Shawn Rhea: Hello, I'm Shawn Rhea, medical technology reporter for Modern Healthcare magazine. We're here today with Jason Launders, senior project officer and a medical physicist at the ECRI Institute, which is a not-for-profit company that evaluates medical products. Jason is here today to talk to us about the changing landscape of computer-assisted detection. Hello, Jason.
Jason Launders: Hello, Shawn.
Shawn Rhea: Computer-assisted detection is a technology that is used to mark questionable areas on a medical image and draw a radiologist's attention to a particular problem. The technology has been used particularly in breast imaging for well over a decade and many companies are anxious to get versions of the product on the market. What are the benefits and what are the limits of CAD technology?
Jason Launders: Well, computer technology is ideally suited to automating repetitive tasks. While the human eye is very well adapted to detecting patterns in images—and is better than computers, in fact—maintaining a high level of concentration is not something that humans do well. There are many cases in which a lesion has been missed that was easily detectable in retrospect, so CAD fills a scan and helps detect overlooked details—in effect acting as a second pair of eyes. The drawback of the technology is that computers are not well-suited at processing image data. A digital image is essentially a [unclear] of numbers that requires computationally intensive processing to identify patterns. And it's not the 2-D representation that we think of as an image. And, really importantly, any noise in that image—that's randomness in the numbers—disrupts the processing. As a result, CAD is susceptible to what we call false positives, which have the effect of needlessly drawing attention to otherwise healthy anatomy in the images. And the radiologist must spend time to inspect and to decide whether to act on those marks that the CAD calls up.
Shawn Rhea: So the technology is essentially seeing things that aren't a problem area.
Jason Launders: Absolutely.
Shawn Rhea: OK. So the Food and Drug Administration [unclear] proposed guides, which to clarify the types of safety and efficacy data that manufacturers of CAD technology need to submit when applying for premarket approval of new technology or [unclear] of products that are—some of those that are already on the market. If the guidance is implemented as written, what effect [unclear] hanging on the development of CAD technology?
Jason Launders: Well, as we at the ECRI Institute see it, the issue with the FDA's guide document is a recommendation for attainable testing. Obviously, some type of testing on clinical images is required; however, the recommendations do not appear to allow the wide range of CAD systems that are now available or becoming available. The same level of clinical testing is being recommended for both [Unclear]-aided detection and the more advanced [Unclear]-aided diagnostics systems, i.e., the systems that provide more than the simple detection or a device that automatically [Unclear] the studies, instead of being viewed by radiologists. In reality, CAD systems occupy a [Unclear] between a simple automation and advanced diagnosis. This type of testing is onerous, time-consuming and very expensive, which seems to be disproportionate for some of the simpler detection algorithms. So, the FDA's guidance is likely to hinder development of new algorithms. So some clearer definitions are required within the guidance.
Shawn Rhea: You mentioned a little earlier that there's a broad spectrum of CAD technology. So how broadly are hospitals and imaging centers currently using CAD? For example, what kinds of imaging studies is CAD being used in conjunction with and [unclear].
Jason Launders: Well, we at ECRI Institute evaluated CAD systems about eight years ago, and since—that was on CAD and [Unclear] systems, and since then they haven't really changed very much. But they have become quite widely used in breast-cancer screening throughout the world in fact. And breast-cancer screening is an ideal application for CAD due to the fact that it is a very high-volume procedure and has a relatively low probability that an individual image will contain a lesion. CAD's also used quite widely in mammography, in which the CAD extracts information from a series of images and plots the data with respect for time, which is a very tedious process and something computers are ideally suited for. Other than that, CAD applications are available for helping colonoscopy and [Unclear]-module detection, and have been approved by the FDA, although we haven't seen their use being widespread mainly due to the lack of reimbursement.
Shawn Rhea: OK, and you hit upon a point that I was hoping to explore a bit. And CAD is not reimbursed as a stand-alone service by payers, but many people do seem to think that it will be an integral part of imaging in the future—
Jason Launders: Yes. It's not reimbursable for most outpatients. It is reimbursed in screening mammography.
Shawn Rhea: But given, though, its limited reimbursement, why do so many providers think that it is a key technology that they're going to be pushing [Unclear].
Jason Launders: Well, I think the main fact is that radiologists miss things every day. In retrospect it's probably to be shown that most imaging studies only hit about 80% sensitivity. And a lot of the findings that radiologists miss are easily visible in retrospect. So a way of cutting down those errors. And the other issue is that more and more images are being generated today have 320-slice CT machines, which are producing many hundreds of images to study, and CAD is seen as a way to help fuel those images in a timely manner—make the work flow within radiology manageable.
Shawn Rhea: So in other words, it would be nearly impossible for a physician to sit down and go through all of those images in the same space of time and with the same level of accuracy as [Unclear]?
Jason Launders: Yes, 'cause a computer just never gets tired and humans do [Laughs].
Shawn Rhea: So what area of imaging do you anticipate CAD technology having an impact on next?
Jason Launders: Well, [Unclear]-module detection, which I just mentioned has been available for some time. However, the full potential of those technologies has yet to be realized in our view. But really, ECRI Institute anticipates that CAD applications will be largely invisible and become rather integrated with the whole image-viewing technologies. So there may well be a kind of reduction of this CAD and image viewing, and just merging all into one. And it will just become a seamless part of the work flow. And maybe in 10 years time, CAD—we won't be talking about CAD as a separate entity.
Shawn Rhea: Well, Jason, I think we've covered some good ground here. I appreciate your taking some time out to chat with us a bit about the future of CAD technology. Again, I'm Shawn Rhea, medical technology reporter for Modern Healthcare magazine, here today with Jason Launders of the ECRI Institute. Thank you for joining us.
