1.Comparative Analysis of Logistic Regression, Support Vector Machine and Artificial Neural Network for the Differential Diagnosis of Benign and Malignant Solid Breast Tumors by the Use of Three-Dimensional Power Doppler Imaging.
Shou Tung CHEN ; Yi Hsuan HSIAO ; Yu Len HUANG ; Shou Jen KUO ; Hsin Shun TSENG ; Hwa Koon WU ; Dar Ren CHEN
Korean Journal of Radiology 2009;10(5):464-471
OBJECTIVE: Logistic regression analysis (LRA), Support Vector Machine (SVM) and a neural network (NN) are commonly used statistical models in computer-aided diagnostic (CAD) systems for breast ultrasonography (US). The aim of this study was to clarify the diagnostic ability of the use of these statistical models for future applications of CAD systems, such as three-dimensional (3D) power Doppler imaging, vascularity evaluation and the differentiation of a solid mass. MATERIALS AND METHODS: A database that contained 3D power Doppler imaging pairs of non-harmonic and tissue harmonic images for 97 benign and 86 malignant solid tumors was utilized. The virtual organ computer-aided analysis-imaging program was used to analyze the stored volumes of the 183 solid breast tumors. LRA, an SVM and NN were employed in comparative analyses for the characterization of benign and malignant solid breast masses from the database. RESULTS: The values of area under receiver operating characteristic (ROC) curve, referred to as Az values for the use of non-harmonic 3D power Doppler US with LRA, SVM and NN were 0.9341, 0.9185 and 0.9086, respectively. The Az values for the use of harmonic 3D power Doppler US with LRA, SVM and NN were 0.9286, 0.8979 and 0.9009, respectively. The Az values of six ROC curves for the use of LRA, SVM and NN for non-harmonic or harmonic 3D power Doppler imaging were similar. CONCLUSION: The diagnostic performances of these three models (LRA, SVM and NN) are not different as demonstrated by ROC curve analysis. Depending on user emphasis for the use of ROC curve findings, the use of LRA appears to provide better sensitivity as compared to the other statistical models.
Adolescent
;
Adult
;
Aged
;
Aged, 80 and over
;
*Artificial Intelligence
;
Breast Neoplasms/*ultrasonography
;
Diagnosis, Computer-Assisted
;
Diagnosis, Differential
;
Female
;
Humans
;
Image Interpretation, Computer-Assisted
;
Imaging, Three-Dimensional/*statistics & numerical data
;
Logistic Models
;
Middle Aged
;
*Neural Networks (Computer)
;
Predictive Value of Tests
;
ROC Curve
;
Sensitivity and Specificity
;
Ultrasonography, Doppler/*statistics & numerical data
;
Ultrasonography, Mammary/*statistics & numerical data
2.Observer Agreement Using the ACR Breast Imaging Reporting and Data System (BI-RADS)-Ultrasound, First Edition (2003).
Chang Suk PARK ; Jae Hee LEE ; Hyeon Woo YIM ; Bong Joo KANG ; Hyeon Sook KIM ; Jung Im JUNG ; Na Young JUNG ; Sung Hun KIM
Korean Journal of Radiology 2007;8(5):397-402
OBJECTIVE: This study aims to evaluate the degree of inter- and intraobserver agreement when characterizing breast abnormalities using the Breast Imaging Reporting and Data System (BI-RADS)-ultrasound (US) lexicon, as defined by the American College of Radiology (ACR). MATERIALS AND METHODS: Two hundred ninety three female patients with 314 lesions underwent US-guided biopsies at one facility during a two-year period. Static sonographic images of each breast lesion were acquired and reviewed by four radiologists with expertise in breast imaging. Each radiologist independently evaluated all cases and described the mass according to BI-RADS-US. To assess intraobserver variability, one of the four radiologists reassessed all of the cases one month after the initial evaluation. Inter- and intraobserver variabilities were determined using Cohen's kappa (k) statistics. RESULTS: The greatest degree of reliability for a descriptor was found for mass orientation (k = 0.61) and the least concordance of fair was found for the mass margin (k = 0.32) and echo pattern (k = 0.36). Others descriptive terms: shape, lesion boundary and posterior features (k = 0.42, k = 0.55 and k = 0.53, respectively) and the final assessment (k = 0.51) demonstrated only moderate levels of agreement. A substantial degree of intraobserver agreement was found when classifying all morphologic features: shape, orientation, margin, lesion boundary, echo pattern and posterior feature (k = 0.73, k = 0.68, k = 0.64, 0.68, k = 0.65 and k = 0.64, respectively) and rendering final assessments (k = 0.65). CONCLUSION: Although BI-RADS-US was created to achieve a consensus among radiologists when describing breast abnormalities, our study shows substantial intraobserver agreement but only moderate interobserver agreement in the mass description and final assessment of breast abnormalities according to its use. A better agreement will ultimately require specialized education, as well as self-auditing practice tests.
Adenocarcinoma/classification/*diagnosis
;
Adenocarcinoma, Mucinous/classification/*diagnosis
;
Adolescent
;
Adult
;
Aged
;
Aged, 80 and over
;
Biopsy
;
Breast Neoplasms/classification/*diagnosis
;
Carcinoma, Ductal, Breast/classification/*diagnosis
;
Carcinoma, Intraductal, Noninfiltrating/classification/*diagnosis
;
Female
;
Follow-Up Studies
;
Humans
;
Middle Aged
;
Observer Variation
;
Predictive Value of Tests
;
Radiology
;
Reproducibility of Results
;
Sensitivity and Specificity
;
Societies, Medical
;
Terminology as Topic
;
Ultrasonography, Doppler, Color/statistics & numerical data
;
Ultrasonography, Mammary/*statistics & numerical data