Parascript® AccuDetect® Computer-Aided Detection (CAD) software helps radiologists read digital mammograms. Using several complementary algorithms and a patented voting method to achieve high sensitivity and low false-positive rates, AccuDetect identifies areas suspicious for breast cancer for further review. AccuDetect improves the performance of radiologists in discriminating between malignant and nonmalignant cases.
Examining the Results
According to the retrospective comparative study, “Malignant Lesions on Mammography: Accuracy of Two Different Computer-Aided Detection Systems,” (Lobbes et al., Clinical Imaging 37 (2013) 283-288), the previous version of the AccuDetect CAD system (AccuDetect Galileo version 4.0.1) showed better overall performance when compared to iCAD’s Second Look (version 7.2) in detecting masses, microcalcifications, and all cancer types, especially in extremely dense breasts. When compared to Second Look in extremely dense breasts (ACR 4), AccuDetect demonstrated 15.4 percent increase in detection of both masses and calcifications with similar operating points for both CAD systems. Most importantly, for total cancer cases in extremely dense breasts, the percentage of cancers correctly identified by AccuDetect was 14.6 percent higher than that of Second Look, again with similar operating points.
These differences can be directly related to the mechanism that each system uses in detecting lesions. Second Look detects culprit lesions on mammograms using image processing, pattern recognition, and artificial intelligence techniques based on the knowledge from thousands of mammograms. In addition to these well-known approaches, AccuDetect system uses multiple independent cancer detection algorithms and unique patented voting methodology (“Voting in Mammography Processing.” US Patent 8,311,296.) to combine its findings. Comparing the results of the multiple image recognition processes allows for the mitigation of the inherent faults of the recognition processes, thus leading to reduced false-positive and false-negative rates.
Find out more about how Parascript® AccuDetect® works:
The Digital Imaging and Communications in Medicine (DICOM) standard facilitates interoperability of medical equipment. Below is Parascript’s DICOM Conformance Statement for Parascript AccuDetect Computer Aided Detection (CAD) software.
SUPERIOR CAD WORKFLOW PERFORMANCE
- Fast processing time – 11 seconds per image,
45 seconds per 4-view study.
(Compared to: iCAD’s SecondLook: 30 seconds per image, 120 seconds per 4-view study; and Hologic’s R2: 30 seconds per image, 120 seconds per 4-view study.)
(P. Tchou et al., “Interpretation Time of Computer-aided Detection at Screening Mammography.” Radiology. October 2010.)
- Priority queuing of preferred devices for important or urgent studies.
- CAD server supports up to four FFDM systems.
DETECTION OF MICROCALCIFICATIONS
- Detects calcification clusters consisting of several calcifications with size between 0.1mm and 0.8mm. in diameter within 2cm2 area.
- Results for calcifications: 90% sensitivity with 0.75 false positives per four-view study.
DETECTION OF SOFT TISSUE DENSITIES
- Detects soft tissue densities with a diameter
between 5mm and 50mm.
- Results for soft tissue densities: 83.3% sensitivity
with 0.95 false positives per four-view study.
OUTSTANDING STANDALONE PERFORMANCE
AccuDetect’s sophisticated CAD algorithms support early, more accurate detection:
- 91% sensitivity with 1.6 false positives per
- 45% specificity (percentage of normal cases
where CAD did not show any marks).
- Delivers very high performance on dense/extremely dense breasts – 85% sensitivity, 43% specificity, 1.5 false positives per four-view study. (The differences in both sensitivity and specificity between AccuDetect CAD results for dense or extremely dense breasts and fatty or scattered fibroglandular breasts are statistically insignificant.)
AccuDetect 7.0 includes a Windows 7, multi-vendor CAD server that provides consistency across all digital mammography systems. Please see the AccuDetect brochure for details.