請用此 Handle URI 來引用此文件:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/84827
標題: | 影像處理於血液透析患者之分析 Image processing on analysis of dialysis patients |
作者: | 郭縉 KUO CHIN |
指導教授: | 蕭浩明 Hao-Ming Hsiao |
關鍵字: | 血液透析,血管阻塞,影像處理,影像放大,遠距光體積變化描紀圖, dialysis,vessel occlusion,image processing,video magnification,remote photoplethysmography, |
出版年 : | 2022 |
學位: | 碩士 |
摘要: | 腎臟疾病在台灣盛行許久,其嚴重程度在世界上的排名更是名列前茅。而在腎臟病嚴重影響到病患的生理機能時,洗腎便是無可避免的。台灣洗腎多數採用血液透析的方式,血液透析透通常需要在手臂放置瘻管增強血液循環中的血流以加速洗腎的效率,但瘻管伴隨的問題是血管的阻塞,故瘻管的健康程度對洗腎病患來說相當的重要。檢查血管阻塞的方法有許多種,但這些方法不外乎都需要專業人員的操作及判讀,甚至是進行侵入式手術。在台灣眾多的洗腎病患下,光是為了檢測瘻管的健康程度便會造成醫療上面的巨大負擔。近年影像處理相關的應用越來越廣,其中在醫療方面的研究更是族繁不及備載。影像的好處在於可以透過非侵入式的方式進行,並且操作簡易。 本研究與台大新竹分院的吳志成主任合作,從台大新竹分院中收取六位的洗腎患者進行研究。此六位患者會在血管通暢和阻塞時各進行影像的拍攝並透過影像處理的方式進行血管健康程度的分析。 本研究拆成兩部分,第一個部分將探討遠距光體積變化描紀圖(rPPG)在手臂上的可行性。此部分對現有的文獻進行討論,以及透過血管擷取的改良方法進行比較,而改良的方法對手臂的rPPG方法可以看出具有發展性。第二個部分會藉由尤拉影像放大的方式,將手臂的細微跳動放大,並透過光流法抓取像素的位移量,透過影片時間內的位移總和轉為色彩圖以直觀的方式觀察血管阻塞時和健康時的差別;另外記錄每一幀下手臂的平均跳動,觀察長時間跳動的變化,本研究透過訊噪比(SNR)以及曲線擬合得到的方均差進行量化,並在訊噪比的量化能從現有的病患資料得到一個初步的分界標準,對未來相關研究開闢一個新的研究方向。 Kidney disease has been prevalent for many years in Taiwan. The epidemic rate is even at the front tier in the world. When kidney disease starts to erode the kidney and cause irreversible damage, having dialysis is inevitable. In Taiwan, the most common way of dialysis is Hemodialysis. Hemodialysis will create outer blood circulation to fasten the speed of dialysis. But by doing so, vascular occlusion will occur easily. There’re lots of ways to examine the severity of occlusion, but most of them require professionals to operate. The examination itself costs lots of medical resources. Recently, image processing has been applied to different medical uses. Image processing can create a non-invasive and professional-less way to monitor our health conditions. In the research, we cooperate with doctor Wu in National Taiwan Hospital Hsin-Chu Branch. Collecting 6 patients having dialysis, we would take their forearms’ video when there’s occlusion occured and occlusion-free. Then will use the technique in image processing to analyze our data. Our research sperate into two methods of analysis. The first method is remote-Photoplethysmography(rPPG). We will use the existing rPPG method to apply on a human’s forearm, then we’ll use a NIR camera to capture a human’s vessel and propose a revised way to improve the performance of the traditional method. The second method is by Eulerian video magnification to magnify the tiny pulse on our skin, and use optical flow to track the movement. We found out that when occlusion occurs, patients’ pulse will have some sparky and irregular signal compared to when they’re occlusion-free condition. We will describe this phenomenon in numeric values of Signal to Noise Ratio(SNR) and Mean Square Error(MSE). We hope by collecting more patients’ data, we can find out a specific feature to examine dialysis patients' health and replace the original examination which cost lots of medical resources. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/84827 |
DOI: | 10.6342/NTU202202892 |
全文授權: | 同意授權(限校園內公開) |
電子全文公開日期: | 2027-08-29 |
顯示於系所單位: | 機械工程學系 |
文件中的檔案:
檔案 | 大小 | 格式 | |
---|---|---|---|
ntu-110-2.pdf 目前未授權公開取用 | 4.44 MB | Adobe PDF | 檢視/開啟 |
系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。