Radiology: New approaches to managing mega datasets
Before considering technical developments, the authors reviewed the basics of image interpretation, pitfalls for various tasks and potential developments that might address challenges.
Detection and localization of any potential abnormalities is the first task of diagnostic imaging interpretation, stressed Katherine P. Andriole, PhD, from the department of radiology at Brigham and Women’s Hospital in Boston, and colleagues. Errors may occur if the radiologist's gaze is not directed to the target or if the target is not “strong” enough to be recognized as an abnormality. The probability of detecting a target correlates with conspicuity.
“Interventions that boost the signal-to-noise ratio would be expected to improve performance,” wrote Andriole, who cited 3D rendering, advanced assistive technology to flag relevant parts of the image set and CAD as positive advances.
The second task, assessing change over time, may be facilitated by multimodality image fusion, image subtraction, texture analysis and the use of motion and color.
Next, visual characterization of the lesion often requires the radiologist to combine input from multiple types of studies. “Technology could help the radiologist correlate the appearance of the same lesion on multiple studies or provide intelligence through decision aids to assist in differential diagnosis,” offered Andriole and colleagues. Electronic teaching files and clinical decision support also could facilitate this task, the authors suggested.
Finally, radiologists need to communicate results to referring physicians. Developments that could streamline and improve results communication include technology that adds illustrative or annotated images to multimedia reports. Andriole also cited guidelines for structuring content such as the Breast Imaging Reporting and Data System (BI-RAD) lexicon as a useful foundation for improving report quality and clarity. Finally, automatic electronic communication systems may boost patient safety.
Emerging advanced technologies
A host of existing and emerging advanced technologies may aid radiologists on multiple fronts, including data management, workflow, navigation, rendering, distribution and image management, explained the authors.
Andriole offered several options to help radiologists obtain a gestalt of the massive datasets. Stereoscopic displays have been shown to produce greater true-positive and fewer false-positive mammography results, and devices to track and adjust the reader’s gaze produce a visual effect similar to stereoscopic viewing, she noted.
Other developments including volumetric rendering, 3D volumetric rendering and pseudo 3D volumetric rendering offer improved paradigms, while 4D and 5D viewing add the elements of time and functional measures. Multimodality fusion, specifically across time points, may help viewers detect changes across serial exams, while holographic rendering enables intuitive interpretation, offered Andriole et al.
The authors listed 3D software as a positive development “that makes it possible to perform advanced postprocessing more easily.” However, they emphasized that these applications must be embedded in the PACS workstation to provide comprehensive utility, accessibility and functionality.
Additional infrastructure should enhance the human-machine user interface, noted the authors, who offered promising developments from other fields, particularly gaming. These include the 3D mouse, joystick, Wii controller, touch-sensitive screens and voice-activated workstations.
Finally, Andriole and colleagues noted that handheld devices and faster cellular or wireless networks could facilitate distribution of images and reports to referring physicians. Developments such as the iPhone, software to push images to handhelds and tools to display information from multiple sources can improve image access.
“Particularly, with the advent of the next generation of web-based systems, service-oriented architecture and cloud computing, referring physicians and radiologists can have secure easy access to both image-based and non-image based patient information in synchronous sessions on compact, portable display devices,” wrote Andriole.
The authors stressed the need for ongoing developments to plug remaining gaps in image management. For example, a tool to integrate the array of nonimaging clinical information specific to the patient or disease could inform the image interpretation process.
“We are approaching a revolution in the role of the imaging specialist in direct patient care. To achieve this, current practice must take full advantage of the technologies of today. This is radiology’s challenge and opportunity,” concluded Andriole.