Nearly a decade has passed since the first computer-aided detection (CAD) product won FDA approval, and still CAD is in a nascent state—at least insofar as market penetration.

“The good news is that familiarity with the concept of CAD is widespread,” says Herta M. Klaman, president and CEO of PlusCAD Inc, Pittsburgh. “We’ve reached the point that now, no one asks what the technology is all about; now they just want to know what it’s going to cost to acquire or utilize it.”

And yet, knowing about the technology is crucial to deciding whether the time to buy into CAD is now or later. A cursory look at some of the products that were displayed at this year’s Annual Meeting of the Radiological Society of North America (RSNA) in Chicago suggests that many prospective CAD users won’t go wrong making a purchase in 2007—even though it might be another few years before the potential of CAD is fully realized.

“There is every reason to expect that soon, radiologists will have a CAD tool to assist with diagnosis of a much broader range of the anatomy than is today the case,” says Kevin E. McBride, vice president of marketing for Riverain Medical, Miamisburg, Ohio.

Agreeing with that assessment is David Sumner, CEO of the Asia markets for Medicsight, London. “Mammography CAD is paving the way for other CAD applications, including colon, lung, brain, and liver,” he says.

The way may be paved, but right now, there is a speed bump in the middle of it. “With the exception of mammography, there is no reimbursement for CAD—and that’s having an effect on research and development of new applications,” says Jeff Collins, CEO of The Medipattern Corp, Toronto. “If we ever get a code for CAD in the United States, there would be a surge in CAD products. For example, it would not be all that difficult to modify a breast ultrasound CAD system to detect thyroid cancers, since the two types of lesions present pretty much the same. But no one is currently trying to package a thyroid CAD product, and the reason is the reimbursement situation.”

Thus, for CAD makers, R&D efforts are centered largely around improving existing products. “The core product, when you get right down to it, is an algorithm, and there still are significant advances to be made in terms of accuracy,” says Ken Ferry, CEO of iCAD Inc, Nashua, NH. “What that means is we’re looking to develop higher sensitivity, higher specificity, and, at the same time, reduce the number of false marks on the image. We think there are still orders of magnitude opportunity for improvement in that space, particularly as it relates to the very difficult-to-find, subtle cancers.”

At the heart of these improvement efforts is a desire to boost physician productivity.

“Radiologists expect CAD to help with workflow so they can quickly make the diagnosis and move on to the next study,” says Daniel D. Bickford, executive vice president of sales and marketing for Confirma Inc, Kirkland, Wash. “What they want from CAD is, of course, good detection specificity; but more so, they want improved tools that will call out areas of interest more quickly. This is especially true for radiologists who spend their time looking at large, multislice image sets. In a breast MR, there are 2,000 images to look at. A good CAD tool will quickly get them to the 15 or so slices, as a volume, that point up the most worrisome curve and lead to diagnosis.”

Workflow rapidity is the name of the game when you boil it all down, Collins insists. “If we can save the radiologist time,” he says, “then his profitability will improve, even without benefit of CAD reimbursement.”

Products to Watch

What follows is a summary of what the major CAD companies displayed at RSNA.

Confirma Inc. Visitors to the Confirma booth at RSNA were shown a work-in-progress demonstration of what the company has accomplished thus far with smart morphologic operators (an effort being conducted in partnership with The Medipattern Corp).

Confirma’s CADstream product for breast MRI, according to Bickford, already comes with a morphological feature, albeit in a less-advanced form than the one the company has in development. The purpose is to help physicians determine the shape of a 3D lesion by looking at the margins of that lesion in relation to an automatically placed elliptical margin, which helps physicians determine the shape or eccentricity of the lesion, he explains.

“The American College of Radiology’s BI-RADS atlas for breast MRI advocates that physicians report both kinetic and morphologic information for lesions identified in a breast MRI study,” Bickford says. “And so the ability to more confidently report the shape and margin of a lesion has been requested by many of our customers.”

Automated features include adaptive image registration in both 2D and 3D modes, multiplanar reformatting, subtractions, angiogenesis maps, curves, maximum intensity projections, volume summaries, portfolio for reporting, and multimodality capability.

“We are continuing to expand the possibilities of optimizing workflow within CADstream,” Bickford says with regard to Confirma’s R&D efforts. “Specifically, we’re further refining our thin client architecture, which permits remote reads. It’s now typical for us to have customers operating multiple imaging centers and reading all of their MRIs off of a single, centralized CADstream server located many miles away. This is possible because we’ve augmented our thin client architecture with significant lossless compression capabilities. Images are zipped, then shot across a 20-megabit connection to the local reader. Crucially, the reader is able to see what all of his colleagues on the network are themselves doing at that exact moment with CADstream. Let’s say 10 studies are in the queue, and I’m at my PC. CADstream shows me which of those studies are being looked at by the radiologist in the cubicle next to me or in the imaging center 20 miles away, so that I can go immediately to any of those studies not being looked at and start in right there. Once I choose a study, that one becomes locked so that no one else can select it. When my colleague is ready for a new one, he’ll only be able to pick one that isn’t already being read by me. That eliminates a lot of potential for confusion and delay.”

The latest release from iCAD, v7.2 of its SecondLook software, identifies fewer false markings of vascular calcifications, as well as other false markings due to film noise/grain and linear structures, for improved workflow.

iCAD Inc. iCAD’s product line is founded upon the SecondLook family of analog and full-field digital mammography CAD systems. In September, iCAD released a new version of the software for SecondLook 300 and SecondLook 700.

“CAD v7.2 is a highly trained algorithm that automatically identifies and marks areas typically associated with cancer,” Ferry says. “The CAD software marks the primary indications of breast cancer in a consistent, logical manner closely aligned with standard mammography training. Benefits include reduced false markings of vascular calcifications and lymph nodes, as well as reduced false markings due to linear structures or film noise/grain. These result in improved workflow and a higher rate of detection. In clinical studies, iCAD has been shown to find up to 72% of the cancers that had been missed on the previous mammogram exam.”

SecondLook systems rely on patented advanced pattern-recognition techniques to find subtle areas of concern, Ferry conveys.

Meanwhile, iCAD is developing a product designed to assist in the detection of colonic polyps. “Our algorithm is really a platform,” Ferry says. “Therefore, it’s possible to adapt pattern-recognition capabilities for use in a virtual colonoscopy application. We think that eventually, the standard screening tool for healthy males and females over age 50 will be CT scanning using pattern-recognition software to detect colonic polyps.”

This new product is not expected to reach the market until the end of 2007—at the earliest. (It still must obtain FDA approval.) “It’s going to be a PMA-type approval process, and quite rigorous at that,” Ferry predicts. “This will take up to a year or more, depending on what the FDA requires of us in order to introduce the product to that market.”

Medicsight USA Inc, Nashville, Tenn. Medicsight used the occasion of RSNA to announce an online, pay-per-case CAD service, according to Sumner. For a modest fee, users will be able to use CAD through any 2D or 3D PC- or laptop-based viewer.

“OEMs have started to introduce CAD when they deliver a CT scanner-and-workstation package,” he explains. “So, for a while, a fair number of radiologists in the United States and around the globe will be equipped with very good CT scanners but not the latest workstation: the one that comes with CAD. Our online, pay-per-case product allows them to use CAD but avoid the six-figure price tag of a new workstation.”

The new Web-based clearinghouse, dubbed MedicExchange, was unveiled at RSNA, and it is expected that radiologists will be able to begin using the service in early 2007.

For those who prefer to own the equipment, Medicsight offers ColonCAD and LungCAD. The former comes with alterable filter settings that identify defects protruding into the colonic lumen and automatically highlight areas of the colon wall containing raised spherical regions. (These are then marked for further investigation.) “The sphericity threshold can be manually set by means of a simple slider bar allowing the user to discriminate between real polyps and other features that sometimes mimic the appearance of polyps, such as retained stool or haustral folds,” Sumner says. “By increasing the sphericity threshold, you can increase specificity for true polyps; by decreasing it, you can increase the overall sensitivity and draw attention to more irregular-shaped objects.”

Sumner contends that a strong suit for his company’s products is the quantity and quality of data upon which the CAD algorithm is based. “The bigger the database of consolidated CT scan information, the higher the sensitivity that it’s possible to achieve,” he says. “For the past 5 years, we’ve been collecting gold standard read data from not just the United States and western Europe, but all over the world. That’s important, because you’re going to use CAD on more than [whites]. The CAD algorithm needs to be able to recognize abnormalities in all body types, from wherever around the globe the patients come.”

The latest version of B-CAD Ultrasound from Medipattern, allows for the anlaysis of multiple lesions within the same breast as well as both breasts within the same report.

The Medipattern Corp. New from Medipattern is B-CAD MRI, a system that earned a Frost & Sullivan award in July of this year for product innovation, in recognition of B-CAD MRI’s ability to facilely integrate with a gamut of mainstream imaging applications, its reasonable pricing, and its abundant mix of features and flexibility.

Foremost among its positives, B-CAD MRI makes adroit use of pattern-recognition technology and has incorporated the BI-RADS guidelines for standardization of interpretation of breast imaging. “Users can select a region of interest, and the B-CAD software will display a gallery of up to six segmented views on the monitor,” Collins says. “Users can select any view and perform an interactive, controlled segmentation in real time for further analysis of anatomy and pathology. The software permits annotation, tagging, measuring, and automatic recording of selected views.”

The morphologic capabilities of B-CAD MRI were derived from those found within Medipattern’s original product, B-CAD Ultrasound (said to be the only FDA-approved ultrasound solution for reviewing breast lesions). “Studies have shown that morphological analyses of breast lesions are a much stronger predictor than kinetic analyses alone,” Collins says. “Our goal is to offer both morphological and kinetic analyses side by side in a belts-and-suspender approach that we feel will result in a better diagnostic package.”

The company’s latest iteration of B-CAD for ultrasound provides for the handling of multiple lesions within the same breast, as well as both breasts within the same report, Collins tells. “Also,” he says, “the report ranks the lesions in descending order, starting with the most suspicious lesion first.”

PlusCAD Inc. Pay-per-case CAD is the only offering of service provider PlusCAD. “Our customers are entirely film-based mammography facilities,” Klaman says. “None have CAD. Some never will. Many plan to eventually acquire it, but right now, they are not ready to make the significant capital investment necessary to obtain a digital mammography system, the digital workstation that goes with the modality, and then the CAD product.”

Klaman explains how her service works. “The client facility performs a mammogram and places the films in an insert folder,” she says. “The films are picked up daily and transported to us at PlusCAD by an express courier service. We digitize the images, perform a CAD analysis, and produce a paper printout. The mammography films and the accompanying CAD printout are then transported back to the mammography facility for interpretation. The facility and the interpreting radiologist bill third-party payors for the mammogram as the primary procedure—but they also include in that same billing an add-on code for the CAD analysis. When reimbursement is received, the facility pays our CAD fee. That fee is lower than the reimbursement, so the facility makes money with CAD without actually having invested in the technology.”

R2 Technology Inc, Sunnyvale, Calif, a Hologic Co. Launched earlier this year was version 2.1 of R2’s ImageChecker CT Lung CAD system. This update features enhancements to the product’s pulmonary artery patency exam tool, which now provides volume calculations for filling defects. This tool is available separately for physicians who prefer focusing on potential pulmonary artery-filling defects rather than lung-nodule detection, the company conveys.

V2.1 also includes expanded PACS integration. Chiefly, users will find that they now can send direct to an enterprise PACS workstation the results of lung-nodule detection performed in conjunction with ImageChecker’s tool for temporal comparisons. This allows more efficient side-by-side observation of current and prior CAD results. The tool automatically matches and tracks lung-nodule changes—size measurements, density, and estimated growth-doubling time—in order to minimize intrareader measurement variability and to expedite case review, according to the company.

R2 made available at RSNA an assemblage of clinical studies validating its technology. One such previously presented study demonstrated that ImageChecker detected significant lung nodules in 33% of 100 multislice CT cases previously reported as having no nodules. Another study showed that ImageChecker’s pulmonary artery patency exam tool caught 88% of segmental pulmonary emboli and 78% of subsegmental pulmonary emboli, with a median of four false CAD marks per normal case.

Riverain Medical. Riverain demonstrated continued improvements to its number one product, RapidScreen Digital lung CAD. “It is becoming an even more efficient radiologist tool that can be used on any workstation,” McBride says.

RapidScreen Digital lung CAD from Riverain features a modified algorithm to better detect subtle nodules and cancers in the upper chest.

RapidScreen Digital provides an efficient way for digital x-ray departments to add early lung-cancer detection to their capabilities. It evaluates each digital PA/AP image based on a comprehensive set of unique signatures to rapidly identify nodules that warrant further review. Circled nodules—as small as 9 mm—quickly direct reviewers to areas of concern, McBride says.

“At RSNA, we also showed continued improvement of our algorithm,” he adds, explaining that more than 1,000 chest radiographs—each evidencing T1 lung cancers as confirmed by CT, follow-up, or biopsy, and more than 10,000 cancer-free x-rays as confirmed by 3 to 10 years of follow-up—went into its calculation. “Because of the improvements we’ve made, our product is better able to show subtle nodules and cancers in the upper areas of the chest, which are the areas most problematic for radiologists. Other improvements to the algorithm have enabled us to further reduce false-positives, thereby speeding up the radiologist’s reading of the image.”

McBride stresses that RapidScreen is strictly a review tool that operates in the background. “It does not interrupt department workflow,” he says, “and it’s compatible with all modalities that generate DICOM 3.0-conformant DR, CR, and digitized images.”

Rich Smith is a contributing writer for Medical Imaging. For more information, contact .