| |||||||||||||
|
Home
|
For other articles and previous issues click here. August 15, 2005 X-ray
Assist — Combining CAD With Routine Chest Exams Can Identify
Lung Cancer Sooner Lung cancer is the most common cause of cancer death in the United States. According to the American Cancer Society, 160,000 people die each year from the disease. Diagnosis is devastating, as it is among the most difficult of all cancers to treat. Currently, only 15% of patients survive five years after diagnosis. The survival rate is improved if the cancer can be detected early, when it is still surgically treatable. That has posed a major obstacle to clinicians, as small, early-stage nodules are difficult to identify. However, a new technology that works with common chest radiographs can increase radiologists’ ability to detect lung cancer at its earliest stages, leading to earlier treatment. The RapidScreen system applies a computer-aided detection (CAD) algorithm to a common chest x-ray to identify possible cancer nodule sites for the radiologist to study further. The basic idea is that routine chest x-rays, which are taken for many reasons, can be quickly evaluated with RapidScreen, providing additional opportunity to catch nodules early with little disruption to radiologist workflow. Experts estimate that early detection and subsequent treatment could save as many as 54,000 lives per year. “Statistics indicate that if you find the early stage 1 cancer, you could increase the five-year survival rate to 49%,” says Sam Finkelstein, president of Riverain Medical, LLC, the Miamisburg, Ohio-based company that markets RapidScreen. “Today, because clinicians only find about 15% of stage 1 cancers, the mortality rate is extremely high.” Deus Technologies, located in Rockville, Md., developed the technology offered by Riverain Medical, a developer of CAD systems, imaging software solutions, and information services. The RapidScreen RS-2000 system has been shown in clinical studies to improve radiologists’ detection of nodules in the 9- to 14-millimeter range by as much as 21%. “Early-stage cancer, by definition, is nodules from 9 to 30 millimeters,” says Finkelstein. Computerized Collaboration As the name suggests, CAD, the so-called “second set of eyes,” assists radiologists in identifying abnormalities. Computer algorithms analyze images for signs of specific diseases. It is a collaborative process between the computer and the radiologist. Most CAD systems are designed to be a second reader, complementing radiologists’ initial studies by marking areas for further analysis. In the case of RapidScreen, CAD aids radiologists in detecting solitary pulmonary nodules, potentially lung cancer at its earliest stages. “The major benefit is seen with the nodules that are smaller because those are the ones radiologists have more trouble seeing,” says Matthew Freedman, MD, associate professor of oncology at Georgetown University Medical Center in Washington, D.C. Freedman serves as senior medical advisor for Riverain Medical and was clinical director for the trials submitted to the FDA for premarket approval of RS-2000. According to Riverain Medical, RapidScreen employs an automatic multistep image analysis processing system that involves a series of patented algorithms and proprietary classification technologies to identify regions that may indicate cancer. “With CAD, we are not manipulating or enhancing an image,” explains Finkelstein. “Rather, we are taking the data from that image and we are evaluating and analyzing it with the algorithm to mark regions of interest that aid in detection.” He adds that it took 10 years to develop the specific algorithm that includes 87 unique signatures of cancerous abnormalities. Digital and Analog Configurations In less than one minute, the digitized image appears on the RS-2000’s monitor. The system then generates a computer printout of the image that can be included with the original x-ray film. Physicians can compare the original film and the RS-2000 image and focus on the regions the system has circled—regions that may not have been identified during the initial review. The RS-Digital configuration is a fully DICOM (digital imaging and communications in medicine) -compatible platform that allows connectivity to existing workstations. It utilizes the productivity and connectivity features of a digital department and provides rapid CAD results. It also processes dual energy subtraction radiographic images. A DICOM-formatted chest image is sent through the RS-Digital. Processing is done within the system so department workflow isn’t interrupted. Following processing, CAD markings are either “burned” in to generate a derived image or become an integrated part of the original DICOM image. All image data that goes into the network is transferred to the picture archiving and communication system (PACS) server. The data becomes associated with the original radiographic study and are immediately available for retrieval and viewing on a PACS workstation. Essentially, this is how it works: an x-ray machine on the DICOM network automatically sends an image to the RS-Digital, which then processes the image for nodule detection and produces and stores output objects with marked regions of interest. The RS-Digital system can receive images from up to five origin devices. It can be upgraded to enable additional origin devices. Output objects are then automatically sent to destination devices on the DICOM network. Up to 25 destination devices can be defined. The RS-Digital sends a storage commitment request to the destination devices. As Finkelstein says, the RS-Digital features an algorithm, based on neural networks and fuzzy logic, designed to identify and classify 87 features associated with solitary pulmonary nodules. The system searches a chest x-ray for structures smaller than 30 millimeters that may indicate lung nodules. When such a structure is identified, the system places a marker over it. “The algorithm automatically takes the data from the film or the digital image, runs it through the digital scanner, and, based on the signatures, identifies nodules and gives the weighted average,” explains Finkelstein. “The image isn’t viewed as an image but as data. That is where the unique signatures come in. Information is added, in terms of the areas of markings, so that radiologists take a closer look. It is not something like an image magnification. Rather, it is actually using the data within the image itself.” Clinical Trials Tests have demonstrated that RS-Digital required less than 30 seconds for the complete processing cycle, from when an image was sent by an origin device until the results were available for display on a workstation. According to Freedman, after the initial detection with RapidScreen, technologies such as computed tomography may be used to further characterize and confirm nodules. Both Finkelstein and Freedman indicate that RapidScreen technically is not a screening tool. Rather, it is better defined as a case-finding tool. “It identifies nodules that appear suspicious, therefore it is a case-finding detection tool,” clarifies Finkelstein. Radiologists, he explains, would use RapidScreen in their normal workflow. They wouldn’t necessarily be looking for cancer but for abnormalities within the chest radiograph. “In case finding, the machine help makes it more likely you’ll find the cancer,” explains Freedman. “But that doesn’t mean that it was something you couldn’t see. It is finding something that you may have seen if you looked at the film on another day, or that one of your colleagues might have found.” Ease of Use It is also cost effective. “From a practical point of view, you don’t have to buy any new equipment,” says Finkelstein. “RapidScreen is easily introduced into existing radiography equipment, whether it’s film-based or digital.” In 2004, Garden City Hospital in Michigan became the first facility in the United States to offer the technology. The hospital installed the analog RS-2000 configuration. Director of radiology Jim Williamson, MSRT, ARRT, described the system as easy to install and “incredibly simple” to learn. He also indicated that after an initial period of reluctance, physicians became more enamored of the new technology. “At first, the doctors weren’t very accepting, but as they became more accustomed to it, they became more aware of the capabilities it provided,” he recalls. “Now they realize that it’s more of a help than any kind of hindrance.” Though the hospital doesn’t strictly define RS-2000 as a screening service, it still markets the technology as a proactive measure. Its physicians recommend the service to patients who: • smoke tobacco; • are exposed to secondhand smoke; • experience frequent exposure to asbestos; • have lived in a radon-contaminated house; • work with cancer-causing agents such as radioactive material, chemicals such as arsenic and coal products, fuels such as gasoline, or diesel exhaust; • have recurring inflammation of the lungs due to tuberculosis or pneumonia; • have a personal or family history of lung cancer; • eat a diet low in fruits and vegetables; or • live in an area with heavy air pollution. Garden City Hospital is located in an industrialized area, and physicians have been gratified by a positive response from the community, says Williamson. As far as reimbursement, a Category III CPT code (0152T) was recently assigned to CAD for lesion detection, including lung cancer. Consequently, the code may now be submitted to third-party payors for reimbursement for using CAD to review chest radiographs. Moving Forward “Radiologists are always having new technology thrust upon them,” he says. “So, for them, the challenge is how to use that technology in ways that doesn’t interrupt their workflow. One of the reasons we are excited about this product is because it isn’t just about finding the early lung cancers; it is about assisting the radiologists in their workflow, and giving them the tool to improve the diagnosis, which they are already excellent in doing.” Looking to the future, Freedman sees CAD as something that is going to be applied to more and more areas. “As people figure out how to write the algorithms, every part of medical care is going to involve intelligent systems to help detect disease and to make decisions. Doing this with chest radiographs is just part of the picture,” he comments. Further, Freedman believes CAD will ultimately be an inherent part of every imaging modality. “I think it could happen faster than people think it will, but it will definitely occur.” — Dan Harvey is a freelance writer based in Wilmington, Del. |
 |
3801 Schuylkill Rd • Spring City, PA 19475 Publishers of For the Record All rights reserved. |