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, 永井 秀雄⁴

Kiyoka OMOTO¹, Kouichi ITOH¹, Xiangyong CHENG², Yi WANG¹, Nobuyuki TANIGUCHI¹, Iwaki AKIYAMA³, Shin OTSUKA⁴, Hirobumi MIZUNIMA⁴, Shigeto OGURA⁴, Hideo NAGAI⁴

¹自治医科大学臨床検査医学, 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 Laboratory Medicine, Jichi Medical School, 3311-1 Yakushiji, Minamikawachi-machi, Kawachi-gun, Tochigi-ken 329-0498, Japan, ²Mitani Sangyo Co., Ltd. 2-6 Asahidai, Tatsunokuchi-machi, Nomi-gun, Ishikawa-ken 923-1211, Japan, ³Department of Electrical Engineering, Shonan Institute of Technology, 1-1-25 Tsujido-Nishikaigan, Fujisawa-shi, Kanagawa-ken 251-8511, Japan , ⁴Department of Surgery, Jichi Medical School, 3311-1 Yakushiji, Minamikawachi-machi, Kawachi-gun, Tochigi-ken 329-0498, Japan

キーワード : Automated Breast Cancer Diagnosis System (ABCD system) , Breast tumor, Computer-aided diagnosis (CAD), Preoperative predictability, Three-dimensional ultrasound imaging

We recently developed the Automated Breast Cancer Diagnosis (ABCD) system, which is based on three dimensional (3D) ultrasound imaging technology. Here we evaluated clinically its effectiveness in preoperative estimation of malignancy. Specifically, we used five statistical parameters in applying the system to 64 lesions and use our results to place each tumor into one of three classes. Thirty-five benign tumors and 29 malignant tumors were studied. The benign tumors comprised 17 cysts, 14 fibroadenomas, and 4 mastopathic lesions; the malignant tumors included 20 ductal carcinomas, 2 mucinous carcinomas, 1 metastatic lesion, and 6 others. The probe's position and orientation were tracked during scanning using a magnetic field positioning sensor that was mounted on the probe. This information, together with the ultrasound image video signal was then fed into the computer. Next, coordinate transformation and fuzzy reasoning techniques were used to extract and render a 3D image of the tumor area. Pmax was calculated using five quantitative parameters from this tumor area and was then used to place each tumor into one of three classes: class A, Pmax=0 to 15; class B, Pmax=16 to 32, and class C, Pmax=33 to 100. Each class was later compared with results obtained on pathological examination. In addition, the system's sensitivity, specificity, and accuracy were calculated by defining classes A and B as negative (-) and class C as positive (+). Using Pmax, 16 benign and 3 malignant tumors were classified as class A; 12 benign and 3 malignant tumors, class B; and 7 benign and 23 malignant tumors, class C. The system's sensitivity is therefore evaluated at 79%; specificity, at 80%; and accuracy, at 80%. The preoperative predictability results indicated potential value of this system in screening for breast cancer.