陳中明2006-07-262018-06-292006-07-262018-06-292002-07-31http://ntur.lib.ntu.edu.tw//handle/246246/22358In the third year, we have developed a new segmentation scheme, which is capable of extracting multiple hepatic tumors simultaneously from an ultrasound image. Moreover, we have combined the segmentation results and the classification scheme to achieve a classification accuracy of 85.78% with leave-one-out crossvalidation by using only texture information. As a future study, it is believed that the performance can be further improved by incorporating the characteristics of the boundary vicinity information.在第三年計畫中，我們發展出可以從 超音波影像中同時分割出多個腫瘤的技 術，並且結合了分割的結果與分類的技術 而達到85.78%的良惡性腫瘤辨識率。此一 結果僅使用了紋路訊息。作為研究的下一 個工作，我們相信加入邊緣附近的訊息將 可進一步的改良辨識率。application/pdf229818 bytesapplication/pdfzh-TW國立臺灣大學醫學工程學研究所Computer assisted diagnosisHepatic tumorUltrasound imageImage SegmentationData Mining電腦輔助診斷肝臟腫瘤超音 波影像影像分割資料發掘以資料發掘與視覺模型為基礎之電腦輔助超音波肝臟腫瘤之區別診斷(3/3)Computer Assisted Differential Diagnosis of Hepatic Tumors in Ultrasound Images Based on Data Mining and Vision Models (3/3)reporthttp://ntur.lib.ntu.edu.tw/bitstream/246246/22358/1/902213E002123.pdf