|
|||||||||||||||
|
Home
|
It Came From Outer
Space In college, Matthew J. Kuhn, MD, took several astronomy courses, where he learned about how astronomers detect new objects in the night sky by taking photographs and observing changes over time. In the mid-1980s, Kuhn was a neuroradiology fellow at Massachusetts General Hospital in Boston. The hospital was updating the way its radiologists read films. Management made a file room person responsible for hanging films so the radiologists only had to come there to read them. “It seemed like it would be more efficient because you didn’t have to pick up the film, find its orientation, put it in the right place, and find the old films for comparison,” Kuhn recalls. However, an older radiologist objected to the procedure change. “He said he didn’t like it because he made most of his findings just picking up the x-ray to hang it,” Kuhn says. He realized the older radiologist had a point. “There is something about motion that it draws our attention to abnormalities,” Kuhn says. He kept that idea in the back of his mind. Twenty years later, Kuhn is chairman of the radiology department at St. John’s Hospital in Springfield, Ill., and he is working to perfect a technique similar to that used in astronomy, which could make it easier for radiologists to detect new lesions, as well as size changes of known lesions, on imaging studies. Kuhn, also a professor of radiology, neurology, and neurosurgery at Southern Illinois School of Medicine, calls the technology BlinkRadiology because every 0.3 seconds it blinks, or alternates, between a prior and new image of the same area. As a result, the differences “blink” at the observer. Astronomy Ties Other disciplines have used the technology as well. Geologists look at satellite photos of land forms to detect changes of coastlines, and governments use aerial views to detect surreptitious development. The technology did not really have a place in radiology as long as images were analog because it was considered too cumbersome. “It couldn’t really easily be applied in a time-efficient manner to radiologic images,” Kuhn says. Now that images can be digitized, the technology becomes more practical. However, manufacturers have not developed the software necessary to superimpose and blink diagnostic images. While Kuhn is hopeful someone will work with him to commercially release the software, he has devised his own homemade solution: He downloads patient images into Microsoft PowerPoint and is able to blink them that way. It is possible to toggle images manually, but it is more effective when the images blink automatically, he says. Kuhn is convinced that if blink technology catches on, the method will significantly improve the accuracy of radiologists’ reports. The blink method is not only useful for looking for new lesions or abnormalities, Kuhn says, but it also helps detect the growth of lesions. “You can see a lesion get bigger or smaller very easily this way,” he says. “It’s almost as if you are looking under a microscope and watching the lesion grow before your eyes.” Side by Side Kuhn explains that BlinkRadiology takes advantage of the human brain’s instinctive attraction to motion. An area of the retina known as the receptive field is specifically designed for motion detection, “yet it typically goes unused when looking at medical imaging,” he says. “This is like your mother always told you: ‘Use more of your brain.’” Examples of how it works can be seen on Kuhn’s Web site at www.blinkradiology.com. The method also has “a Zen kind of thing to it,” Kuhn says. Typically, he says, radiologists find lesions, but with the blink method, “it’s as if the lesions are finding us. You can’t miss [them]. That’s why I say it has an Eastern philosophical bent to it.” Blink technology also helps readers overcome their biases when dictating reports. Studies of radiologists have proven that readers have a tendency to stop looking once they think they have found what they are looking for. “They are not always looking for the unexpected,” Kuhn says. With blink technology, the expected and unexpected jump out, making it less likely the reader will miss something. The technology works best when the images are taken of the same area from the same angle. “If obtaining the images [is] done with the same protocol each time, then blinking is simple,” Kuhn says. He says the human eye can easily accommodate for small differences. “Even if one chest image is an inspiration and one is expiration, for example, our eyes are easily able to accommodate for that.” It doesn’t have to be just two images either, he says. “It could be old, newer, and newest.” The time between images is irrelevant as well. “We’ve seen new lesions that have occurred when we compared images from years and years ago,” Kuhn says. CT, MR, Nuc Med Work Best Ultrasound may be the one exception. “Ultrasound is probably the worst because of the differences in the angles in the plane. It’s the most inexact for blinking, and I haven’t done that much with it because the [images] are so dependent on the person holding the transducer,” Kuhn says. Most of his work to date involves comparing older and newer studies, but Kuhn is trying to find a way to get motion involved if only one image is available. “That speaks to what the older radiologist at Mass General was saying about making a diagnosis even before the film was embedded in the light box. What if there was a way to slightly offset the image by a degree or two? Could we use the computer to make it sort of wiggle? I have no idea if that idea will work, but that’s where I’m heading right now,” he says. Kuhn says that while radiologists are meticulous and methodical when it comes to looking at images, they make mistakes. Studies show that most errors occur in the bottom right of the image. Kuhn says that could be because radiologists are trained to read from left to right, starting at the top. By the time they get to the bottom, they are not as cautious. This technology would make the differences stand out no matter where they are on the images, he says. Comparative Study In the study, two radiologists were presented with 80 pairs of images containing 246 abnormalities. Each pair of images was either plain CR, CT, or MRI images obtained with the same patient at different times. Each observer was asked to note any abnormalities and changes between the images using conventional series and stack formats typical of contemporary PACS interpretation styles. Then, they were asked to view the pairs using the digital blink comparator technique. The two radiologists detected 185 and 197 abnormalities, respectively, using conventional PACS interpretation techniques. Using the digital blink technique, they detected 223 and 232 abnormalities, respectively. The blink technology resulted in an improved detection rate of 20.5% and 17.8%, respectively. In late June, Kuhn will attend the Computed Assisted Radiology and Surgery meeting in Berlin, Germany, where he will report on a second study specific to MRI and CT exams that found the use of digital blink comparator results improved a radiologist’s ability to detect new lesions and/or changes when comparing new and prior studies. The study involved two radiologists who detected 187 and 180 abnormalities, respectively, using conventional PACS interpretation techniques and 210 and 204 abnormalities, respectively, using digital blink technology. They reported an improved detection rate of 12.3% and 13.3%, respectively. David Channin, MD, chief of imaging informatics at Northwestern University’s Feinberg School of Medicine department of radiology in Chicago, agrees that BlinkRadiology shows promise. He believes it is an interesting use of an idea from another domain. “I think it needs further evaluation to document exactly how valuable it is, but it is certainly something that is worth pursuing further,” he says. Channin is working on a project similar to what Kuhn is doing. It would allow readers to flicker through a stack of images, he says. Channin says he is optimistic about the results. “People should start using it,” he says, “because, with a little more evaluation, we can see if it has the potential to be a radical change in the practice of radiology.” Kuhn says most manufacturers are less enthused about the idea because implementing it would require making changes in software code. “Working with large companies, it’s difficult to get them to budge until things are scientifically proven,” Kuhn says. However, if blink technology continues to prove beneficial, he hopes that some manufacturer will want to be the first with a workstation that makes it simple to use. “I’m not implying,” Kuhn says, “that blinking new and old images will solve all reading problems, but it might go a certain distance toward helping reduce misses.” — Beth W. Orenstein is a regular contributor to Radiology Today. She writes from her home in Northampton, Pa. For an interactive demonstration of how BlinkRadiology works, visit www.BlinkRadiology.com. Then post your thoughts at www.RadiologyToday.net/forum.
|
|
3801 Schuylkill Rd • Spring City, PA 19475 Publishers of Radiology Today All rights reserved. |