Online Journal
IF値: 1.878(2021年)→1.8(2022年)


Journal of Medical Ultrasonics

にて英文誌のFull textを閲覧することができます.


2000 - Vol.27

Vol.27 No.02

Original Article(原著)

(0119 - 0129)


Study of Automated Breast Tumor Extraction and Diagnosis Using Three-Dimensional Ultrasonic Imaging: Multivariate Logistic Regression Analysis with Multiple Parameters

尾本 きよか1, 伊東 紘一1, 程 相勇2, 王 怡1, 谷口 信行1, 秋山 いわき3, 大塚 紳4, 水沼 洋文4, 小倉 重人4, 金澤 暁太郎4

Kiyoka OMOTO1, Kouichi ITOH1, Xiangyong CHENG2, Yi WANG1, Nobuyuki TANIGUCHI1, Iwaki AKIYAMA3, Shin OTSUKA4, Hirobumi MIZUNUMA4, Shigeto OGURA4, Kyotaro KANAZAWA4

1自治医科大学臨床病理学, 2三谷産業株式会社, 3湘南工科大学電気工学科, 4自治医科大学外科

1Department of Clinical Pathology, Jichi Medical School, 2Mitani Sangyo Co., Ltd., 3Department of Electrical Engineering, Shonan Institute of Technology, 4Department of Surgery, Jichi Medical School

キーワード : Automated breast cancer diagnosis system (ABCD system), Breast tumor, Computer-aided diagnosis (CAD), Multivariate logistic regression analysis, Three-dimensional ultrasonic imaging

We are developing a new method of breast cancer screening that we call our Automated Breast Cancer Diagnosis (ABCD). This system uses computerized three-dimensional (3D) imaging techniques and statistical analysis to base diagnosis on data obtained from ultrasound images. Here we use this system to investigate ten parameters and their effectiveness and to show their effectiveness in determining the malignancy of tumors. Twenty-nine benign tumors and 32 malignant tumors were studied. The benign tumors compried 8 cysts and 21 fibroadenomas; the malignant tumors were 23 ductal carcinomas, 2 special carcinomas, 1 malignant lymphoma, and 6 other types of lesions. The procedure requires the simultaneous acquisition of both the ultrasonic image data and the position and orientation of the probe for each slice. This data is transferred to a computer, where the tumor surface is determined using fuzzy reasoning and relaxation techniques. The extracted tumor image is then rendered in 3D, allowing interactive manipulation and observation. A significant distinction between benign (0.57±0.25, 2.08±0.12, 0.76± 0.25) and malignant tumors (0.78±0.33, 2.22±0.16, 0.58±0.29) was obtained for all three parameters (Sz/Sxy, M-D, and Vei/V). A malignancy probability expression is calculated using multivariate logistic regression analysis in combination with the five parameters (Sz/Sxy, M-D, Vei/V, 3D-D/W, and S/Vindex). Good results were obtained when this expression was applied to three benign and two malignant tumors additional tumors.