Imaging Invaluable for the Correct Diagnosis

Ayso H. de Vries, Rogier E. van Gelder, and Jaap StokeR


9.1 Introduction 117

9.2 Three-dimensional Reconstruction 117

9.3 Three-dimensional Display Methods 118

9.3.1 Conventional 3D Display 118

9.3.2 Alternative 3D Display Methods 119

9.4 2D and 3D Reading are Complementary 120

9.5 Fecal Tagging and Electronic Cleansing 123

9.6 Primary 2D and Primary 3D: Difference in Accuracy? 123

9.6.1 High Prevalence Population 123 Two-dimensional Methods 123

9.6.2 Low Prevalence Population 124 3D Methods 124 2D Methods 124 2D vs 3D Methods 124 Discussion 125

9.7 Review Time 125

9.8 Conclusion and Future Development 126 References 127


In evaluation of computed tomography (CT) colo-nography (virtual colonoscopy) examinations there are basically two principles of reviewing: it can be done two-dimensionally (2D) or three-dimension-ally (3D) (Fig. 9.1).

The simplest 2D approach is to view the axial helical CT images without any additional processing, although in general this will be performed with multi planar reformatting (MPR), or a combination of both MPR and 3D display. The method is named primary 2D if 3D is used for problem solving.

A. H. de Vries MD; R. E. van Gelder, MD; J. Stoker, MD, Professor of Radiology

Department of Radiology, Academic Medical Center, University of Amsterdam, Meibergdreef 9, 1105 AZ G1-226, Amsterdam, The Netherlands

Alternatively, evaluation of CT-colonography examinations can be done in a primary 3D approach, in which an (endo)luminal view of the colon is combined with a requisite 2D method.

In this chapter the pros and cons of 3D reviewing of colonography examinations are discussed. This discussion will be based on data of CT colonography only since no comparative studies have been published on MR-colonography yet.

Three-dimensional Reconstruction

Currently there are two rendering techniques used to produce virtual colonoscopy; surface rendering and volume rendering.

Surface rendering uses a specific attenuation coefficient to define the air-lumen interface. This is done by attaching an opacity value of 0 (complete transparency) to voxels that have an attenuation coefficient under a certain threshold and an opacity value of 1 (no transparency at all) for voxels above this threshold. Consequently, surface rendering assigns all structures either into luminal air or colonic wall, depending on the attenuation coefficient selected. By raising the threshold, more voxels of the colonic mucosa will be assigned to the lumen since the attenuation coefficient in transition zone (mucosa) typically lies between that of air and soft tissue.

Unfortunately, the attenuation coefficient varies within patients as well as with pathologic findings, which decreases the detail seen on the image. A second drawback is the method's sensitivity to noise and artifacts (Hopper et al. 2000).

The second method is volume rendering. This method categorizes voxels into multiple groups on the basis of their attenuation coefficient. Unlike surface rendering, volume rendering allows the transition voxels (colonic mucosa) between air and bowel wall to be reconstructed as a separate, specific structure. Although this method requires more computer

Fig. 9.1. Image of a 14 mm polyp in the cecum displayed at axial 2D (left), MPR (middle) and 3D (right)

power, it provides smoother rendering and provides the possibility to peer into the colon wall (Hopper et al. 2000). Most of the commercially available software is based on this method.

Until recently most implementations of both rendering methods did not allow for interactive visualization but now interactive visualization has been introduced. In contrast to real time rendering, image sequences are generated off-line for later diagnostic examination by the physician. Viewing positions are generated at regular intervals along the central path which generates the illusion of flying through the lumen of the colon (virtual colonoscopy). This central path is generated in a (semi) automatic way through the lumen of the colon. Several different methods for the calculation of the centerline are used, though comparative studies are lacking. If the colon is discontinuous because a segment is collapsed or entirely filled with fecal material or fluid, the centerline may need to be manually adjusted.

Three-dimensional Display Methods 9.3.1

Conventional 3D Display

The first 3D visualization method for CT colonogra-phy was adopted from conventional endoscopy, and has a similar disadvantage: areas behind haustral folds are not easily visualized. As a consequence, in conventional 3D methods the colon needs to be evaluated in both antegrade and retrograde directions in order to visualize sufficient colonic surface. With a fly-through in one direction substantial parts of the colonic wall and therefore polyps will be obscured by haustral folds, as happens in optical colonoscopy (Fig. 9.2). Two-directional fly-through evaluation reduces these unseen colonic areas considerably (Paik et al. 2000; Vos et al. 2003). However, even with a two-directional fly-through, substantial parts of the colonic wall that potentially harbor polyps remain non-visualized. Interactive evaluation can overcome this problem at the expense of an additional, and often substantial, increase in reading time.

An other solution to this problem is to use an algorithm that identifies areas that are not visualized during bidirectional evaluation. These areas, indicated for example by color, are presented to the observer after completing the bidirectional evaluation (Fig. 9.3). For practical purposes, the areas can be presented in descending order of size. The observer can then decide whether to skip small-sizes areas as these will not hide clinical relevant polyps.

An alternative approach to reduce unseen areas is to increase the viewing angle of the virtual camera. Consequently more colonic surface is displayed. A major drawback is the resulting distortion, especially at the edges, that prevents the use of these large viewing angles (Fig. 9.4).

Fig. 9.2. Schematic shows areas in black that are missed in conventional 3D view. Reprinted with permission of Vos et al.
Fig. 9.3. Endoluminal view with "Missed Region Tool." This feature allows the reader to investigate areas (indicated in pink) that were not previously viewed during conventional fly through. (Figure courtesy of Viatronix, Stony Brook, NY)

Alternative 3D Display Methods

The ideal 3D display mode shows the complete colonic surface (so in theory no polyps can be missed) in a time efficient way and without image distortion (so polyps can be recognized as such).

Several groups have studied alternative 3D methods, which all have in common a less 'colonoscopy'-like representation of the colonic surface than the conventional fly-through.

A so-called "flattening method" can be used to straighten and flatten the colon mathematically. Beaulieu et al. (1999) evaluated a method called "Panoramic endoscopy" (Fig. 9.5). With this method the inner colonic surface is depicted as a flattened structure. The camera is rotated around the path in

60° increments, which generated six image panels at an interval of 3 mm along the central path. When these image panels are displayed side by side, they depict a panoramic view of the colonic wall.

In a comparative study of display methods for CT colonography with the use of simulated polyps of different size and morphology, panoramic endoscopy had a significantly higher sensitivity (90%) than bidirectional virtual endoscopy (68%) for polyps 7 mm or larger. Part of the difference could be explained by the invisibility of some lesions during bidirectional virtual endoscopy since this method does not visualize the complete colonic surface.

"Virtual colon dissection" is an improvement of the "Panoramic endoscopy" method evaluated by Hoppe et al. (2004). In this method the virtual camera captures one quarter of the circumference of the complete colon length at each viewing position with a 90° camera field of view that can be rotated in 45° increments. Eight contiguous panels displaying the colonic circumference are generated by rotating the virtual camera in 45° increments around the path.

In a study of 20 patients (31 colonic lesions of which 9 were 10 mm or larger) reviewed by 2 radiologists, Hoppe et al. reported a sensitivity of 67% and 89% for virtual colon dissection and 89% and 100% for axial 2D interpretation. The authors concluded that although virtual colon dissection may facilitate detection of colonic polyps in isolated cased, its detection rate is not superior to axial 2D interpretation. This can mainly be attributed to failed rendering of a high number of insufficiently distended colonic segments or regions with residual feces in this series.

Virtual dissection, as depicted in Fig. 9.6, is an image display method that is another improvement of the "flattening method". Virtual dissection is a method that has recently become available and has, according to our knowledge, not been evaluated yet. It opens and straightens the entire lumen of the colon along the longitudinal axis with a 45° overlap

Fig. 9.4. Distortion illustrated of a polyp at four different endoluminal views of 160°, 120°, 90° and 60°
Fig. 9.5. "Panoramic endoscopy" display. Reprinted with permission of Beaulieu et al.

on each side. The potential advantages of this "flattening method" include the easy overview over a substantial part of the colonic surface and complete visualization of the entire mucosal surface. This may lead to a reduction in interpretation time.

A drawback of the method is that straightening of a curved structure like the colon results in distortion of the colonic surface. Moreover it does not display the forward and backward viewing directions. Consequently, the frontal site and back site of structures are not visible. This problem can be overcome by combining the method with a conventional 3D method.

The "unfolded cube" method (Fig. 9.7) is a solution that tackled the problem of colon coverage in a different way. It was published by SErLiE et al. (2001). In this method the colonic surface is projected on a cube. On the cube faces, 90° views are projected. By folding out the six images onto a single plane (unfolded cube display) the complete field of view is rendered.

In a comparative series 99.5% of the colonic surface was displayed with this technique (Vos et al. 2003) as compared to 93.8% with antegrade and retrograde reviewing with a conventional 3D (120°) endoluminal view. Bidirectional reviewing is not mandatory with this method and therefore the additional evaluation time compared to 2D reviewing is less compared to the additional evaluation time of conventional 3D reviewing (vAn Gelder et al. 2004c).

2D and 3D Reading are Complementary

Two- and three-dimensional display methods must be considered as complementary instruments to evaluate CT colonography.

If a suspicious area is detected when using a primary 2D review method, a 3D snapshot can be used to obtain more information about the nature of the abnormality (e.g. folds and ileocecal valves) (Figs. 9.8 and 9.9).

On the other hand, a 3D method needs to be complemented by a method that can assess the heterogeneity and level of the attenuation values within an area of interest. Information about the attenuation values of a suspected lesion is mandatory for an accurate differentiation between a polyp and fecal material (Fig. 9.10). In practice this means that the method is combined with an axial or MPR 2D method (Fig. 9.11) or a method that demonstrates attenuation values in a 3D view (Pickhardt 2004) (Fig. 9.12).

Obviously 2D and 3D methods are complementary, and a combination of both methods is essential for review of CT colonography. The question that remains, however, is which method is preferable for the primary review?

Fig. 9.8. Complex fold that resembles a polyp in the sigmoid colon in axial 2D (left), but is not in a conventional 3D view (right)
Fig. 9.9. Protruding ileo-cecal valve seen on axial 2D (left) and well recognizable on 3D (right)
Fig. 9.10. Polyp in the transverse colon, seen on 3D (left) proves to be tagged fecal material in a 2D axial view (right)
Fig. 9.11. Screen panel that combines a conventional endoluminal view (bottom left), an unfolded cube display (center), an axial 2D display (top left), a 2D MPR display (top center) and an overview of the colon (top right)
Fig. 9.12. Translucency rendering applied to a 3D image (right). The polyp shows a red interior which is indicative of soft tissue attenuation
Managing Diverticular Disease

Managing Diverticular Disease

Stop The Pain. Manage Your Diverticular Disease And Live A Pain Free Life. No Pain, No Fear, Full Control Normal Life Again. Diverticular Disease can stop you from doing all the things you love. Seeing friends, playing with the kids... even trying to watch your favorite television shows.

Get My Free Ebook

Post a comment