Computer-aided detection, often used for nodule detection, is becoming more integrated with PACS, a trend that streamlines workflow. Integration of the two systems is gaining more widespread use as facilities try to maximize the potential that coordinating the processes can achieve, which will let radiologists refer to CAD images during interpretation.
Lack of integration between PACS and CAD has been the biggest barrier to the use of CAD technology, according to Dr. Heber M. MacMahon, director of thoracic imaging at the University of Chicago Medical Center in the U.S., a hub of CAD development.
“[CAD technologies] have not been tightly integrated into PACs, so it takes a lot of time to evaluate,” MacMahon said. “If it takes more than a few minutes to bring up the interface and extract results, people won't use it in their practice.” CAD and PACS developers have worked together to improve the software so that nodule detection can be done more quickly and reliably, he said.
Japanese researchers at Kyushu University hospital have integrated a server used for moving temporal-subtraction and nodule-detection images into PACS, allowing current and prior CAD images to be automatically loaded on display terminals, according to a report outlining their four-year experiment (J Digit Imaging 2008;21[1]:91-98).
The image-loading technique can easily be applied to other CAD systems, such as those for mammography, gastrointestinal exams, and CT, said Dr. Shuji Sakai, an associate professor of radiology at the university's school of medicine.
The report, however, did mention several problems associated with the integration process, including excessive false-positives, difficulties with image orientation, and low image quality when temporal subtraction was performed on images obtained on different devices.
The integrated systems allow all data to be transferred from one PACS to another, all of which are connected to a CAD manager.
The advantage of the integrated systems is that no interface is required between the systems, allowing data from CT chest images to be sent automatically to the CAD manager, which starts the necessary utility process. The CAD manager then determines where the data are coming from and sends the results to the correct PACS.
“It works in the background. It's just a box somewhere in the closet,” Herzog said. “Nobody has to work with it.”
When Herzog opens a lung study in PACS to see the original images, the integrated system displays a second image with CAD markers on it, showing every potential pulmonary nodule with a circle around it.
While Herzog considers his CAD systems to be very robust, he would like to see future systems do more than just detect nodules.
“Perhaps they can do measurements on nodules, especially based on CT data,” he said. “We need diameters, we need volume.”
