Home

Cover Story

Table of Contents

E-Newsletter

Article Archive

Editorial Calendar

Datebook

Writers' Guidelines

Orgs/Links

Opinion Polls

Reprints

Forum

For other articles and previous issues click here.

July 5, 2004

CAD: Driving CT Colonography to Clinical Settings
By Dan Harvey

Computed tomography colonography, commonly known as virtual colonoscopy or CTC, is generating strong interest because its minimally invasive nature could make it more widely utilized than conventional colonoscopy, which would, in turn, save more lives.

Viatronix received FDA 510(k) clearance for its V3D-Colon system this spring, but the virtual procedure is still considered an emerging technology. It poses challenges that make it impractical for widespread clinical use in a screening setting, including a high false-positive rate. False positives can lead to unnecessary follow-up studies. Also, reading times can be long, as virtual colonoscopy requires an enormous number of images. Difficult cases can take from 40 minutes to one hour to read. For virtual colonoscopy to be practical in a busy clinical setting, reading times must be greatly reduced. But the perceived potential and results from ongoing studies keep researchers focused on developing techniques that will make virtual colonoscopy widely acceptable in a screening environment.

Researchers are investigating computer-aided detection (CAD) technology in hopes of addressing the specificity question and possibly the time issue. Vendors have already started developing products for a latent market. Still, a lot of work lies ahead, and CAD for virtual colonoscopy—like the procedure itself—is still in the early stages of development.

Abraham H. Dachman, MD, for one, feels that CAD could be the vehicle that takes virtual colonoscopy into the general clinical setting. “CAD will be extremely helpful in increasing radiologists’ confidence and accuracy of interpretation, and because of that, it will go a long way in helping virtual colonoscopy become more widely available,” says Dachman, who is director of CT in the department of radiology at the University of Chicago, where some of the most important research into CAD for virtual colonoscopy has been conducted.

Colorectal Cancer and Compliance
In the United States, colorectal cancer is the third most common form of cancer diagnosed and the second-leading cause of cancer-related death. Each year, nearly 55,000 men and women die from the disease and approximately 120,000 new cases are diagnosed. Despite those dismaying statistics, it could be avoided. Colon cancer takes a long time to develop. Its precursor—the adenomatous polyp—can take up to 10 years before it evolves into a cancerous lesion. This long incubation period would appear to make it the one form of cancer most easily policed by preventive screening. Early detection results in successful treatment of colon cancer for almost all patients. Several screening options—including fecal occult blood testing, sigmoidoscopy, and colonoscopy—are widely available and relatively inexpensive. The reason the statistics remain so high is a general lack of compliance.

As it turns out, only approximately one-third of colorectal cancers are found at early stages. Many people ignore the reasonable and prudent screening guidelines advanced by the American Cancer Society (ACS) because the prospect of a colonic examination is unpleasant. Medical professionals hope that virtual colonoscopy’s “patient-friendly” elements will improve the compliance rate and, in turn, reduce the colon cancer death rate.

However, virtual colonoscopy has its limitations that hold it back from widespread acceptance and application. Both the ACS and the American Gastroenterological Association believe it’s still an emerging technology in need of technological improvements and more clinical testing. Physician turf issues aside, the challenges to virtual colonoscopy most often relate to its accuracy and the required amount of interpretation time.

Researchers—most notably at the National Institutes of Health (NIH) in Bethesda, Md., and the University of Chicago—are addressing those challenges by applying CAD to virtual colonoscopy, developing complex algorithms and CAD schemes to decrease false-positive rates and interpretation times and to improve reader performance.

Advantages and Limitations
Several studies support virtual colonoscopy as a safe and accurate alternative to colonoscopy for detection of polyps. Because it is less invasive, there is little risk of injury, and it requires no anesthesia. In addition, subjects involved in studies report that they prefer virtual colonoscopy over conventional colonoscopy. Virtual colonoscopy involves less preparation and recovery time, and the discomfort level is lower.

Also, recent studies have indicated that it can be just as effective as conventional colonoscopy for detection of polyps 5 millimeters and larger in the average-risk population. Moreover, it’s viewed as a viable alternative for patients who can’t undergo the procedure or have had a failed colonoscopy.

But virtual colonoscopy also involves a significant learning curve. CT images are notoriously hard to read, requiring special training. Currently, many radiologists are not educated in how to perform virtual colonoscopy, and not many courses are offered. Also, when a virtual exam identifies polyps, a conventional colonoscopy must then be scheduled to remove them. If the screening procedure is a conventional colonoscopy, polyps can be removed in the same procedure.

“To have this [virtual colonoscopy] be an acceptable procedure, the workflow components offered by CAD are essential for use in the clinical setting,” says Susan Wood, PhD, vice president of CT products for R2 Technology of Sunnyvale, Calif., a company involved in the development and commercialization of CAD technology. “You not only need increased accuracy but a more efficient overall workflow, too. Otherwise, you have an application that takes 20 or 30 minutes to get through. That’s not acceptable in a clinical setting, especially in a screening environment.”

CAD’s purpose in colonography is to reduce both perceptual error and interpretation times and make it easier to identify lesions. “The role of CAD is to improve the accuracy of the radiologist to detect colonic polyps and to improve the productivity of the radiologist searching for these polyps,” says Wood.

CAD is currently being used with other modalities, most notably mammography, where it addressed the same problems associated with virtual colonoscopy. It has helped identify regions of interest and presents these regions in a fashion that fosters more accurate assessment. CAD is still widely considered a research tool still in the relatively early stages of development. Still, in many clinical trials, the technology has produced encouraging results.

For CTC, the primary functions of CAD are identifying polyps according to characteristic features and distinguishing detected sites as polyps or false positives. “CAD functions by analyzing the mucosal surfaces and looking for things that look like polyps or masses,” says Dachman. “It may be refined to look for flat lesions as well. But current shape-based analysis is looking for rounded structures.”

Two useful features that help accomplish this are:
1. Surface shaping. This identifies and characterizes polyps using a curvature assessment algorithm to define its shape. Colonic polyps typically have a rounded contour and an elliptical peak shape.
2. CT attenuation. This helps to distinguish polyps from false positives. False positives typically have low CT attenuation and polyps tend to have soft tissue attenuation.

Increasing Sensitivity
Dachman believes CAD will help pick up missed lesions and increase sensitivity, but he doesn’t feel it will have a significant impact on interpretation time. “It may increase or decrease it, but only slightly,” he says.

Like Dachman, Ronald M. Summers, MD, PhD, chief of the imaging process group at the NIH, is also unsure that CAD will greatly affect interpretation time. “I think it’s unclear whether it will help speed up the interpretation of the exam,” he says. “It depends on whether it is used on the first-reader or second-reader paradigm, which is being discussed in the CAD community.”

Summers explains that most CAD is used in the second-reader paradigm, where the physicians analyze the study, make a decision, then consult the CAD and change their decision if they wish. In the first-reader paradigm, the CAD reads the case and the physician looks only at the results and makes a decision. “In radiology, we don’t have any examples of the first-reader paradigm, so I am not sure if the CAD will actually speed up interpretation,” says Summers.

But Summers thinks virtual colonoscopy will significantly impact accuracy. “My feeling is that it can reduce interpretation error, and that may improve the overall quality of the exams in terms of reducing error rates,” he says. “It may also improve diagnostic confidence, which can make an exam more acceptable to physicians who are thinking about performing it.”

Similarly, Dachman believes CAD will help increase radiologists’ confidence and interpretation accuracy. “That will go a long way in helping CTC become more widely available,” he says.

Pioneering Researchers
Since 1994, Dachman has been involved in CAD-CTC research at the University of Chicago department of radiology, which has been one of the top pioneers in this area. The department has developed cutting-edge CAD software to help identify polyps and masses. One of its most important accomplishments has been the development of computerized false-positive reduction schemes that automatically locate suspected polyps for quick and accurate detection of suspicious lesions.

The department’s ongoing studies have produced some of the most compelling information about CAD. Recently, its researchers conducted an observer-performance study that demonstrated how CAD can provide a computerized “assistant” to radiologists in detecting colonic polyps. In the study, researchers collected 20 data sets (10 abnormal sets with 11 polyps ranging in size from 5 to 12 millimeters and 10 normal data sets).

These sets were read by four experts twice—once with the CAD system and once without. Without the system, they missed 42% of the polyps. With CAD, they detected 75%.

The NIH is also at the vanguard of CAD-CTC research. In 1997, when investigating CAD for CT bronchography, researchers recognized that the technology was readily transferable to colon polyp detection. That led to research and a series of studies that indicated the feasibility of CAD-CTC. NIH researchers recently conducted a study on how CAD worked with contrast-enhanced CTC.

The study involved 29 patients with suspected polyps or masses who underwent CTC (contrast and noncontrast) and conventional colonoscopy. The results were then compared. The researchers found that CAD detected 27 of the 34 polyps or masses confirmed by colonoscopy. They also found that supine contrast-enhanced CT performed slightly better than the noncontrast prone exam. They concluded that there was a higher sensitivity and a lower false-positive rate with the CAD scheme. Further, the contrast improved the sensitivity. But Summers emphasizes that the findings were only preliminary. “It is early in the research,” he says. “Contrast may be useful for characterizing lesions, but there really is no data yet.”

At this point, Summers finds it hard to speculate on how long it will take for CAD to bring CTC into widespread use. But he reveals that companies have been actively involved in research and development of CAD products for virtual colonoscopy, and he expects to see the fruits of their efforts become more visible in the next year or two.

“It’s hard to say when a wide application will occur, but a number of companies are working on products. My guess is that at the next RSNA, you’re going to see CAD highlighted a lot more,” he comments.

Industry Development
One such company is R2 Technology. At RSNA 2003, the company exhibited its investigational CAD tools for virtual colonoscopy. Around the same time, it entered into a licensing agreement with the University of Chicago for the CAD tools developed in its radiology department. “We are in the process of developing the colonoscopy application as an adjunct to our current ImageChecker CT product line,” says Wood. “Right now, it’s still in development, but we anticipate having a product in a year.”

The company’s patented ImageChecker CT CAD system features a CT display workstation that combines dedicated software and hardware that provide tools to aid radiologists in reviewing CT exams. The company reports that the workstation receives CT exams and DICOM structured reports of CAD-identified areas using DICOM protocols. According to the company, radiologists can view screens that summarize their findings after review of each exam, and the information can be saved to a local database in the workstation for later review.

Siemens Medical Solutions of Malvern, Pa., is currently developing a CAD component for its syngo Colonography software package. According to the company, the syngo technology combines enhanced digital image processing tools with optimized workflow and reporting capabilities. The package helps clinicians in studying intra-lumenal and extra-lumenal views, as well as the colon wall.

Siemens reports that it has been working with researchers from New York University and The Cleveland Clinic in developing its CAD work-in-progress. “The reason we’re doing it,” explains Ingo Schuecking, from Siemen’s CAD division, “is that we believe that CAD, as a new feature, will be able to enhance the CT colonography software. What is important is the high performance of the algorithm in detecting lesions with not too many false positives. The second thing is efficient workflow. The user can evaluate the findings very quickly and determine whether they are true positives or false positives. This will make it an acceptable in the busy radiology department.”

Looking Ahead
Virtual colonoscopy demonstrates potential as a useful screening test for colon cancer. Its “patient-friendly” elements make it a more attractive option to the screening population that has demonstrated an aversion to colonic testing. As such, it could ultimately reduce the death rate associated with the disease. CAD might help CTC evolve into an accepted procedure in the clinical setting, and many medical experts believe improved CAD systems can be the key to the widespread acceptance of virtual colonoscopy as a practical and effective screening tool.

— Dan Harvey is a contributing editor for Radiology Today.

Subscribe to Radiology Today Magazine!