Two peas in a pod: digital AP, radiology

Excerpted from CAP Today – October 2007

Michael Montalto, PhD, of GE Global Research, spoke on GE’s vision of the future at the June 2007 CAP Foundation conference, "Futurescape of Pathology," in Chicago. Dr. Montalto spoke of how similarities between pathology and radiology point to a digital future for pathology. Dr. Montalto is the program leader for GE’s molecular imaging and diagnostics advanced technology program. His remarks follow.

The short story is, if you want to know where anatomic pathology is going to be in the future, look at radiology. It’s not that much different. Many folks within GE, when they hear about the rumblings within anatomic pathology, think to themselves: "We’ve heard that all before. This is nothing new."

Fundamentally, pathology and radiology are very similar. They use imaging devices. Their diagnoses are based primarily on anatomy. A subset of cases is reviewed for molecular information. In the case of anatomic pathology, we’re talking about immunohistochemistry, which is still the minority of cases in pathology, although we’re all anticipating there to be major growth in that space. Neither of the molecular techniques is very quantitative. There’s much room for improvement for quantitation in all molecular techniques, whether it be noninvasive imaging or ex vivo.

One of the similarities—and this was a shock to the folks within GE-is the image size for digitized anatomical pathology slides. The average radiology exam is maybe 100 megabytes, and the volume is about 200,000, for maybe 20 terabytes a year within the radiology department, on average. Digital pathology slides are about one to five gigs in size at the right kind of resolution for diagnostics. The case volume being the same, you’re looking at 10 to 20 times the amount of data-storage and data-handling requirements than for radiology. So although both radiology and pathology have very large image sizes, pathology is by far the heavyweight in terms of the need for large images.

That is a wake-up call to a company like GE—that if anatomic pathology were to go digital, we’d have to go way beyond what we’re already doing in radiology. Handling large volumes of images is key. We knew that in radiology some 20 years ago. We understand that now in pathology.

Another similarity with radiology is that workflow can be improved. Many people say of radiology when it was digitized, "Well, the real reason they went digital is because they got rid of film." But it was the improvements in workflow in radiology then and that exist today that drove the adoption of the technology.

So, can that happen in pathology? You bet. There is an enormous amount of workflow inefficiencies in the current lab. You have to get glass to a person with a microscope to do reviews, and then if you want to get consultations, you need to marry the glass with the patient data and move the glass to another microscope. In cases where you’re shipping the glass to reference labs, all the patient data and glass have to go and come back. If everything were digital, all that work would go away, and you would be able to access all of your cases and digital images at the same time.

Other similarities with radiology: There’s simply some things that you just can’t do unless you are digital. In the early days of radiology, you couldn’t convince a radiologist that they were going to have 64-slice CT and they were going to have to hang all of those images up on their light box. They would have looked at you like you had two heads. There was just no way. Now, thinking about putting all those images on light boxes is ridiculous. Could that happen in anatomic pathology? Absolutely.

But this is really the bottom line: Can you improve the quality of care with digital? I don’t think anybody would argue that the quality of care in radiology improved with digital imaging. With the advent of CT, we know we can get much earlier diagnoses, and we can improve the quality of care with that. One example—and there are tons of examples— is a New England Journal of Medicine paper on a study in which CT screening was done for lung cancer in asymptomatic patients. Those patients went on to be treated surgically, and there was a significant difference in outcomes in those who were or would have been asymptomatic versus those who presented with symptoms.

There is an analogy in anatomic pathology in lung cancer published in the New England Journal, not too long ago, where they went way beyond the standard anatomic pathology kinds of measurements. This was using, at Yale, automated image analysis of a particular biomarker and showing that nuclear localization of that biomarker can stratify patients in stage one non-small-cell lung cancer, and certainly those patients could be treated in a different way. And this is fully automated. So the quality of care can absolutely be increased if we have digital information.

The reasons why digital path won’t happen could have been "why digital radiology won’t happen" 20 years ago. Five years ago, a lot of this was very much true. A lot of people were not thinking that digital anatomic pathology could happen, and there were great arguments as to why it wouldn’t happen. I think the tide is turning.

One of these examples is the technology myth, that the technology simply isn’t there to scan the volume of slides needed to make a diagnosis, or that the quality isn’t there, so the perception is that the throughput isn’t fast enough. But we can scan a standard slide in 30 seconds. This is more than adequate for the majority of pathology labs in the country. You could probably have two or three machines and do more. And scan times will eventually get faster.

Another argument against it: We can’t browse a slide fast enough. I think that most browsing now has virtually no latency, or the latency is sufficiently low enough that you can make a diagnosis. We know that’s true. There have been papers published to that effect.

And another: Data storage is too great and too expensive. That’s no longer true. It’s about $1,000 per terabyte for storage, and that will continue to drop precipitously over time. There is a cost associated with going digital. If you want to know about it, talk to a radiologist. They’re a very capital-equipment-intensive department. This is not free, so you have to be getting something out of it.

But how expensive is glass? I would ask you to go back to your departments and pull out the numbers, if they exist. Somebody is collating that glass for sign-out, preparing it for shipping, receiving it, re-preparing it, re-collating it for sign-out again, re-packaging it, sending it back to the originating institution, finding lost cases-which nobody likes to talk about, preparing cases for tumor boards, which many people are doing, waiting for lab results.

Of course, there is lost revenue to competitors who are digital or are going digital. And there’s the opportunity cost of the throughput that you’re not getting because you are doing these other things. Add that up and ask, "How expensive is it to have glass?"

Another argument: The "no-value" myth. We think there is probably very real value in both workflow efficiency, reading anytime anywhere, doing QA and consults, load balancing in cases of multi-institutional hospitals—which is becoming more prevalent, and retrieving secondary reads. Just increasing diagnostic accuracy with real-time consultations. Annotating, localizing, and computer-assisted review is going to be the future if we go digital. There is real value, beyond monetary value. And this is not just our perception. We’ve spent time trying to learn from you about whether or not this is real.

If adoption is to occur, why, when, and how would it take place? There has to be an evolution. There are certainly selective pressures that are going to have to drive adoption of this technology and the evolution of this market. One of the major pressures is that post-genomic information has led to molecular diagnostics, some of which will not be done in the anatomic pathology space, though much of it could be if anatomic pathology embraces the technology.

You have to look to new ways to improve your throughput and maintain your costs. The whole computational power and the space of computation is growing, while the cost of storage is going down. That’s another pressure. That’s not only existing for you but also for other clinical departments. And, of course, a retiring workforce and a general subspecialty shortage—or a growth of certain subspecialties that are not always located where you need them.

Those are the pressures. How will it happen? Today, many anatomic pathologists use a camera on a scope. That’s the state-of-the-art for the most part. Eventually the real gateway is going to be the enabling of full digital, whole-slide imaging of whole cases. Whether or not the entire volume of slides happens, or a select number of cases happens first—probably the latter, we’ll see whole-slide imaging potentially for secondary review.

Eventually, as people get more comfortable with the technology and the technology is more readily available, it will be a no-brainer to move into primary diagnostics for some of those cases that you can do some other way, that do not require the comfort of the microscope. And then eventually this could move into a 100 percent digital environment.

What we see now, in 2007, is the very beginning of what potentially is the adoption of digital pathology within anatomic pathology. I don’t know if adoption will be fast or slow. I’m hesitant to say it will be fast. There’s a lot that has to be converted within the anatomic pathology space for it to be fast. But if it’s like radiology, it will be slow but very steady.

When is this going to happen? I’m not sure we’re 100 percent there yet with all the technology that’s needed for labs to go digital. We certainly have scanners, local storage to handle those images, software that can do automated image analysis or semiautomated image analysis, and imaging browsing. We have not built a total workflow infrastructure to go along behind all of this—which we would look at, like we did radiology, to be the foundation of widespread adoption and change.

But there’s going to be the front face, which is going to be the client side that you see, that’s going to be specific to anatomic pathology, or radiology or cardiology, or whatever the new ‘ology’ happens to be that moves to adopt digital. So this is all built on one major seamless network that’s sharing data and sharing images. That alone could drive more of the convergence or collaborative discussion between radiologists and anatomic pathologists.

Some things are not like they were and are in radiology. As I said, no real clinical infrastructure exists in anatomic pathology for workflow reporting standardization. In terms of the technology, I think we’re almost there. The primary data are not digital. We’ll never go glassless. We may in the very distant future, but I think in the foreseeable future we’re going to have slides to deal with. So the primary data will come from that.

And then there is customer acceptance. I’m probably sitting in front of a technologically savvy group by virtue of the fact that you’re here. But many other pathologists are not here and probably don’t choose to be here, and are not necessarily too concerned with any of this. There’s a big learning curve that has to happen with anatomic pathology in order for it to go digital.

So here’s the summary: big similarities between radiology and pathology. We understand this. There are great advantages to going digital, including workflow efficiencies and improved patient care. Digital will enable and drive advanced molecular diagnostics within AP. It’s just something that has to happen, like it happened in radiology, and then you’ve got all these advanced techniques that were embraced once digital occurred.

We believe anatomic pathology is poised for a digital revolution because of the pressures that will drive change. From our perspective, we think adoption is going to be slow, if it takes place, and dictated by the benefits of the total end solutions that are connected to digital and hospital infrastructure.