KMID : 1114620130100010012
|
|
Journal of the Korean Society for Breast Screening 2013 Volume.10 No. 1 p.12 ~ p.16
|
|
Computer-aided Detection System in Automated Breast Volume Scanner: Development and Evaluation of the Effectiveness
|
|
Kim Jeoung-Hyun
Cha Joo-Hee
|
|
Abstract
|
|
|
Purpose: The aim of this study was to develop a computer-aided detection (CAD) system for breast lesion in automated breast volume scanner and to evaluate the performance of the proposed CAD system.
Materials and Methods: 18 volume data (which composed of 6 sets of anteriorposterior and both oblique positions) of breast phantom were obtained by automated breast volume scanner. Gaussian pyramid and median filtering was applied for down sampling and noise removal from the image. After boundary detection and removal, mass candidate were detected and segmented using Otsu¡¯s threshold. Then the features of segmented objects are extracted. The cut off values of three features (Eccentricity, perimeter/area, cluster size) were determined by gradual increase or decrease of the threshold. The overall performance was evaluated using 6 sets of volume data.
Results: The overall accuracy of our CAD system was 89% (standard deviation; 8.61%). The number of false positives was 2¡¾0.63 per a volume data.
Conclusion: Our CAD system has relatively high accuracy, and further development with more data including 3D features would generate more improved CAD.
|
|
KEYWORD
|
|
Automated breast volume scanner, Computer-aided detection system
|
|
FullTexts / Linksout information
|
|
|
|
Listed journal information
|
|
|