CAD: Helping radiologists make a more accurate, reliable, rapid diagnosis
Computer-aided-detection (CAD) systems for chest imaging are emerging and will play a pivotal role in the future of radiology, discussed Matthew Brown, PhD, assistant professor, department of radiology, UCLA’s David Geffen School of Medicine, at yesterday’s SCAR University session that focused on image processing, 3D and CAD. The session is part of the annual meeting of the recently renamed Society for Imaging Informatics in Medicine taking place this week in Austin, Texas.
“Lung nodule detection is currently the primary application of computer aided detection systems,” Brown said. The goal of these CAD systems is automated detection of lung nodules in x-ray and CT imaging, acting as a “second opinion” for radiologists.
“The motivation for thinking about screening for nodule detection is for the early diagnosis of lung cancer,” Brown said. “Another motivation [for use of the technology] is just the sheer number of images that we are required to deal with.”
It is well known throughout the imaging community that today’s scanners – compared with older generation systems – produce more images with much thinner slices. CAD advocates believe that the overwhelming number of images being generated by multidetector CT scanners and results showing improved radiologist sensitivity for nodules suggest that CAD will some day become part of daily radiologist workflow.
The advantage of CAD for chest imaging is that the intelligent software automatically identifies and marks the location of lung nodules for review by a radiologist. According to Brown, when using CAD to provide a second opinion, radiologists experience an increase in sensitivity for small nodules, and can mark false positives. Identification of false-positives typically leads to unnecessary second-looks by radiologists.
Commercially available CAD systems for chest x-ray images can detect nodules as small as 5mm in diameter but exhibit their best performance for nodules around the 10 – 30mm range, Brown said. Commercial CT-based systems typically report nodules around 4mm in diameter, he explained.
Recent CAD analysis tools that can help radiologists with this ‘second opinion’ include quantitative assessment tools to measure nodule shape size, attenuation statistics and 3D visualization capabilities for CT data.
Brown also discussed how commercial CAD systems can receive DICOM image data and typically output their results in DICOM objects. At the same time, these objects are usually not readable by standard PACS workstations, so a dedicated CAD workstation is required for viewing.
The future entails a fully integrated workstation with CAD capabilities within the PACS workflow “I think that people are working hard to make that happen,” Brown said.
“This is a very hot topic both in research and commercially,” Brown continued. “There are new applications that are improving CAD performance in terms of sensitivity, specificity and speed.”
“Lung nodule detection is currently the primary application of computer aided detection systems,” Brown said. The goal of these CAD systems is automated detection of lung nodules in x-ray and CT imaging, acting as a “second opinion” for radiologists.
“The motivation for thinking about screening for nodule detection is for the early diagnosis of lung cancer,” Brown said. “Another motivation [for use of the technology] is just the sheer number of images that we are required to deal with.”
It is well known throughout the imaging community that today’s scanners – compared with older generation systems – produce more images with much thinner slices. CAD advocates believe that the overwhelming number of images being generated by multidetector CT scanners and results showing improved radiologist sensitivity for nodules suggest that CAD will some day become part of daily radiologist workflow.
The advantage of CAD for chest imaging is that the intelligent software automatically identifies and marks the location of lung nodules for review by a radiologist. According to Brown, when using CAD to provide a second opinion, radiologists experience an increase in sensitivity for small nodules, and can mark false positives. Identification of false-positives typically leads to unnecessary second-looks by radiologists.
Commercially available CAD systems for chest x-ray images can detect nodules as small as 5mm in diameter but exhibit their best performance for nodules around the 10 – 30mm range, Brown said. Commercial CT-based systems typically report nodules around 4mm in diameter, he explained.
Recent CAD analysis tools that can help radiologists with this ‘second opinion’ include quantitative assessment tools to measure nodule shape size, attenuation statistics and 3D visualization capabilities for CT data.
Brown also discussed how commercial CAD systems can receive DICOM image data and typically output their results in DICOM objects. At the same time, these objects are usually not readable by standard PACS workstations, so a dedicated CAD workstation is required for viewing.
The future entails a fully integrated workstation with CAD capabilities within the PACS workflow “I think that people are working hard to make that happen,” Brown said.
“This is a very hot topic both in research and commercially,” Brown continued. “There are new applications that are improving CAD performance in terms of sensitivity, specificity and speed.”