張璞曾臺灣大學:電機工程學研究所黃鴻鈞Huang, Hung-ChunHung-ChunHuang2007-11-262018-07-062007-11-262018-07-062005http://ntur.lib.ntu.edu.tw//handle/246246/53130血管攝影方式目前分成侵入式(invasive)和非侵入式(non-invasive)兩種,而磁振血管造影(magnetic resonance angiography, MRA)具備非侵入式的優點,所以較易為人所接受,目前臨床上患者多透過MRA來做為血管診斷的先期篩檢。由於國人飲食生活的改變,罹患糖尿病的比例有增高的趨勢,而糖尿病患者多伴隨有下肢周邊血管動脈阻塞的現象,所以透過MRA我們可以清楚瞭解糖尿病患者下肢動脈形態,自主幹至分枝由粗而細,狹窄栓塞及阻塞之處。本論文則是針對MRA之下肢膝部動脈血管影像,嘗試利用二維結構之SOM(Self-Organizing Map)、LVQ(Learning Vector Quantization)類神經網路對膝部動脈血管拓樸形態之辨識,期能作為先期快速篩檢、輔助醫生診斷的工具,証實經過3組PAOD(peripheral arterial occlusive disease)病患和1組正常人影像,共計20張不同角度的影像的實例測試,辨識率高達85%,確實能有效診斷,並作為專家診斷系統並縮短病人等檢查報告時間,有助於整體醫療技術的改善及增進整體醫療的效率。The ways of angiography are divided into two kinds at present: the invasive type and the non invasive type. Because the magnetic resonance angiography (MRA) has advantages of the non invasive type, thus people can accept MRA more easily. Presently, to diagnoses for the initial stage triage of the blood vessel on clinic by MRA mostly. We to be allowed to see clearly that the shape of lower limb artery which like the dendrite and the blood vessel is thick from the trunk to the thin branch, also we can see the narrow embolism and the blocked place through MRA. This study is aiming at the image of artery of blood vessel by MRA assay, and is attempting to use two-dimensional structure of SOM and LVQ to make out topologies for the shape of artery of blood vessel. We expect that MRA could be useful tools for earlier on the quick triage and auxiliary diagnosis of doctors. By actual examples truly prove that patients after peripheral arterial occlusive disease (PAOD) treatment can diagnose effectively, shorten the time of patients waiting for reports and improve the whole efficiency of the medical treatment system.第一章 緒論 1.1 簡介 1 1.2 研究動機 2 1.3 研究目的 5 1.4 研究方法 7 1.5 論文架構 7 第二章 類神經網路原理 2.1 類神經網路架構 9 2.2 SOM類神經網路原理及架構 10 2.3 LVQ類神經網路原理及架構 14 2.4 SOM與LVQ結合模式 20 2.5 文獻回顧 22 第三章 濾波暨其他相關原理 3.1 小波濾波 23 3.2 高斯低通濾波 28 3.3 EXP Enhance 29 3.4 Otus method 30 第四章 系統架構設計 4.1 系統架構 33 4.2 濾波模組 34 4.3 SOM類神經網路 38 4.4 LVQ類神經網路 45 第五章 辨識系統實驗結果 5.1 測試影像暨測試方法 48 5.2 辨識結果 51 第六章 總結 6.1 結論 54 6.2 未來研究方向 55 參考文獻 564228619 bytesapplication/pdfen-US磁振血管造影自組映射學習向量周邊動脈阻塞MRASOMLVQPAOD利用類神經網路於輔助膝部動脈血管之MRA影像辨識Identification of Knee Artery in MRA Images Using Neural Networksthesishttp://ntur.lib.ntu.edu.tw/bitstream/246246/53130/1/ntu-94-P92921012-1.pdf