Radiologists Should Read CT Colonography

This article defines the necessary skill set and knowledge base required for accurate computed tomography colonography (CTC) interpretation. The components of the interpretative process as well as the various strategies currently employed are discussed. The role of extracolonic evaluation as an integral part of this examination is also covered. Within this context, the question of whether a radiologist or gastroenterologist is better suited to interpret this examination is explored.

The debate regarding computed tomographic colonography (CTC) interpretation has elicited strong opinions and statements. Who is qualified to read CTC has become a central issue between radiologists and gastroenterologists in recent years, particularly with the maturation of CTC into an effective colorectal cancer screening modality. In general, radiologists have an inherent advantage from prior imaging training, particularly in computed tomography (CT), whereas gastroenterologists are better versed in clinical topics surrounding the issue of colorectal cancer screening and hold unique experiences from colonoscopy. Although there are several important attributes of a quality CTC reader, it is clearly apparent to experienced CTC readers that the core of interpretation rests on a skill set rooted in the principles of cross-sectional imaging. Without a firm grasp of cross-sectional skills, evaluation of a CTC examination cannot be done. Any attempt to do so is inherently flawed and will ultimately lead to poor quality. This view forms the foundation of the argument that “radiologists should read CTC” or perhaps more correctly stated, “are better qualified to interpret CTC.” In arguing this position, this article explores the components necessitating effective CTC interpretation, illustrating the need for cross-sectional imaging skills and exposing the common misconception regarding the role of the 3-dimensional (3D) endoluminal image. The issue of evaluation of extracolonic findings with CTC interpretation is also discussed. Once these concepts have been defined, the authors discuss whether radiologists or gastroenterologists would better be able to fulfill the requirements to provide quality interpretation.

Components of CTC interpretation

Despite the moniker of virtual colonoscopy, CTC is not truly virtual endoscopy. The postprocessing of CTC data can create images (ie, the 3D endoluminal view) that simulate colonoscopy ( Fig. 1 ); however, the interpretative skill set for CTC significantly differs from that of endoscopy. CTC is a low-dose, noncontrast CT examination. CTC has been optimized for luminal pathology where the bowel has been cleansed, any residual stool tagged, and the colon distended. The acquired thin section 2-dimensional (2D) CT images (typically 500 in each series; both supine and prone series are obtained) are then directly evaluated much in the same fashion as a standard CT examination as well as being postprocessed into various 3D series for additional evaluation.

There are 2 basic tasks in CTC interpretation. The first involves that of detection, where a list of potential polyps is generated. It is important to realize that many of these polyp candidates are not true soft tissue polyps but instead represent colonic folds or stool (tagged or untagged) ( Fig. 2 ). The second task involves evaluating these polyp candidates to winnow down the list to the few true soft tissue polyps. The specific steps to accomplish these 2 tasks are different, dependent on the particular interpretative approach employed (primary 2D or primary 3D). However, both strategies require a solid foundation in cross-sectional imaging. Although there are proponents for one approach over another, data from the ACRIN 6664 study suggest that the best CTC readers employ both strategies as needed for a given examination.

A primary 2D approach is so named for the use of the source 2D images during the polyp detection task. Image review is similar to the method of routine CT evaluation. Here, the 2D CT images (typically in the axial plane) are viewed in stack mode. By scrolling through the dataset, possible polyp candidates are identified as focal soft tissue protrusions into the colonic lumen that persist between the supine and prone series. The need for a strong cross-sectional skill set is obvious. It is important to realize that there are elements of characterization intertwined with detection with a primary 2D approach. Here, the assessment of the internal soft tissue make-up of the polyp candidate is undertaken during the detection phase. A suspected polyp is then confirmed at 3D where the 3D series is viewed to confirm morphology. In other words, the soft tissue structure represents a focal polypoid structure and does not elongate out into a colonic fold ( Fig. 3 ). Thus, detection is by the 2D series and confirmation (of morphology) is at 3D.

The alternative method of CTC interpretation is a primary 3D approach. The basis of this strategy involves the use of a postprocessed series, which converts the CT data into a 3D perspective. This strategy has only been possible in recent years because of the marked advances in underlying computer hardware and software. Likely in part to the similarity of appearance to the optical colonoscopic image, a common misperception has arisen wherein it is thought that CTC interpretation can be accomplished by viewing the “virtual endoscopic” images alone without any assessment of the 2D images. In other words, the 3D postprocessing of the CT source data can compensate for a lack of 2D interpretative skills and knowledge of underlying cross-sectional principles. As is discussed below, such is not the case. Cross-sectional skills are crucial for this approach.

In the primary 3D approach, the postprocessed 3D image series (such as the endoluminal fly through) is used to detect potential polyps. Thus, morphology is used as the sole marker for detection without input from other factors such as internal composition. This approach has been cited as one of the significant factors accounting for improved polyp detection rates. The rationale is that the postprocessing of the source data allows for easy recognition of focal polypoid structures from the colonic haustra. As opposed to a 2D approach for which the mental translation of a 3D structure from 2D data can be tedious or difficult, it is done here automatically by the computer, allowing for an easier detection pattern for the CTC reader. Perceptual errors where a polyp is mistaken for a fold are less likely at a primary 3D approach ( Fig. 4 ). Although this strategy is more likely to capture a true soft tissue polyp, it generates a much larger number of pseudopolyps. Because the internal lesion composition is not considered at this step, any structure with a polypoid morphology will be reconstructed to mimic a true polyp ( Fig. 5 ).

Thus, the second task of CTC interpretation (ie, polyp confirmation) holds even greater importance with a primary 3D interpretative schema and ultimately determines whether a quality interpretation results. Although 3D tools such as translucency rendering can suggest whether a polyp is of soft tissue density, only the 2D images with the use of cross-sectional skills can confirm the true polyps from the large list of polyp candidates generated by this approach. If a reader cannot perform this step competently, too many false positives will result and an unacceptably high number of CTC examinations will be erroneously referred to therapeutic colonoscopy.

Although a complete listing and description of the required cross-sectional skills for characterization is beyond the scope of this paper, it is helpful to cover the basic skills used in the determination of a classic polyp in order to gain a sense of the these requirements. Once detected, a potential polyp is confirmed at CTC by demonstrating 2 criteria on the 2D series. First, the candidate is fixed in location between the supine and prone series ( Fig. 6 ). A strong grasp of cross-sectional anatomy is required to make this determination. Often, the colon shifts in location between the supine and prone series, particularly those portions on an elongated mesentery. The reader must use the few internal colonic landmarks (eg, counting haustra from a definable point such as a diverticulum) and the relationship to the extracolonic structures to convince himself or herself that the polyp candidate is either fixed in location (ie, a polyp) or moves (ie, stool). Second, the internal make-up of the polyp candidate should be a homogeneous soft tissue density to distinguish from untagged or partially tagged stool (see Fig. 6 ). Again, a solid basis in cross-sectional principles is required to make this determination, as artifacts such as beam hardening or volume averaging must be taken into account for a given 2D appearance of the polyp candidate, otherwise leading to incorrect interpretations ( Fig. 7 ).