Concurrent read CAD software cuts breast image study time
Using a computer-aided detection system to read automated breast ultrasounds can cut the time a radiologist needs to interpret images.
Screening mammography is an important step in the early detection of breast cancer. However, it’s harder to read mammography screens in women with dense breast tissue. In 2012 the Food and Drug Administration approved the first ABUS system to use in combination with screening mammographies. ABUS systems automatically scan the breasts in six volumetric data sets, thus eliminating operator dependence on handheld breast ultrasound equipment.
Using both ABUS and mammography improves a radiologist’s ability to correctly read the images and detect cancer, but it’s challenging to interpret automated breast ultrasounds (ABUS) images because all six datasets need to be reviewed. There are also image artifacts, such as dead pixels or ghosting, that can degrade the image’s quality or hinder diagnosis. As a result, it takes 2.9 to 9.5 minutes to interpret an ABUS study.
CAD systems have been used to help radiologists interpret images, but many of them are used after the radiologist has conducted his or her initial review, which may help in diagnostics but not reduce the interpretation time.
The researchers, led by Yulei Jiang, from the Radiology Department at the University of Chicago, posited that using a current read CAD at the same time as a radiologist interpreted the initial ABUS images might cut down on review time.
They used a new CAD system (QVCAD) that concurrently reads the ABUS raw data using proprietary algorithms and highlights potentially clinically significant findings superimposed on the ABUS image. It is intended for use during the initial interpretation of ABUS studies, not after the radiologist has conducted the initial review.
In the study, 18 Mammography Quality Standards Act-certified radiologists from nine different states interpreted a cancer enriched set of 185 screening ABUS studies, 52 of which included cancer and 133 that did not. Each reader interpreted each case twice, once without using the CAD system and once with it. Some readers read the raw ABUS images first; others read the CAD-aided ones first.
Using the CAD system “significantly” reduced the time it took to interpret the ABUS images. The mean interpretation time per case was three minutes 33 seconds without the CAD system and two minutes 24 seconds with it, a reduction in interpretation time of 67 percent of the time without the CAD system, or just over a minute saved.
Importantly, there was no significant difference in sensitivity or specificity between interpretations with and without CAD use. The number of cases in which the radiologists correctly identified the presence of cancer was similar. There was also a clear trend toward fewer false positive findings when the readers read ABUS studies with the CAD system.
The study appears in the August issue of the American Journal of Roentgenology.
“The interpretation time saving is possible because it is a concurrent read; the time saving is not possible in a second-read scenario. Interpretation of ABUS studies is complex and time-consuming. The CAD system evaluated in this study helps streamline and reduce the complexity of this interpretation task. The demand on a CAD system to not lead to a loss in diagnostic accuracy, especially in sensitivity, is greater for a CAD system intended for concurrent read than for second read. This study showed that the QVCAD system meets this demand and can be used safely to reduce interpretation time and improve workflow efficiency,” the study authors say.