are used as unary and pair-wise costs respectively, and included in a Markov Random Field (MRF) model to estimate the maximum a posteriori fold labelling assignment. The method achieved fold matching accuracy of 96.0% and 96.1% in patient cases with and without local colonic collapse. Moreover, it improved upon an existing surface-based registration algorithm by providing an initialisation. The set of landmark correspondences is used to non-rigidly transform a 2D source image derived from a conformal mapping process on the 3D endoluminal surface mesh. This achieves full surface correspondence between prone and supine views and can be further refined with an intensity based registration showing a statistically significant improvement (p<0.001), and decreasing mean error from 11.9 mm to 6.0 mm measured at 1743 reference points from 17 CTC datasets.

Keywords: CT colonography; Haustral fold; Landmark; Markov random field; Registration.

Graphical abstract

Fig. 1

Virtual colonoscopy (left), external (right) and internal (bottom) views of segmented haustral folds with marked centres. Red and blue sections represent fold and non-fold labelled vertices respectively. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 2

Internal views of rendered haustral fold in the prone (a) and supine (c) view, with their corresponding depth map images (b) and (d).

Fig. 3

Virtual camera with restricted set of degrees of freedom for optimisation. Image shows view along (left) and perpendicular to (right) the axis of the colon.

Fig. 4

Parameterisation based on Frenet-Serret (left) and Rotation Minimising (right) frames.

Fig. 5

Calculation of relative angle of rotation of haustral fold position with respect to an adapted frame.

Fig. 6

Distance between folds against variance in centreline (left) and rotational displacement (right) between corresponding fold pairs.

Fig. 7

Images of the endoluminal surface produced from the conformal mapping technique (case 11) showing the full registration with B-spline initialisation. The colour scheme shows the Shape Index (SI) and the vectors show the displacements generated from the landmark registration. Images show: (a) the source (prone) image; (b) the ambiguous vector direction on the source image; (c) the sorted displacements; (d) the source image vertically aligned to reduce displacements; (e) the source image with displacement vectors and regular grid; (f) the result of the landmark B-spline initialisation with transformed image and grid; (g) the refinement with the intensity based registration (with same grid); (h) the target (supine) image.

Fig. 8

Surface rendered examples of a subset of the cases used for validation. Top row shows prone view. Bottom row shows supine view. Cases shown from left to right are: 9, 14 and 16.

Fig. 9

Normalised distributions of labelled folds with respect to normalised centreline distance (caecum at 0, rectum at 1) in the reference standard (left) and the MAP labelling (right).

Fig. 10

External surface renderings of the transverse colon in the supine image of case 16. The set of reference standard points in the supine view (blue) and the corresponding points transformed from the prone view (green) and shown using the results from the LSI w/ IBS (left) and BSI w/ IBS (right) registration methods. The red lines show the Euclidean distance error. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)