In a retrospective study involving 30 plain radiographs and one chest radiologist, a team from Massachusetts General Hospital (MGH) found a high level of concordance between interpretations performed on an iPad 2 using home-grown viewing software and those performed with a conventional PACS workstation.
"Our preliminary evaluation shows that the image quality and accuracy of interpretation was excellent when [our] iPad-based application was used for visualizing chest radiographs," said presenter Dr. Supriya Gupta. "This could potentially serve as an aid for radiologists for making clinical decisions."
The U.S. Food and Drug Administration (FDA) has cleared a number of mobile viewing apps for diagnostic use in the past year, but until the recent clearance of MIM Software's Mobile MIM application for radiography, all have been for use with CT, MRI, and, in one case, PET. Unlike those modalities, radiography's native resolution far exceeds the iPad's display capabilities, making its use more challenging on the tablet computer's relatively small display.
However, the utility of mobile apps for emergency diagnosis in radiology is increasing, and chest radiographs are an important part of forming these diagnoses, Gupta said.
As a result, the MGH group sought to evaluate plain chest films using an internally developed iPad-based image viewer called RadPad. RadPad is a native iOS app that's currently available inside the institutional firewall network at MGH with restricted user-specific download and use, Gupta said. RadPad does not store or cache images on the iPad.
Digital x-ray images originally have a resolution of 4,200 x 3,408 pixels, with a file size of approximately 27 MB. After compression at a 19-1 ratio, the typical image has a compressed size of around 1.5 MB. While RadPad first decompresses an image at a size of 1,050 x 852 so that it can be quickly displayed on the screen, decompression at full resolution continues in the background in a process that takes approximately 4.3 seconds, Gupta said.
"The display performance is better with iOS5," she said.
After initial decompression, users can window/level and pan/zoom the images at full speed. When the images have been fully decompressed, they are automatically put into the display.
For the purposes of their study, the researchers used RadPad's research settings, which remove patient information. In a retrospective evaluation, 30 cases of plain chest radiographs performed to rule out pneumothorax following tube placement in the intensive care unit were randomly selected from the institution's radiology database. Blinded to the final diagnosis, a chest radiologist read the studies initially using RadPad and then on a PACS workstation one week later.
The first 10 cases were used for quality control and user feedback, and the application was then redesigned to incorporate the feedback.
In addition to reading the studies, the chest radiologist graded the remaining 20 cases for image quality, with 5 meaning an excellent diagnostic-quality image, 4 meaning less than diagnostic but good quality, 3 meaning good only for a preliminary report, 2 meaning not adequate for a preliminary opinion, and 1 meaning unacceptable.
Download speed was also rated: 5 meant excellent speed, almost instantly; 4 meant good speed, better than PACS; 3 meant speed equivalent to a PACS workstation; 2 meant speed less than a PACS workstation, but not slow; and 1 meant very slow.
Observer readings were recorded as either normal or having the presence of pneumothorax. Interpretation accuracy on the iPad was then manually compared with the report findings produced on the PACS workstation, and intraobserver variability was assessed using kappa statistics.
For the 20 cases, the observer's findings on the iPad and PACS workstation were highly concordant and differed in only one case (r = 0.8, p < 0.0001, 95% confidence interval: 0.5-0.9), Gupta said. The kappa value was 0.8, with a 0.9 percentage of overall agreement.
Image quality was found to be excellent (5) by the radiologist, while download speed was also considered excellent (ranging from 4 to 5). Average download speed was less than one minute for both the PACS workstation and the iPad, and there was no statistically significant difference in average interpretation time between the PACS (1.2 minutes) and the iPad (1.44 minutes).
More research studies now need to be performed on a larger scale, including some that involve more radiologists, Gupta said. The researchers believe, however, that the iPad and various other high-resolution mobile devices could be used for diagnosis and consulting for thoracic emergencies and trauma, in addition to potentially serving as a triage tool. The devices could also be useful for patients who have been transferred, and for assessing chest tube placements.
Further studies to measure luminance, image quality, and noise in accordance with international standards and guidelines for plain radiographs are also required, Gupta said.
"Radiologists' expectations are evolving in the direction of better [image] quality for rendering rapid reads following FDA approval," Gupta said. "Addressing these concerns can actually help us realize the utilization of mobile imaging."