The British Journal of Radiology, 82 (2009), 475–481

Quantitative assessment of colonic movement between prone and supine patient positions during CT colonography 1

S PUNWANI, PhD, MRCP, FRCR, 1S HALLIGAN, 2 MRCP, FRCR and D HAWKES, PhD, FInstP, FREng

MD, FRCP, FRCR,

3

D TOLAN,

MRCP, FRCR,

1

S A TAYLOR,

MD,

1

Department of Specialist Radiology and, 2Centre for Medical Image Computing, University College London, London and Department of Clinical Radiology, Leeds General Infirmary, Leeds, UK

3

ABSTRACT. This paper aims to quantify changes in colonic length and positional change between supine and prone CT colonography (CTC) studies in order to aid development of image registration techniques. CTC studies in 20 patients (10 men and 10 women) with technically adequate distension were analysed using an image analysis workstation. Spatial co-ordinates of colonic landmarks were determined in both prone and supine orientations using a three-dimensional colon model view and centreline positions. Change in the co-ordinate position of colonic segments between supine and prone scans was calculated using the superior mesenteric artery as a fixed point of reference. There was no significant difference in total colonic length for subjects between prone and supine positions, nor any significant difference overall when men were compared with women. However, significant differences between sexes for individual segments were found; the ascending colon, descending colon and rectum were significantly longer in men and the sigmoid colon was longer in women. The transverse colon was the most mobile segment during positional change, with an average displacement between supine and prone scans of 4.6 cm (standard deviation, 0.48 cm) for men and 4.1 cm (standard deviation, 0.4 cm) for women. Consistent patterns of colonic positional change between supine and prone orientations were present and were thought to be most likely the result of abdominal compression. We concluded that there is minimal variation in colonic length between prone and supine orientations. Consistent patterns of colonic displacement with patient position suggest that predictable forces act upon the colon. Understanding these forces will facilitate image registration for CT colonography.

Computed tomography colonography (CTC) is an emerging technology for the diagnosis of colorectal cancer and polyps in both screening and symptomatic populations. Average sensitivity and specificity using metaanalysis for the detection of polyps > 6 mm in size have both been reported at 86% [1]. Computer-aided detection (CAD) algorithms have recently become commercially available and can increase the sensitivity of polyp detection, but possibly at the expense of specificity [2–6]. During CTC, data are acquired with the patient in both prone and supine positions in order to redistribute gas, residual fluid and faeces, and combined analysis of both data sets has been shown to improve the sensitivity of polyp detection by human observers [7]. However, current CAD algorithms do not use supine–prone matching to improve detection characteristics because colonic movement between orientations is assumed to hinder the ability to register the two data sets accurately. Some authors have attempted to register the prone and supine acquisitions by using the colonic centreline, but match errors of up to 5.9 cm have been observed [8]. More sophisticated registration employing non-rigid registration techniques [9] may reduce such Address correspondence to: Professor Steve Halligan, Professor of Gastrointestinal Radiology, Department of Specialist Radiology, Podium Level 2, University College Hospital, 235 Euston Road, London NW1 2BU, UK. E-mail: [email protected]

The British Journal of Radiology, June 2009

Received 26 February 2008 Revised 8 May 2008 Accepted 20 May 2008 DOI: 10.1259/bjr/91937173 ’ 2009 The British Institute of Radiology

errors and improve accuracy but, at the time of writing, little is known about the degree to which colonic position changes between prone and supine acquisitions. Colonic mobility in vivo has been determined previously using barium enema by authors attempting to predict difficulty at subsequent colonoscopy [10]. However, patient positioning differs from CTC and, most notably, the colon contains a significant volume of highdensity barium suspension when enema is performed, which may influence mobility. Quantitative assessment of colonic positional change between supine and prone orientations during CTC is therefore needed to provide data for development of non-rigid registration techniques. The aim of this study was to quantify the degree and direction of colonic displacement between supine and prone scans in men and women undergoing CTC.

Methods and materials Permission was obtained from the local ethics committee for the use of anonymous patient data. A single observer searched a local CTC database retrospectively and selected clinical studies that demonstrated subjectively good colonic distension on both prone and supine data sets. Individual patient data sets were displayed using a CTC workstation (Viatronix 3D, 475

S Punwani, S Halligan, D Tolan et al

Stony Brook, NY) and colonic distension assessed subjectively using a three-dimensional (3-D) rendered overview of the colon. Studies with . 3 cm of continuous undistended lumen, or with multiple small areas of collapsed colon on either the supine or prone data sets, were excluded (representing approximately 40% of the data set). In total, 20 patient studies were selected: 10 men (mean age 54.6 years, SD 5.9) and 10 women (mean age 59.3 years, SD 9.1). CTC examinations were performed using a standard technique. In brief, all patients underwent full bowel preparation before the procedure and were examined in the prone and supine positions following intravenous administration of hyoscine-N-butylbromide (Buscopan, Boehringer Ingelheim, Germany) to abolish bowel peristalsis. Colonic insufflation was achieved using carbon dioxide and an automated insufflator (ProtoCO2L, E-Z-EM, Bicester, UK). Examinations were performed using a 64-detector-row machine (Siemens Somatom 64, Siemens Medical Solutions, Berkshire, UK) and the following scan parameters: all 64 detector rows used 0.6-mm collimation, 50 mA reference tube current, 100 kV, a 0.5 s rotation time, a 26.9 mm rotation, a 1.4 pitch factor, a 0.7 mm increment, a B20f kernel and a 50 cm field of view and with dose modulation active. Data were reconstructed into 1 mm slices using a smooth reconstruction algorithm. Each individual study was loaded into the colonic visualization software (V3D-Colon; Viatronix, Stong Brook, NY) and six pre-defined colonic landmarks identified by the observer on a combination of the 3-D segmented colon view and the prone and supine studies (window: width, 1500 HU; level –200 HU) (Table 1, Figure 1), thus dividing the colon into five segments. Centreline x, y and z co-ordinates were then determined at each colonic landmark in both prone and supine orientations, using the ‘‘World co-ordinates’’ setting of the image analysis workstation. For each colonic segment, three further centreline co-ordinate points were defined by trisection of the centreline between successive landmarks (Figure 2). The superior mesenteric artery (SMA) origin coordinates were used as a fixed point of reference between prone and supine studies because this structure is a fixed retroperitoneal structure and easily identifiable on CT images. The distance between the individual centreline points x, y and z co-ordinates and the SMA origin was calculated for supine and prone studies for each individual patient. The change in position of the centreline relative to the SMA origin between supine and prone scans was then determined, and any change in location

of the centreline between supine and prone studies was attributed to colonic positional change.

Data analysis Total colonic length was defined as the centreline length measurement between the landmarks identifying the origin of the appendix and the anorectal margin. The length of each colonic segment was defined as the centreline length between successive colonic landmarks as defined in Table 1. Positional changes in total colonic length and segmental colonic length were then determined by calculating the difference between corresponding prone and supine measurements. Colonic displacement was determined by calculating the magnitude of displacement between supine and prone scans using triangulation and Pythagoras’s theorem. For each colonic segment, the mean change for all segmental centreline points for all subjects was calculated for the x, y and z directions. To determine whether a preferential direction of displacement existed, the mean value in each direction was assessed to determine whether it was significantly different from zero using a single-sample two-tailed Student’s t-test. We also plotted the distance from the SMA origin as a function of each centreline co-ordinate (i.e. the x-axis representing the distance along the centreline from the anus and the yaxis representing the distance from the SMA origin), and estimated the degree of displacement of each individual point that occurred between prone and supine studies.

Results Mean total colonic length was 184.2 cm (SD 26.9 cm) in the prone position and 183.9 cm (SD 22.6) in the supine position for men, and 175.7 cm (SD 33.3) in the prone position and 173.6 cm (SD 32.7) in the supine position for women. There was no significant difference in total colonic length between the prone and supine position for men (p 5 0.98) or women (p 5 0.87), nor was there any significant difference when mean lengths between men and women were compared (p 5 0.54 prone, p 5 0.42 supine). However, significant differences were found between men and women when the lengths of individual segments were considered. For example, in both orientations, the ascending and descending colon and rectum were significantly longer in men (Table 2). The sigmoid colon was longer in women, but this finding attained significance only in the supine position (Table 2). When

Table 1. Anatomical colonic landmarks Colonic landmark

Criteria

Anorectal margin Rectosigmoid junction

Centreline position at the anorectal junction Centreline position at the distal most maximum point of inflexion (located approximately at the S3 vertebral body level) Centreline position at which the distal descending colon angles forward (located approximately at the pelvic brim) Centreline position of the leftmost, most cranial inflexion point Centreline position of rightmost, most cranial inflexion point Centreline position at caecal pole origin of appendix

Sigmoid–descending junction Transverse–descending junction Ascending–transverse junction Appendix origin

476

The British Journal of Radiology, June 2009

Colonic movement between prone and supine CTC studies

Figure 1. A three-dimensional model view of a colon illustrating fixed colonic landmark selection: (a) rectosigmoid junction, (b) sigmoid–descending junction, (c) splenic flexure, (d) hepatic flexure, (e) appendiceal origin identified on two-dimensional views and (f) shaded individual colonic segments.

individual segmental length was compared between prone and supine orientations, the descending colon was significantly shorter for women in the supine position (27.7 cm, SD 8.9) than in the prone position (30.8 cm, SD 9.2; p , 0.01), and the ascending colon was significantly shorter for men in the prone position (30.7 cm, SD 6.7) than in the supine position (34.0 cm, SD 9.1; p 5 0.04). All other comparisons yielded nonsignificant changes. Concerning colonic displacement, Table 3 shows the mean displacement of the aggregated centreline points that occurred when shifting between the prone and supine positions for both men and women. There was no significant difference between mean displacement of any individual segment for male and female groups but, when the two groups were compared, the rectum was significantly more mobile in men than in women (Table 3). The transverse colon was the most mobile structure during positional change, with an average displacement between supine and prone scans of 4.6 cm (SD 0.48) and 4.1 cm (SD 0.4) for men and women, respectively. The preferential x, y and z direction of displacement for individual colonic segments for both men and women combined is shown in Figure 3. There was a tendency for the rectum and sigmoid to move to the right and anteriorly upon positional change from supine to The British Journal of Radiology, June 2009

prone. The ascending, descending and transverse colonic segments demonstrated a preference for posterior displacement on prone positioning. Furthermore, a caudal displacement of sigmoid, transverse and ascending colonic segments was found with prone positioning. There was no preferential transverse (x) direction of displacement for descending, transverse and ascending colonic segments, and no preferential anteroposterior (y) direction of displacement for the descending colon. The mean displacement between prone and supine positions for each individual data point on the centreline is shown in Figure 4 for men and women separately. It was apparent that displacement was greatest in the transverse colon and least in the hepatic and splenic flexures.

Discussion Colonic length has been investigated previously using barium enema studies. Sadahiro et al [11] estimated colonic length at 125.87 ¡ 15.95 cm for Japanese men and 132.83 ¡ 15.69 cm for Japanese women. Saunders and colleagues [10] also used barium enema to obtain estimates of up to 157 cm (upper 95% CI 162; lower 95% CI 152.3) in Caucasians. Intraoperative measurements of colonic length have found a mean of 113.5 cm for men and 114.9 cm for women [12], although cadaveric studies 477

S Punwani, S Halligan, D Tolan et al

Transversedescending Ascendingtransverse D3 T1 A1

D2

T3

D1

A2 A3

S2

T2

Sigmoiddescending

S3

S1 Appendix origin

Rectosigmoid

R3 R2 R1

Anorectum

Figure 2. A three-dimensional model view of a colon illustrating centreline points.

have produced corresponding estimates of 180 cm for men and 157 cm for women [13]. We could identify only a single study from the CTC literature the purpose of which was to estimate colonic length [14]: Hanson and colleagues compared total colonic length from patients in whom colonoscopy was incomplete with those in whom it had been successful and found that colons were longer in the incomplete group (210.8 cm vs 167 cm, respectively). The authors did not assess movement between prone and supine acquisitions. The colonic length measurements in the present study are similar to those found in the CTC study performed by Hanson et al [14], although they are longer than those recorded by others using barium enema [10]. This discrepancy is likely to be a consequence of the radiographic foreshortening that occurs during barium enema filming because some portions of the colon are perpendicular to the radiographic detector. In contrast, CTC

centreline determinations of colonic length is a calibrated 3-D intraluminal measure and likely to be more accurate. Furthermore, it is possible that insufflation of the colon during CTC stretches the colon, thereby increasing its length when compared with intraoperative measurements despite traction applied to the colon during surgery. Additionally, it has been proposed that the cadaveric colons can stretch more because muscle tone is absent [12]. We hypothesise that a CTC-derived estimate of colonic length from a well-distended colon is likely to reflect true colonic length more accurately than previously described methods. We found no change in total colonic length for men or women when supine and prone positions were compared. We found preferential directions of displacement for individual bowel segments, implying that predictable forces act upon the distended colon and operate in predictable directions. Colonic disposition may be

Table 2. Comparison of mean segmental colonic lengths for men and women in both prone and supine orientations Orientation

Segment

Length (cm) in men ¡ SD (n 5 10)

Length (cm) in women ¡ SD (n 5 10) Probabilitya

Prone

Rectum Sigmoid Descending Transverse Ascending Rectum Sigmoid Descending Transverse Ascending

14.30 46.28 42.93 49.99 30.72 14.09 45.43 42.34 48.08 34.01

10.83 ¡ 2.55 58.18 ¡ 11.92 30.8 ¡ 9.21 52.86 ¡ 14.23 23.07 ¡ 5.43 10.73 ¡ 3.21 57.08 ¡ 9.49 27.68 ¡ 8.94 54.44 ¡ 12.76 23.68 ¡ 9.65

Supine

¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡

3.10 16.08 13.31 5.32 6.67 2.95 13.22 13.07 4.51 9.08

0.01 0.08 0.03 0.56 0.01 0.03 0.04 0.01 0.15 0.03

SD, standard deviation. Student’s t-test.

a

478

The British Journal of Radiology, June 2009

Colonic movement between prone and supine CTC studies Table 3. Mean displacement of individual colonic segments occurring between prone and supine orientations for both men and women Centreline segment

Distance (cm) in men (SD)

Distance (cm) in women (SD)

Probabilitya

Rectum Sigmoid Descending Transverse Ascending

3.6 3.1 3.5 4.6 3.7

2.9 3.2 3.1 4.1 3.7

0.02 0.68 0.51 0.21 0.70

(0.72) (0.25) (1.11) (0.48) (0.65)

(0.43) (0.74) (0.83) (0.40) (0.49)

SD, standard deviation. a Student’s t-test.

affected by many factors, including gravity, abdominal constraint, elasticity of the abdominal wall and diaphragmatic musculature, pelvic tilt, intraluminal contents and the degree of colonic distension. We found that when moving from a supine to a prone position, the rectum had a tendency to move right, anteriorly and cranially. Because the rectum is a fixed retroperitoneal structure, this observation seems likely to be a result of changes in the pelvic position relative to the SMA when turning. The sigmoid colon demonstrated a similar tendency to move right and anteriorly when moving to a prone position, but travelled caudally rather than cranially. Descending, transverse and ascending colon segments moved posteriorly when changing from the supine to prone position. We had expected to observe the opposite,

owing to the effects of gravity, but believe our findings are a consequence of anterior abdominal compression when lying prone, which displaces intra-abdominal contents posteriorly. Like the sigmoid, the transverse and ascending colon also move caudally when prone. It is possible that this is also a predictable effect of compression of the anterior abdominal wall. Therefore, our findings suggest that, when distended, the position of the majority of the colon is dictated primarily by changes in compressive effects, rather than by the effect of gravity. Rectal and sigmoid colon positional changes seem most dependent on pelvic tilt. Lateral colonic movement for transverse, ascending and descending colonic segments is less predictable and may be related to the length of mesocolon.

25

57

20

15

10

43

41

33

5

39

28

37

y

33

0 Ascend

Trans

Descend

Sigmoid

Rectum

–5

31

x

z

37 41

49

–10

40

39 50

–15

–20

Figure 3. Mean direction of displacement for individual colonic segments in each of the x-, y- and z-axes for men and women (n 5 20). Positive x, y and z displacements signify movement of the colon towards the left, anteriorly and caudally, respectively, in relation to the body. White striped columns represent no statistically significant direction of displacement. All other bars show a mean displacement that was statistically significantly different from zero. Numbers represent points that have moved in the preferential direction from a total of 60. Descend, descending; Trans, transverse; Ascend, ascending.

The British Journal of Radiology, June 2009

479

S Punwani, S Halligan, D Tolan et al Rectum

Sigmoid

Descending

Transverse

Ascending

100

75

50

25

0 0

25

75

50

100

(a) Rectum

Sigmoid

Descending

Transverse

Ascending

100

75

Figure 4. Graphical display for men 50

25

0 0

25

50

75

100

(b)

Limitations of our study include an intentional selection bias towards studies with well-distended colons on both the supine and prone scans, and single observer analysis. Although additional observers are preferable, the measurements made were largely objective and we do not believe that using more observers would substantially alter our data. Additionally, reproducible identification of our chosen colonic landmarks in supine and prone orientations in the same patient was not assessed. However, centreline measurements for these points did not change significantly, suggesting that manual matching was accurate. We have not evaluated positional changes in a collapsed or semidistended colon because technically optimal CTC is performed in a welldistended colon. It is possible that segmental mobility of a poorly or non-distended colon may be less predictable and possibly more mobile on its mesenteric attachment as a function of poor distension. 480

(a) and women (b) of the degree to which colonic landmarks moved when turning between the prone and supine positions. The x-axis represents the distance of each data point on the centreline as a percentage of the total length from the anus. The y-axis represents the mean degree of displacement in millimetres that occurred at each data point when the prone and supine positions were compared, with 95% error bars for each individual point.

Conclusions This study demonstrates and quantifies a predictable change in colonic segmental position from supine to prone orientations during CTC that is influenced predominantly by abdominal compression and pelvic motion rather than gravity. Our estimates are likely to reflect total colonic length more accurately than previous methods employed in the distended colon.

References 1. Halligan S, Altman DG, Taylor SA, Mallett S, Deeks JJ, Bartram CI, et al. CT colonography in the detection of colorectal polyps and cancer: systematic review, metaanalysis, and proposed minimum data set for study level reporting. Radiology 2005;237:893–904. 2. Halligan S, Altman DG, Mallett S, Taylor SA, Burling D, Roddie M, et al. Computed tomographic colonography:

The British Journal of Radiology, June 2009

Colonic movement between prone and supine CTC studies

3.

4.

5.

6.

7.

assessment of radiologist performance with and without computer-aided detection. Gastroenterology 2006;131:1690–9. Yoshida H, Masutani Y, MacEneaney P, Rubin DT, Dachman AH. Computerized detection of colonic polyps at CT colonography on the basis of volumetric features: pilot study. Radiology 2002;222:327–36. Summers RM, Johnson CD, Pusanik LM, Malley JD, Youssef AM, Reed JE. Automated polyp detection at CT colonography: feasibility assessment in a human population. Radiology 2001;219:51–9. Halligan S, Taylor SA, Dehmeshki J, Amin H, Ye X, Tsang J, et al. Computer-assisted detection for CT colonography: external validation. Clin Radiol 2006;61:758–63, discussion 764–5 Summers RM, Yao J, Pickhardt PJ, Franaszek M, Bitter I, Brickman D, et al. Computed tomographic virtual colonoscopy computer-aided polyp detection in a screening population. Gastroenterology 2005;129:1832–44. Yee J, Kumar NN, Hung RK, Akerkar GA, Kumar PR, Wall SD. Comparison of supine and prone scanning separately and in combination at CT colonography. Radiology 2003;226:653–61.

The British Journal of Radiology, June 2009

8. de Vries AH, Truyen R, van der Peijl J, Florie J, van Gelder RE, Gerritsen F, et al. Feasibility of automated matching of supine and prone CT-colonography examinations. Br J Radiol 2006;79:740–4. 9. Hawkes DJ. Algorithms for radiological image registration and their clinical application. J Anat 1998;193:347–61. 10. Saunders BP, Halligan S, Jobling C, Fukumoto M, Moussa ME, Williams CB, et al. Can barium enema indicate when colonoscopy will be difficult? Clin Radiol 1995;50:318–21. 11. Sadahiro S, Ohmura T, Yamada Y, Saito T, Taki Y. Analysis of length and surface area of each segment of the large intestine according to age, sex and physique. Surg Radiol Anat 1992;14:251–7. 12. Saunders BP, Phillips RK, Williams CB. Intraoperative measurement of colonic anatomy and attachments with relevance to colonoscopy. Br J Surg 1995;82:1491–3. 13. Underhill BM. Intestinal length in man. Br Med J 1955;2:1243–6. 14. Hanson ME, Pickhardt PJ, Kim DH, Pfau PR. Anatomic factors predictive of incomplete colonoscopy based on findings at CT colonography. AJR Am J Roentgenol 2007;189:774–9.

481