Radiologists can reduce ABUS interpretation time by using concurrent-read CAD system
Use of a concurrent-read computer-aided detection (CAD) system when interpreting automated breast ultrasound (ABUS) examinations helps improve efficiency without negatively affecting accuracy, according to a new study published by the American Journal of Roentgenology.
The authors tasked 18 radiologists with interpreting 185 ABUS studies twice, once with the FDA-approved CAD system and then again, four weeks later, without the CAD system. Overall, with the CAD system, the radiologists’ area under the ROC curve (AUC) was 0.848 and the mean interpretation time was 2 minutes and 24 seconds. Without the CAD system, the AUC was 0.828 and the mean interpretation time was 3 minutes and 33 seconds.
These numbers represent a “statistically noninferior” difference in AUC, the authors noted, but the shift in interpretation time is noteworthy.
“The faster interpretation of ABUS studies with the CAD system saves slightly over one minute per case on average,” wrote author Yulei Jiang, PhD, department of radiology at the University of Chicago, and colleagues. “Almost all readers will be faster with the CAD system, and some will save more time than others. To take advantage of this time saving, it is important that diagnostic performance in cancer detection not be compromised.”
Jiang and colleagues added that no significant difference is expected in the sensitivity or specificity of radiologists, but “some readers may experience individually significant improvement in specificity with the CAD system.”
The authors also explained one key limitation of their research: the ABUS studies were acquired with older ABUS equipment.
“A new generation of ABUS equipment has replaced the flat transducer with a curved transducer,” they wrote. “Therefore, the performance of the CAD system in interpretation of ABUS studies acquired with the newer generation of equipment must be evaluated. If the computer performance of the CAD system in new ABUS cases remains similar to that in this study, then the conclusion of this study may also apply to new ABUS studies because interpretation of new ABUS studies in the conventional manner remains challenging.”
Michael has more than 18 years of experience as a professional writer and editor. He has written at length about cardiology, radiology, artificial intelligence and other key healthcare topics.