Feature | November 05, 2008

Computer-Aided Detection in Screening Mammography Enhances Performance of a Single Reader

Large Scale, Historical Controlled Study Finds CAD Increases Sensitivity in Cancer Detection with Only a Small Increase in Recall Rate

Stock image

Hologic, Inc. published a large scale breast screening trial comparing a single radiologist using computer-aided detection (CAD) to double reading without CAD.

The author of the study is Dr. Matthew Gromet of the Breast Imaging Section of Charlotte Radiology, a large imaging practice (approximately 66,000 screening mammograms a year) in Charlotte, North Carolina. The study , published in the on-line version of the American Journal of Roentgenology, the scientific journal of the American Roentgen Ray Society, appeared in the April 2008 issue of the research journal. The study compared the recall rate, sensitivity, positive predictive value, and cancer detection rate for single reading with CAD, versus double reading without CAD. Biopsy and pathology data for positive cases were also compared.

Dr. Gromet found that a single reader with CAD had a statistically significant increase in sensitivity (11 percent) and a smaller increase in recall rate (4 percent), when compared a single reader without CAD assistance. He also found that single reading with CAD review, when compared with independent double reading, resulted in a not statistically significant increase in sensitivity but with a statistically significant lower recall rate. With manpower constraints limiting the use of double reading, Dr. Gromet concludes that “CAD appears to be an effective alternative that provides similar, and potentially greater, benefits.”

The results were based on the performance of nine radiologists at Charlotte Radiology with a mean level of experience in mammography of 15 years. The mammograms in the study were screen-film studies obtained on Hologic mammography equipment. The CAD technology used in the study was the Hologic R2 ImageChecker CAD system. A total of 231,221 exams were included in the study; 49 percent of the exams were double read without CAD and 51 percent were single read with CAD.

“The Gromet study is a valuable addition to the body of knowledge on the performance of radiologists reading screening mammograms with CAD assistance,” says Ronald A. Castellino, MD, FACR, Chief medical officer for Hologic. “A strength of this study is the large number of malignancies as well as the large number of cases in both arms of the study (double reading without CAD vs. single reading with CAD). The number of CAD cases is nearly four times greater than that reported in the Fenton paper published in the New England Journal of Medicine last year. An additional strength is the use of data from the same pool of experienced radiologists with considerable CAD experience.”

A copy of this study may be found at: http://www.ajronline.org/aheadofprint/04_07_2812_WI.pdf
Note on Sensitivity, Double Reading and CAD

The efficacy of screening mammography is enhanced when sensitivity is high and the recall rate is low. Double reading, though used by few practices in the United States, is a method for increasing sensitivity commonly used in European screening programs. CAD has become increasingly popular as an alternative way to increase sensitivity, since double reading is time-consuming and because of a shortage of radiologists.

CAD has the potential to increase sensitivity of screening mammography by marking areas of interest that might otherwise be overlooked by interpreting radiologists. However, CAD also may place false marks, identifying areas of suspicion that are not cancer. A small fraction of false marks may result in unnecessary patient recalls for further evaluation. The major goal for CAD is to reduce oversights when screening mammograms are read; the radiologist is still responsible for lesion analysis and final interpretation of an examination.

 

Related Content

In a demonstration on the exhibit floor of the SBI symposium, Koios software identified suspicious lesions in ultrasound images

In a demonstration on the exhibit floor of the SBI symposium, Koios software identified suspicious lesions in ultrasound images. Photo by Greg Freiherr

Feature | Artificial Intelligence | April 19, 2019 | By Greg Freiherr
Commercial efforts to develop...
Artificial Intelligence Performs As Well As Experienced Radiologists in Detecting Prostate Cancer
News | Artificial Intelligence | April 18, 2019
University of California Los Angeles (UCLA) researchers have developed a new artificial intelligence (AI) system to...
Videos | Breast Imaging | April 18, 2019
In a keynote lecture at the Society of Breast Imaging (SBI)/American College of Radiology (ACR) 2019 Symposium, ...
Fatty tissue and breast density may be considered in the context of many factors that affect the occurrence and detection of breast cancer

Fatty tissue and breast density may be considered in the context of many factors that affect the occurrence and detection of breast cancer. Permission to publish provided by DenseBreast-info.org

Feature | Breast Imaging | April 18, 2019 | By Greg Freiherr
When planning a screening program to detect the early signs of breast cancer, age is a major consideration.
Oxipit Introduces Multilingual Support for ChestEye AI Imaging Suite
News | Artificial Intelligence | April 16, 2019
The CE-certified ChestEye artificial intelligence (AI) imaging suite by Oxipit is now available in seven European...
iCAD Appoints Stacey Stevens as President
News | Radiology Business | April 16, 2019
iCAD Inc. recently announced that Stacey Stevens has been named president. As president, Stevens will have expanded...
compressed breast during mammography.
360 Photos | 360 View Photos | April 16, 2019
A 360 view of a simulated breast compression for a...
A smart algorithm has been trained on a neural network to recognize the appearance of breast cancer in MR images

A smart algorithm has been trained on a neural network to recognize the appearance of breast cancer in MR images. The algorithm, described at the SBI/ACR Breast Imaging Symposium, used “Deep Learning,“ a form of machine learning, which is a type of artificial intelligence. Graphic courtesy of Sarah Eskreis-Winkler, M.D.

Feature | Artificial Intelligence | April 12, 2019 | By Greg Freiherr
The use of smart algorithms has the potential to make healthcare more efficient.