講演抄録/キーワード |
講演名 |
2009-01-19 13:45
Using Tumor Morphology to Classify Benign and Malignant Solid Breast Masses:Speckle Reduction Imaging (SRI) versus Non-SRI Ultrasound Imaging ○Hsien-Chang Chuang(Tunghai Univ.)・Dar-Ren Chen(Changhua Christian Hospital)・Yu-Len Huang(Tunghai Univ.) MI2008-64 |
抄録 |
(和) |
In this paper, we attempted to compare the diagnostic performance of speckle reduction imaging (SRI) ultrasound and non-SRI ultrasound using sonographic technique to classify benign and malignant breast tumors by morphology. This study evaluated 110 breast lesions. A total of 72 benign and 38 malignant breast tumor images with pathologic proven cases were analyzed. The suspicious tumor contours on both SRI and non-SRI ultrasound images were manually sketched by experienced physicians. Twenty practical morphologic features from the extracted contour were calculated and a support vector machine (SVM) classifier identified the breast tumor as benign or malignant. Conventional binormal receiver operating characteristics (ROC) curve analysis used overall morphologic features from the breast lesions in SRI and non-SRI ultrasound. The area under the ROC curve (Az) was 0.8105 and 0.8241, respectively. This difference was not statistically significant (p = 0.343). The sensitivity was 78.9% and 84.2% (p = 0.554), respectively; the specificity was 73.6% and 70.8% (p = 0.881), respectively; the difference were not statistically significant. According to our study, both non-SRI and SRI methods are helpful to classify benign and malignant breast tumors by morphology, and the diagnostic performance is almost identical without statistical significance. |
(英) |
In this paper, we attempted to compare the diagnostic performance of speckle reduction imaging (SRI) ultrasound and non-SRI ultrasound using sonographic technique to classify benign and malignant breast tumors by morphology. This study evaluated 110 breast lesions. A total of 72 benign and 38 malignant breast tumor images with pathologic proven cases were analyzed. The suspicious tumor contours on both SRI and non-SRI ultrasound images were manually sketched by experienced physicians. Twenty practical morphologic features from the extracted contour were calculated and a support vector machine (SVM) classifier identified the breast tumor as benign or malignant. Conventional binormal receiver operating characteristics (ROC) curve analysis used overall morphologic features from the breast lesions in SRI and non-SRI ultrasound. The area under the ROC curve (Az) was 0.8105 and 0.8241, respectively. This difference was not statistically significant (p = 0.343). The sensitivity was 78.9% and 84.2% (p = 0.554), respectively; the specificity was 73.6% and 70.8% (p = 0.881), respectively; the difference were not statistically significant. According to our study, both non-SRI and SRI methods are helpful to classify benign and malignant breast tumors by morphology, and the diagnostic performance is almost identical without statistical significance. |
キーワード |
(和) |
speckle reduction imaging / breast sonography / tumor classification / tumor segmentation / support vector machine / / / |
(英) |
speckle reduction imaging / breast sonography / tumor classification / tumor segmentation / support vector machine / / / |
文献情報 |
信学技報, vol. 108, no. 385, MI2008-64, pp. 33-36, 2009年1月. |
資料番号 |
MI2008-64 |
発行日 |
2009-01-12 (MI) |
ISSN |
Print edition: ISSN 0913-5685 Online edition: ISSN 2432-6380 |
著作権に ついて |
技術研究報告に掲載された論文の著作権は電子情報通信学会に帰属します.(許諾番号:10GA0019/12GB0052/13GB0056/17GB0034/18GB0034) |
PDFダウンロード |
MI2008-64 |