Evaluation of Computer Aided Diagnosis (CAD) in the Interpretation of Mass Lesions seen on Breast Sonography.
- Author:
Tae Il YOON
1
;
Shin Hye PARK
;
Jong Suk BYON
;
Hee Boong PARK
Author Information
- Publication Type:Original Article
- Keywords: Computer aided diagnosis; breast sonography; breast cancer
- MeSH: Breast Neoplasms; Breast*; Diagnosis*; Diagnosis, Differential; Female; Humans; Korea; Retrospective Studies; ROC Curve; Ultrasonics; Ultrasonography
- From:Journal of Breast Cancer 2006;9(2):110-114
- CountryRepublic of Korea
- Language:Korean
- Abstract: PURPOSE: Sonography for the diagnosis of breast cancer, is important in Korea because Oriental women have denser breast and their breast cancer generally occurs at a younger age. We investigate the potential usefulness of computer aided diagnosis(CAD) as a tool for physicians to use in the differential diagnosis of the breast mass lesions seen on ultrasound examinations. METHOD: We evaluated a series of pathologically proven breast tumors that were detected in our private clinic from May 2004 to April 2005. A total of 134 breast tumors (55 malignant and 79 benign tumors) were retrospectively evaluated. We located the regions of interest (ROI) on two ultrasonic images (the vertical and longitudinal views) for each mass lesion and we evaluated the malignant score with using the SonoEye2DTM(Cad Impact Inc. Korea) system. Triple measurements were performed for each image and we then analyzed the data as two groups. The first group had their data analyzed by using the two perpendicular images of each mass and the second group had their data analyzed by using the one image with the longest diameter of the mass. RESULT: The sensitivity of CAD on ultrasound images was 81.8%-96.4% for the first group and 74.6-87.3% for the second group and the specificities were 53.2-73.4%, and 59.5-68.4% respectively. The area under the receiver operating characteristic curve(AZ) was 0.811-0.816 for the first group and 0.780-0.800 for the second group. CONCLUSION: The CAD system for the ultrasound images of breast lesions was useful in the diagnosis of breast cancer. The diagnostic accuracy for the analysis of two images was higher than for the analysis of one image.
