Through Thick and Thin: The MGH Search for a 3D Solution
Imagine that you run a large city, and that up until now, the sole transportation available has been buses—but with the price of automobiles coming down, citizens suddenly want to drive cars. 
Keith Dreyer, DO, Phd This situation is somewhat analogous to what’s happening to 3D imaging in the radiology department at Massachusetts General Hospital (MGH), Boston. For a decade, MGH has spent large sums of money creating a world-class 3D imaging lab. Technologists have been rigorously trained to provide radiologists of various subspecialties with intricate, highly augmented 3D studies. According to Keith Dreyer, DO, PhD, vice chair of radiology and informatics at MGH, of the 750,000 radiological exams done by the hospital each year, more than 50,000 now go through 3D postprocessing in the lab. To create these studies, technologists sit at specially designed 3D workstations where they follow highly evolved protocols to prepare the images for interpretation. The huge advantage of the 3D lab model is that radiologists are being fed image sets already prepared for them. The radiologists can move quickly and efficiently from one case to the next without having to create views themselves. The difficulty with the lab model, however, is exactly that—radiologists can’t interact with the studies to create their own views. Like passengers on the bus, they are dependent on the technologist drivers. If they want a special view, they have to request it. The bus has to turn to let them off at that stop. Dreyer estimates that 90% of the studies rendered through the MGH 3D lab require no further work. It’s the remaining 10% that have to go back to the technologists for refinement. This wouldn’t be necessary (or, at least, it would be less necessary) if the radiologists could reformat the images themselves. Fat and Thin The 3D workstations where the technologists work run as fat clients. They contain the software necessary to create the 3D studies, and anyone who wants to create such a study has to go to the fat-client workstation and create the study there. Now, new thin-client technology has evolved, enabling the end user—the radiologist or referring physician—to create 3D studies at a conventional radiology workstation, or even at a desktop PC. For many hospitals, this thin-client technology has been a godsend; smaller institutions can have their 3D renderings without gigantic outlays for a postprocessing lab, Dreyer notes. Because MGH has its lab already in place, however, and because it is such a large institution producing so many exams, the adoption of thin-client technology has been troublesome. One of the problems has been scalability. Thin-client technology involves sending image data from the modality to a server (most often, from modality to PACS to a series of servers) that contains software to allow the 3D postprocessing of that image data. This centralized computer server acts as a master workstation where the image data can be manipulated prior to interpretation. The centralized server uses Web-based technology to stream the 3D data out to other workstations or PCs, where radiologists or clinicians can manipulate the 3D images to suit their specific needs. With a thin client, the passengers on the bus become drivers of their own cars. If they need to remove bone electronically to study vascular structure, they don’t need to wait for a technologist. Radiologists using a thin client can interact with the original 3D studies (unlike the set views they receive from the lab), zeroing in on what they want to see. As Dreyer notes, one of the problems with interactive thin clients is that the vendors have not yet solved the problem of allowing simultaneous access to studies by a large number of users. When an institution needs concurrent access for only a dozen or so end users, the thin-client servers work well. Dreyer says, however, “I’ve got 5,000 physicians who might all want to look at the same time—and realistically, I probably have 100 who might all want to look at the same time, so it’s really difficult for me. That’s why I’ve been working with vendors for years, so I can scale. It’s a real challenge, right now, to scale these things up to a big size, so that’s why we’ve been at this for a while.” Enter Visage One company that has built its advanced visualization platform from the ground up on thin-client technology is Massachusetts-based Visage Imaging, a subsidiary of the Australia-based informatics organization Pro Medicus Ltd. Early on, Dreyer was named head of the medical advisory board at Visage. He says that he quickly began to stress the importance of scalability. “It was one of my self-serving goals to say, ‘You’ve got to make this thing scalable,’” he explains. “The company was, by far, the most advanced in having that ability to scale up there. That’s why we went with it, even though it was new to the market,” he says. For two years, Dreyer says, MGH worked with Visage in clinical trials to achieve functionality that would meet the hospital’s needs. “It takes a long time to make these applications. It takes a long time to make a fantastic vessel package, or a cardiac package, or a stroke package. These are clinical applications we’ve been chipping away at for several years,” Dreyer says. Still, unraveling the software knots is taking time. “Let me tell you where I’d like to be,” he says. “I would like to have all CT and MRI available for thin-client visualization. That’s our goal.” Lab Value Ironically, one of the impediments to a thin-client rollout at MGH has been the success of the 3D imaging lab itself. The bus has momentum that keeps the passengers from looking too hard in the direction of the more mobile cars. Gordon Harris, PhD, is the director of 3D imaging services at MGH. He also directs the hospital’s computer-aided radiology diagnostics lab. Like Dreyer, he is on the radiology faculty at Harvard Medical School, which is affiliated with MGH. Harris says that the 3D lab has patiently and persistently developed “30 or 40 different clinical protocols that we offer.” Those protocols meet the needs of most radiologists and referring clinicians at MGH; in fact, he says, there is some resistance to the rollout of an interactive thin-client application. “We’ve been doing this 3D model for 10 years, and there’s not much push for interactive 3D,” he says. “The radiologists have become accustomed to having it done for them. They’re under so much pressure to read and maximize their efficiency that a lot of them have said they don’t want to do any 3D processing themselves. They want the images provided to them. They don’t want to open up another piece of software that’s going to require them to wait even 10 seconds and then interact with that image.” Dreyer agrees that the success of the 3D lab has diluted desire for a thin client, at least within the radiology department. Nonetheless, there are those 10% of cases that go back for refinement and that could be handled interactively if the specialists or referring clinicians had thin-client capability, he notes. “For the department, there are some gains,” he says, “10% of the time.” It’s the hospital as a whole, however, that could really profit from an interactive thin- client deployment, Dreyer adds. “Outside the department, the enterprise solution seems much more valuable to me,” he says, “because it may be difficult to anticipate the needs of everybody for a particular case. Even though you prepared it for a surgeon, someone else might want to look at that study later, and look at a different part of it, so he or she would instantly be able to render a different view using that thin-client application.” Dreyer and Harris agree that a thin client will be more widely deployed at MGH in the near future; Harris says, “soon—definitely this year.” Cardiovascular Views In fact, Harris says, his lab is now involved in thin-client clinical trials aimed at providing interactive capability for cardiovascular radiologists, both at MGH and (potentially) at other hospitals. Thin-client deployment for cardiovascular studies would reduce duplication of effort between lab technologists and cardiovascular specialists. For some cases, they now use their own 3D workstations to create views in parallel with the lab, especially, Harris says, “creating views of the heart rendered to segment out the atrial appendage from the pulmonary vasculature, so that you see the origin of the coronary arteries to the vertex clearly, and some curved reconstruction of the different coronary arteries.” Cardiac imaging seems to be among the most important reasons for having thin-client capability, Harris says, “for the radiologists to be able to make the diagnosis by adjusting the views, to make sure they are confident in what they are seeing.” Life Images Within the radiology department and across the enterprise, the ability of radiologists and other specialists (especially those working with cardiovascular images) to create their own studies interactively is a compelling reason for MGH to implement a thin client. There are also broader uses for the technology that the hospital is eyeing as it deploys two new entrepreneurial offerings in radiology. One of these ventures is called Life Images, which Dreyer oversees. Life Images, Dreyer says, is a service that will store images for Internet access, or as Dreyer puts it, “storage in the sky.” If a patient moves from one locale to another or changes health care providers, Life Images would allow that patient’s imaging to be collected and stored centrally— eventually, as part of regional, national, or even global electronic medical record networks. “We would be able to transfer the images from our PACS or our thin-client visualization system into this Life Images system, which has its own thin-client visualization that serves the images out to anybody with the appropriate authentication for access,” he says. Dreyer believes that the medical record of the future will be service based, offered to consumers through private companies that provide medical-record storage. Microsoft has a solution called HealthVault to store medical records, Google has a competing service (Google Health), and so do other storage providers. “What’s missing in the link today. from companies like Microsoft and Google,” Dreyer says, “is that they don’t have a way to upload and then visualize image data, so that’s what Life Images is doing. That’s where I really think we’re going to use the thin-client solution first.” Tele3D Advantage The second entrepreneurial radiology venture that’s new for MGH is called Tele3D Advantage. It extends the services offered by the 3D Lab to other hospitals and health care providers through teleradiology. As Harris explains, Tele3D Advantage (tele3dadvantage.net) will set up links with other hospitals to send images to MGH. “We will process them according to the protocols we have developed, through very tight interaction between our staff and the radiologist or referring physician,” he says. Harris adds that the company is still in the pilot phase, although it has three sites signed up and connected to study workflow. “We can provide full-time coverage or night and weekend coverage, or coverage for certain kinds of cases,” he says, “while allowing the technologist to focus on scanning and the radiologist to focus on reading. There’s a lot of time and expense that goes into running a clinically useful 3D service, and with our scale and expertise, we can do it much more efficiently, and with much greater impact.” Harris says that the lab is evaluating the role of 3D thin-client technology as part of the Tele3D service. Again, he singles out cardiology as a likely candidate for thin-client applications. As the lab rolls out its internal 3D thin-client application for cardiovascular studies, it can extend that thin-client application to its teleradiology clients. “The idea would be that we would create the standard views first, and then, through the thin client, specialists could interact with those, make their reading more efficient, and reduce the time they have to spend reading cardiac CT angiography,” Harris says. “The other way we are looking at a thin client is for coordinating 3D cardiac imaging at other hospitals and imaging centers.” The message in all this, Dreyer says, is that hospitals and imaging centers have choices when it comes to how they handle 3D imaging. They can build labs (expensive), they can rely on thin-client systems (less expensive), or they can purchase a service like that offered by Tele3D Advantage. For most smaller providers, Dreyer says, the solution will probably be a combination of purchasing lab services from places like MGH and installing thin-client applications for in-house use. One thing is certain: Thin-client technology has made it possible for many more drivers to get on the 3D imaging highway. For MGH, the multiple options and service models associated with advanced visualization have given the facility a larger motive for getting a 3D thin-client application up and running, despite operating a 3D lab that’s already one of the most sophisticated in the country.
Keith Dreyer, DO, Phd This situation is somewhat analogous to what’s happening to 3D imaging in the radiology department at Massachusetts General Hospital (MGH), Boston. For a decade, MGH has spent large sums of money creating a world-class 3D imaging lab. Technologists have been rigorously trained to provide radiologists of various subspecialties with intricate, highly augmented 3D studies. According to Keith Dreyer, DO, PhD, vice chair of radiology and informatics at MGH, of the 750,000 radiological exams done by the hospital each year, more than 50,000 now go through 3D postprocessing in the lab. To create these studies, technologists sit at specially designed 3D workstations where they follow highly evolved protocols to prepare the images for interpretation. The huge advantage of the 3D lab model is that radiologists are being fed image sets already prepared for them. The radiologists can move quickly and efficiently from one case to the next without having to create views themselves. The difficulty with the lab model, however, is exactly that—radiologists can’t interact with the studies to create their own views. Like passengers on the bus, they are dependent on the technologist drivers. If they want a special view, they have to request it. The bus has to turn to let them off at that stop. Dreyer estimates that 90% of the studies rendered through the MGH 3D lab require no further work. It’s the remaining 10% that have to go back to the technologists for refinement. This wouldn’t be necessary (or, at least, it would be less necessary) if the radiologists could reformat the images themselves. Fat and Thin The 3D workstations where the technologists work run as fat clients. They contain the software necessary to create the 3D studies, and anyone who wants to create such a study has to go to the fat-client workstation and create the study there. Now, new thin-client technology has evolved, enabling the end user—the radiologist or referring physician—to create 3D studies at a conventional radiology workstation, or even at a desktop PC. For many hospitals, this thin-client technology has been a godsend; smaller institutions can have their 3D renderings without gigantic outlays for a postprocessing lab, Dreyer notes. Because MGH has its lab already in place, however, and because it is such a large institution producing so many exams, the adoption of thin-client technology has been troublesome. One of the problems has been scalability. Thin-client technology involves sending image data from the modality to a server (most often, from modality to PACS to a series of servers) that contains software to allow the 3D postprocessing of that image data. This centralized computer server acts as a master workstation where the image data can be manipulated prior to interpretation. The centralized server uses Web-based technology to stream the 3D data out to other workstations or PCs, where radiologists or clinicians can manipulate the 3D images to suit their specific needs. With a thin client, the passengers on the bus become drivers of their own cars. If they need to remove bone electronically to study vascular structure, they don’t need to wait for a technologist. Radiologists using a thin client can interact with the original 3D studies (unlike the set views they receive from the lab), zeroing in on what they want to see. As Dreyer notes, one of the problems with interactive thin clients is that the vendors have not yet solved the problem of allowing simultaneous access to studies by a large number of users. When an institution needs concurrent access for only a dozen or so end users, the thin-client servers work well. Dreyer says, however, “I’ve got 5,000 physicians who might all want to look at the same time—and realistically, I probably have 100 who might all want to look at the same time, so it’s really difficult for me. That’s why I’ve been working with vendors for years, so I can scale. It’s a real challenge, right now, to scale these things up to a big size, so that’s why we’ve been at this for a while.” Enter Visage One company that has built its advanced visualization platform from the ground up on thin-client technology is Massachusetts-based Visage Imaging, a subsidiary of the Australia-based informatics organization Pro Medicus Ltd. Early on, Dreyer was named head of the medical advisory board at Visage. He says that he quickly began to stress the importance of scalability. “It was one of my self-serving goals to say, ‘You’ve got to make this thing scalable,’” he explains. “The company was, by far, the most advanced in having that ability to scale up there. That’s why we went with it, even though it was new to the market,” he says. For two years, Dreyer says, MGH worked with Visage in clinical trials to achieve functionality that would meet the hospital’s needs. “It takes a long time to make these applications. It takes a long time to make a fantastic vessel package, or a cardiac package, or a stroke package. These are clinical applications we’ve been chipping away at for several years,” Dreyer says. Still, unraveling the software knots is taking time. “Let me tell you where I’d like to be,” he says. “I would like to have all CT and MRI available for thin-client visualization. That’s our goal.” Lab Value Ironically, one of the impediments to a thin-client rollout at MGH has been the success of the 3D imaging lab itself. The bus has momentum that keeps the passengers from looking too hard in the direction of the more mobile cars. Gordon Harris, PhD, is the director of 3D imaging services at MGH. He also directs the hospital’s computer-aided radiology diagnostics lab. Like Dreyer, he is on the radiology faculty at Harvard Medical School, which is affiliated with MGH. Harris says that the 3D lab has patiently and persistently developed “30 or 40 different clinical protocols that we offer.” Those protocols meet the needs of most radiologists and referring clinicians at MGH; in fact, he says, there is some resistance to the rollout of an interactive thin-client application. “We’ve been doing this 3D model for 10 years, and there’s not much push for interactive 3D,” he says. “The radiologists have become accustomed to having it done for them. They’re under so much pressure to read and maximize their efficiency that a lot of them have said they don’t want to do any 3D processing themselves. They want the images provided to them. They don’t want to open up another piece of software that’s going to require them to wait even 10 seconds and then interact with that image.” Dreyer agrees that the success of the 3D lab has diluted desire for a thin client, at least within the radiology department. Nonetheless, there are those 10% of cases that go back for refinement and that could be handled interactively if the specialists or referring clinicians had thin-client capability, he notes. “For the department, there are some gains,” he says, “10% of the time.” It’s the hospital as a whole, however, that could really profit from an interactive thin- client deployment, Dreyer adds. “Outside the department, the enterprise solution seems much more valuable to me,” he says, “because it may be difficult to anticipate the needs of everybody for a particular case. Even though you prepared it for a surgeon, someone else might want to look at that study later, and look at a different part of it, so he or she would instantly be able to render a different view using that thin-client application.” Dreyer and Harris agree that a thin client will be more widely deployed at MGH in the near future; Harris says, “soon—definitely this year.” Cardiovascular Views In fact, Harris says, his lab is now involved in thin-client clinical trials aimed at providing interactive capability for cardiovascular radiologists, both at MGH and (potentially) at other hospitals. Thin-client deployment for cardiovascular studies would reduce duplication of effort between lab technologists and cardiovascular specialists. For some cases, they now use their own 3D workstations to create views in parallel with the lab, especially, Harris says, “creating views of the heart rendered to segment out the atrial appendage from the pulmonary vasculature, so that you see the origin of the coronary arteries to the vertex clearly, and some curved reconstruction of the different coronary arteries.” Cardiac imaging seems to be among the most important reasons for having thin-client capability, Harris says, “for the radiologists to be able to make the diagnosis by adjusting the views, to make sure they are confident in what they are seeing.” Life Images Within the radiology department and across the enterprise, the ability of radiologists and other specialists (especially those working with cardiovascular images) to create their own studies interactively is a compelling reason for MGH to implement a thin client. There are also broader uses for the technology that the hospital is eyeing as it deploys two new entrepreneurial offerings in radiology. One of these ventures is called Life Images, which Dreyer oversees. Life Images, Dreyer says, is a service that will store images for Internet access, or as Dreyer puts it, “storage in the sky.” If a patient moves from one locale to another or changes health care providers, Life Images would allow that patient’s imaging to be collected and stored centrally— eventually, as part of regional, national, or even global electronic medical record networks. “We would be able to transfer the images from our PACS or our thin-client visualization system into this Life Images system, which has its own thin-client visualization that serves the images out to anybody with the appropriate authentication for access,” he says. Dreyer believes that the medical record of the future will be service based, offered to consumers through private companies that provide medical-record storage. Microsoft has a solution called HealthVault to store medical records, Google has a competing service (Google Health), and so do other storage providers. “What’s missing in the link today. from companies like Microsoft and Google,” Dreyer says, “is that they don’t have a way to upload and then visualize image data, so that’s what Life Images is doing. That’s where I really think we’re going to use the thin-client solution first.” Tele3D Advantage The second entrepreneurial radiology venture that’s new for MGH is called Tele3D Advantage. It extends the services offered by the 3D Lab to other hospitals and health care providers through teleradiology. As Harris explains, Tele3D Advantage (tele3dadvantage.net) will set up links with other hospitals to send images to MGH. “We will process them according to the protocols we have developed, through very tight interaction between our staff and the radiologist or referring physician,” he says. Harris adds that the company is still in the pilot phase, although it has three sites signed up and connected to study workflow. “We can provide full-time coverage or night and weekend coverage, or coverage for certain kinds of cases,” he says, “while allowing the technologist to focus on scanning and the radiologist to focus on reading. There’s a lot of time and expense that goes into running a clinically useful 3D service, and with our scale and expertise, we can do it much more efficiently, and with much greater impact.” Harris says that the lab is evaluating the role of 3D thin-client technology as part of the Tele3D service. Again, he singles out cardiology as a likely candidate for thin-client applications. As the lab rolls out its internal 3D thin-client application for cardiovascular studies, it can extend that thin-client application to its teleradiology clients. “The idea would be that we would create the standard views first, and then, through the thin client, specialists could interact with those, make their reading more efficient, and reduce the time they have to spend reading cardiac CT angiography,” Harris says. “The other way we are looking at a thin client is for coordinating 3D cardiac imaging at other hospitals and imaging centers.” The message in all this, Dreyer says, is that hospitals and imaging centers have choices when it comes to how they handle 3D imaging. They can build labs (expensive), they can rely on thin-client systems (less expensive), or they can purchase a service like that offered by Tele3D Advantage. For most smaller providers, Dreyer says, the solution will probably be a combination of purchasing lab services from places like MGH and installing thin-client applications for in-house use. One thing is certain: Thin-client technology has made it possible for many more drivers to get on the 3D imaging highway. For MGH, the multiple options and service models associated with advanced visualization have given the facility a larger motive for getting a 3D thin-client application up and running, despite operating a 3D lab that’s already one of the most sophisticated in the country.