CT colonography: size reduction of submerged colorectal polyps due to electronic cleansing and CT-window settings

Computed tomography colonography (CTC) is the radiological examination of choice for the diagnosis of colorectal neoplasia [1].

The evaluation of CTC examinations may be limited by residual stool and fluid, and thus either simulating or obscuring polypoid colonic lesions. By oral administration of positive contrast media, residual faecal material is “tagged”, i.e. rendered hyperdense. Soft-tissue lesions submerged within hyperdense tagged faecal residue can be depicted on 2D images. Faecal tagging has been shown to increase the sensitivity as well as the specificity of CTC and is now considered a mandatory part of bowel preparation [2, 3]. However, colonic mucosa that is obscured by tagged faecal residue is not accessible to endoluminal three-dimensional (3D) views and requires additional time-consuming two-dimensional (2D) evaluation. An approach to overcome this limitation includes “electronic cleansing” (EC) software algorithms designed to digitally subtract tagged faecal residue from a CTC dataset. In fact, EC has been shown to improve colonic evaluation and polyp detection [4‐6]. However, it has not been determined in vivo, as yet, whether EC affects the size of polyps that are submerged under tagged faecal residue. Related size differences may influence clinical reporting of polyps and, thus, patient treatment.

The purpose of this retrospective study was to assess whether electronic cleansing of tagged residue and different CT-window settings influence the size of colorectal polyps in CTC studies and whether related size differences lead to different grading into clinically applied size categorisations.

Little is known about the effect of EC on the morphological features of lesions submerged under faecal residue. From a clinical point of view, the most important morphological factor is polyp size. The 10-year risk for colorectal cancer increases with polyp size [13]. Due to lack of a pathohistological assessment, the polyp size represents the key criterion by which to estimate clinical relevance using CTC criteria [3]. Based on the recent European guidelines, there is general consensus that all polyps ≥6 mm need to be reported and endoscopic polypectomy should be recommended, while polyps <6 mm may be ignored [1, 3]. According to the CT Colonography Reporting and Data System (C-RADS), a categorisation scheme for colonic and extracolonic findings and for follow-up recommendations that may be applied for screening, patients with polyps 10 mm or larger require colonoscopy with resection (C3 category). However, for small lesions that measure between 6 and 9 mm, and if there are less than three in number (C2 category), CTC surveillance may be offered to the patient [14]. Therefore, the size of a polyp as measured by the radiologist has a direct influence on the selection of the appropriate therapeutic procedures [15].

We found that both EC and narrow window-level settings reduce significantly the size measurements of polyps. The size reduction was largest when EC was performed in soft-tissue windows and it was smallest using a colon window.

In addition, differences in window-level settings showed substantial influence on the size of submerged polyps, an observation that has already been previously described by Slater et al. [16]. The polyp size was largest when measured in wide window settings, such as colon or bone windows, and decreased when measured in soft-tissue windows, both before and after EC. Size differences of polyps in different window settings are believed to be related to the partial volume effect, which leads to a density gradient at the contrast media–polyp interface, with the window level and width defining how much of this gradient is included and, consequently, how much will be visible [17]. Size reductions after electronic cleansing most likely result from a slight transformation of the shape of the line attenuation profile of the polyps during the digital subtraction process.

Not surprisingly, the combination of both EC and the use of narrow window settings led to the largest decrease of polyp size. Compared to measurements taken in an unsubtracted colon window, which is generally agreed to display the largest diameter of the polyp most accurately [3], EC, performed in a soft-tissue window, led to a size reduction of 27.6% for polyps ≥6 mm and of 13.3% for polyps ≥10 mm.

Size reductions in polyps can be the source of misclassification of lesions into the wrong size category and, therefore, of underestimation of lesions of potential clinical relevance. Within each of the three window settings, EC led to a significant decrease in the number of polyps measuring ≥6 mm. However, when compared with the reference size measurements, taken in the unsubtracted colon window, the number of polyps shifting to smaller size categories increased further due to the additional size reduction related to CT-window changes, with 20.8% and 52.1% of submerged polyps ≥6 mm in bone and soft-tissue windows, respectively.

If being applied in the C-RADS system, EC would have led to a shift of 5, 7 and 17 patients from the C2 to the C1 category within colon, bone and soft-tissue windows, respectively, and 5 patients from the C3 to the C2 category in a soft-tissue window, when compared to the categorisation in an unsubtracted colon window.

While several studies have focused on the impact of EC on feasibility, sensitivity or interpretation time reductions [4, 5, 18‐21], only one study has addressed the topic of a potential change in the size of submerged lesions [22]. Zalis et al. [22] evaluated in vitro the effect of EC software on size measurements of polyps in a colon phantom, repeatedly scanned without intraluminal contrast and with various bowel contrast material concentrations, a study design that cannot be applied in vivo. The mean measurement error of 0.6 mm assessed by two readers in 11 submerged polyps ≥8 mm was not significant at 200 and 500 HU, but increased to a significant level of 1.1 mm at 560 HU and 2.2 mm at 840 HU. Smaller lesions were not considered in this study.

Although not specifically investigated, other authors have also reported morphological effects on submerged lesions that were associated with EC. Serlie et al. [21] reported recognisable changes in the intraluminal appearance of polyps that were related to EC. Furthermore, the degeneration of bowel wall and colonic folds, as well as the distortion or degeneration of submerged polyps, were also found separately in different previous studies [4, 20]. Cai et al. [19] investigated an EC algorithm that potentially reduced EC-related artefacts due to incomplete subtraction at the air/contrast media interface. Strategies to further reduce EC-related artefacts and pitfalls may include the use of EC schemes that are based on the material decomposition capability of dual-energy CT, as shown in a recent phantom study [23, 24].

The results of this study show that measuring of submerged polyps after EC may account for shifting polyps to smaller size categories, a finding that is enhanced by using an inappropriate window setting. In the most unfavourable combination, when performing EC in a soft-tissue window setting, more than 50% of submerged lesions, measured as ≥6 mm, and 30% of submerged lesions ≥10 mm can be underestimated as <6 mm and as <10 mm, and therefore are either completely ignored or not resected.

This study has limitations. Some examiners may favour 3D over 2D measurements [25, 26]. It is, however, technically not possible to assess the influence of EC on 3D measurements because submerged polyps are surrounded by tagged residue, impairing the 3D reconstructions, which, however, are based on 2D image data. There has been expert agreement that the maximal diameter of lesions should be primarily estimated using 2D views, which were considered to be most reliable [3, 27]. Polyp size may vary within different 3D perspectives or appear larger due to contrast coating [18, 28].

A generally accepted “gold standard” for true polyp size would be desirable, since this would enable not only relative but also absolute size comparisons. However, endoscopic, histopathological or CT estimates are not wholly accurate and are affected by the way the measurement is obtained [3, 17, 29]. CT measurements, are, however, considered to provide a solid and reproducible estimate that is believed to range between endoscopic and pathohistological measurements [17, 30]. Finally, the decision for clinical treatment is solely based on CT measurements [17].

The presented results are limited to the applied algorithm for EC. It is possible that results may be different when using other algorithms for EC. However, since the basic principle of EC might also be applied in other approaches, it cannot be excluded that changes in lesion characteristics will also appear with other algorithms, as supported by observations in previous studies [4, 17, 19‐21].

In conclusion, EC as well as narrow CT windows significantly reduce the size of polyps submerged in tagged residue. To avoid underestimation of polyp size, measurements of submerged polyps should be performed on unsubtracted image data in a colon-window setting.