Progress in Refining Virtual Colonoscopy for Colorectal Cancer Screening
Article Outline
Evidence-based guidelines recommend that all asymptomatic, average-risk women and men be offered one of five options to screen for colorectal cancer: fecal occult blood testing, flexible sigmoidoscopy, the combination of fecal occult blood testing and flexible sigmoidoscopy, barium enema, or direct colonoscopy.1, 2 Most gastroenterologists now prefer the option of direct screening with colonoscopy because it clearly is the most accurate way of detecting early cancers, and of both detecting and resecting most premalignant polyps.3 However, many healthy asymptomatic people seem reluctant to endure the inconvenience and perceived discomfort of colonoscopy. In addition, a recent comprehensive study by the Centers for Disease Control and Prevention (CDC) indicates that the capacity to conduct direct colonoscopy screening of the 41 million eligible average-risk Americans over age 50 years who have not yet been screened may be severely lacking.4 That study showed that, if half the currently available U.S. colonoscopy capacity was dedicated completely just to screening, it would take 10 years to complete the task and reach a steady state.
Virtual colonoscopy (VC), or computed tomography (CT) colonography, is a very promising, relatively new technique that likely soon will bolster our screening capacity and increase overall screening compliance. VC uses data obtained from rapid helical CT scanning of the abdomen to create computer-reformatted 2- and 3-dimensional images of the colon. These images can be rotated for different views and even combined for a complete 3-dimensional view of the colon that then can be rapidly “flown through,” thus simulating conventional optical colonoscopy. The history of VC still is fairly recent and improvements and refinements continue to be made each year that move it closer to being included as an additional screening option in the guidelines. The technique initially was developed and introduced by Vining at Wake Forest University. To the sounds of Wagner’s “Ride of the Valkyries,” he showed his first colonic scans at the 1994 annual meeting of the Society of Gastrointestinal Radiologists.5 The next VC presentation occurred at the 1994 National Cancer Institute’s International Workshop on Colorectal Cancer Screening. Vining states that “at this meeting, the gastroenterologists in attendance were having a great time bashing the radiologist’s defense of the barium enema.”6 When he introduced the VC concept, he began his presentation with, “it’s the bottom of the ninth inning, score is 3-0 in favor of the gastroenterologists, bases are loaded, and a new radiologist is up to bat.” It was clear that the gastroenterology community, after seeing virtual colonoscopy in action, realized that a new radiology procedure could replace barium enema and impact the future of colorectal screening. This imaginative new method of imaging the large bowel has undergone remarkable development and improvement since that dramatic inauguration over 10 years ago.
VC has several obvious advantages over conventional colonoscopy. Examination time is much shorter and there is no need for IV conscious sedation with its attendant cost and complications. Patients may return to their regular activity soon after their scan. The procedure has little immediate risk, allows scrutiny of both sides of the bowel wall and of bowel folds, and precisely localizes lesions. It can examine the proximal colon when colonoscopy is incomplete, such as when a distal obstructing cancer prevents passage of a colonoscope. Disadvantages of virtual colonoscopy include still the need for a very thorough bowel cleansing preparation, a somewhat disagreeable gas distention of the colon, and some radiation exposure. Also, these complex scans currently require appreciable expensive radiologist time to set up and read. Lastly, VC is diagnostic only; whenever a clinically significant neoplasm is found, the patient must undergo a subsequent colonoscopy to biopsy or resect the lesion.
The accuracy of VC has been compared with that of conventional colonoscopy in a number of prospective studies. Several earlier studies, performed in high-prevalence populations, have demonstrated good sensitivity for detecting polyps ≥1 cm in diameter.7, 8, 9 However, even in the best of these series, sensitivity and specificity for detecting smaller polyps fell off rapidly. Other studies have not achieved comparable results. Three recently published comparison studies reported sensitivities for detecting large (≥1 cm) polyps of only 46%–59%.10, 11, 12 These studies, however, have been criticized for having poor quality control, including scans performed several years before the results were actually published, or for using suboptimal or dated methodology.
A recent multihospital study reported by Pickhardt et al in a low-prevalence, largely average-risk population of 1233 asymptomatic adults represents a major breakthrough in the development of VC for use in colorectal cancer screening.13 In this large comparison study, 6 experienced radiologists used multidetector CT scanners and a commercially available CT colonographic computer system (Viatronix, Stony Brook, NY) that creates a 3-dimensional endoluminal display for the initial detection of polyps, followed by rapid confirmation of findings with corresponding 2-dimensional images. Patients underwent a standard colonic preparation and consumed 500 mL of dilute barium for solid-stool tagging and 120 mL of diatrizoate solution for opacification of retained luminal fluid. This preparation allowed the computer to differentiate between retained stool and polypoid defects, and for electronic fluid cleansing. Remarkably, the sensitivity of VC in this study was 93.8% for adenomatous polyps at least 1 cm in diameter, 93.9% for polyps at least 8 mm in diameter, and 88.7% for polyps at least 6 mm in diameter. The authors concluded that VC, using this advanced methodology, was as accurate as conventional colonoscopy for the detection of clinically important colorectal polyps (≥6 mm) in asymptomatic, average-risk adults.
Now these investigators have used the data sets generated by this highly successful study to test a computer software system that is capable itself of finding colorectal polyps using algorithms that detect irregularities in the colon wall contour. This computer-aided polyp detection system (CAD), developed by computer and radiology researchers at the National Institutes of Health, and previously tested in pilot studies at the Mayo Clinic, Rochester, Minnesota, was first adapted or “trained” to detect polyps by applying the program to a randomly selected one-third of the of the study’s data sets, and the resulting program then was used to read the images from the remaining two thirds of screened patients (the “test” group). It should be noted that, although this is a re-analysis of images and data from a previously completed and published study, Pickhardt et al included their plan to use the study’s scans and comparison data to later test the performance of this CAD system in their initial Institutional Review Board application. As reported in the current issue of Gastroenterology, for the “test” set of CT colonographic scans, CADs per polyp and per patient sensitivities were both 89.3% for detecting retrospectively identifiable adenomatous polyps at least 1 cm in diameter.14 The CAD system’s per patient sensitivity for detecting 8-mm and 10-mm adenomas were not significantly different than those of optical colonoscopy reported in the original Pickhardt study comparing the performance of the two modalities. The successful application of CAD to a randomly selected portion of Pickhardt et al’s screened patients indicates that the method likely is generalizable to newly generated colorectal scans of comparable quality (previous feasibility studies of CAD performed computer “training” and “testing” on the same data sets).
How will the development of high-performance CAD technology help make VC acceptable for population-based colorectal cancer screening? Pickhardt et al already have shown that VC without CAD, using their advanced methodology of stool-tagging, electronic fluid cleansing, and software that uses primarily 3-dimensional detection of polyps with 2-dimensional confirmation, performs as well as optical colonoscopy for the detection of colorectal neoplasia. However, other centers have reported considerable radiologist intraobserver variability in interpretation of VC scans as an important cause of false-negative results, and that the reading of scans is often time-consuming and therefore both stressful and expensive.10 Some trials have compensated for intraobserver variation (or radiologist errors of perception) by using consensus interpretation by two or more readers to boost sensitivity. This practice, however, is unlikely to be practical or economical in clinical community screening practice. Current charges for screening VC are slightly higher than that of an abdominal/pelvic CT scan. However, if the indication for an examination is screening, additional colonoscopies will be needed in 10%–20% of patients to assess findings or resect polyps. In these cases, a more cost-effective approach might be just to do an initial colonoscopy that is both diagnostic and therapeutic in a single sitting with a single bowel preparation. To compete with direct colonoscopy screening, therefore, the price of a screening VC probably would have to drop substantially below that of conventional colonoscopy in order not to dramatically increase the overall cost of a screening program. Therefore, the use of an effective CAD program during the radiologist’s reading of a CT colonography scan likely would not only improve sensitivity of the study for detecting neoplasia, but might also reduce total expensive reading time, thus allowing the total cost of VC screening to be reasonably contained.
An important disadvantage of the CAD performance in the current study was its relatively low specificity (a high rate of false-positive results). The impact of these false-positives (2.1 and 6.7 false-positives per patient at the 10-mm and 8-mm polyp thresholds, respectively) depends on how quickly the reading radiologist can review CAD “hits” and how difficult it is to decide if a CAD hit is true or false. Although the authors have not yet systematically studied this problem, they state in the current article’s discussion that most false-positives were readily identified to be normal structures, such as the ileocecal valve or normal colonic folds. Furthermore, few (only 0.9%) of the CAD false-positives coincided with false-positives contained in the radiologist’s reading. Further quantitative study of this issue clearly is needed. Other current limitations of CAD are failure to detect flat polyps, and a lower sensitivity for detecting polyps located in the fluid-filled part of the colonic lumen compared with polyps located in the air-filled part.
In summary, VC continues to rapidly evolve into an accurate and practical method for screening for colorectal cancer that should bolster our capacity to screen the large, as yet unscreened average-risk population. Automated computerized programs to detect neoplasia (CAD) promise to improve the accuracy of VC scans; at the same time, they likely will reduce expensive radiologist reading time. If there concomitantly also is widespread adoption of the advanced methods used in the original Pickhardt study, VC undoubtedly soon will be added to the colorectal cancer guidelines’ menu of acceptable screening options. Vining earlier likened the colorectal cancer screening debate to a “baseball game between gastroenterologists and radiologists.” This observer predicts a much more cooperative effort between the two specialties during the next few years. By “playing on the same team,” gastroenterologists and radiologists hopefully can resolve remaining issues that will allow VC and optical colonoscopy to improve the level of screening of our underscreened at-risk population. After further refinements, VC likely will provide needed screening capacity and will improve screening compliance. Colonoscopy demand likely will increase, not decrease, as screening rates rise, but it will become more frequently therapeutic.14
References
- . Colorectal cancer screening and surveillance: clinical guidelines and rational. Update based on new evidence . Gastroenterology . 2003;124:544–560
- . American Cancer Society guidelines for the early detection of cancer (update of early detection guidelines for prostate, colorectal, and endometrial cancers) . CA Cancer J Clin . 2001;51:38–75
- . Rationale for colonoscopy screening and estimated effectiveness in clinical practice . Gastrointest Endosc Clin N Amer . 2002;12:65–76
- . Is there endoscopic capacity to provide colorectal cancer screening to the unscreened population in the United States? . Gastroenterology . 2004;127:1661–1669
- Vining DJ, Gelfand DW. Non-invasive colonoscopy using helical CT scanning: 3D reconstruction, and virtual reality. Presented at the 23rd Annual Meeting, Society of Gastrointestinal Radiologists, Maui, HI, February 1994.
- . Virtual colonoscopy (the inside story) . In: Dachman AH editors. Atlas of virtual colonoscopy . New York: Springer-Verlag; 2003;p. 8–10
- . A comparison of virtual and conventional colonoscopy for the detection of colorectal polyps . N Engl J Med . 1999;341:1496–1503
- Optimization of CT colonography technique (prospective trial in 180 patients) . Radiology . 2000;216:704–711
- . Colorectal neoplasia (Performance characteristics of CT colonography for detection in 300 patients) . Radiology . 2001;219:685–692
- . Prospective blinded evaluation of computed tomographic colonography for screen detection of colorectal polyps . Gastroenterology . 2003;125:311–319
- . Computed tomographic colonography (virtual colonoscopy) (a multicenter comparison with standard colonoscopy for detection of colorectal neoplasia) . JAMA . 2004;291:1713–1719
- . Analysis of air contrast barium enema, computed tomographic colonography, and colonoscopy (Prospective comparison) . Lancet . 2005;365:305–311
- . Computed tomographic virtual colonoscopy to screen for colorectal neoplasia in asymptomatic adults . N Engl J Med . 2003;349:2191–2200
- . Computed tomographic virtual colonoscopy computer-aided polyp detection in a screening population . Gastroenterology . 2005;129:1832–1844
PII: S0016-5085(05)02187-6
doi:10.1053/j.gastro.2005.10.031
© 2005 American Gastroenterological Association. Published by Elsevier Inc. All rights reserved.
Refers to article:
- Computed Tomographic Virtual Colonoscopy Computer-Aided Polyp Detection in a Screening Population , 04 October 2005
