English  |  正體中文  |  简体中文  |  Items with full text/Total items : 69561/69561 (100%)
Visitors : 23072333      Online Users : 382
RC Version 7.0 © Powered By DSPACE, MIT. Enhanced by NTU Library IR team.
Scope Tips:
  • please add "double quotation mark" for query phrases to get precise results
  • please goto advance search for comprehansive author search
  • Adv. Search
    HomeLoginUploadHelpAboutAdminister Goto mobile version


    Please use this identifier to cite or link to this item: http://ir.lib.ncu.edu.tw/handle/987654321/82140


    Title: 醫學影像的精細結構自動化辨識定位及其三維影像重構之研究;Research on Automatic Identification and Localization of Medical Imagery Morphology and Its 3-Dimensional Image Reconstruction
    Authors: 陳健章
    Contributors: 國立中央大學生醫科學與工程學系
    Keywords: 生物科學;計算生物;生物資訊;影像處理;影像重構;統計方法;核磁共振影像;非監督學習;Lucas-Kanade 光流法;像素連結法;前後偵測法;密度泛函理論;資料密度泛函方法;辨識;定位;視覺化;Biological sciences;Computational biology;Bioinformatics;Image processing;Image reconstruction;Statistical methods;MRI;Unsupervised learning;Lucas-Kanade optic flow;Pixel connectivity;Forward-backward detection;Density functional theory;Data density functional method;Identification;Localization;Visualization
    Date: 2020-01-13
    Issue Date: 2020-01-13 14:17:24 (UTC+8)
    Publisher: 科技部
    Abstract: 本三年期計畫之目標為以本期計畫所發展之資料密度泛函方法為基礎,發展一種可自動化用於精準識別、定位和三維可視化於大尺度醫學影像中之微小多病灶的低計算複雜度方法。根據前期研究(已刊登於Scientific Reports、Applied Sciences等),我們已成功地連接了量子力學和人工智能技術,進而創建了一種稱為資料密度泛函方法的自動化演算法。亦基於本期MOST計畫之研發成果,我們已成功整合三種於人工智慧領域中,重要且實用之技術:Lucas-Kanade光流,像素連通性和廣義梯度近似。且已使用該方法解決了如影像對齊、像素標記等醫學成像技術中常見之問題。對於特定問題的解決技術,亦在MATLAB和Python環境之平台上建立了相關技術的自動演算法。一系列SCI期刊論文和會議論文業已發表在跨學科領域,如機器視覺,康復輔助,步態分析,動態物體跟踪等方面。然而,這些研究成果和相關的演算法也指出了資料密度泛函方法無可避免的劣勢。由於本方法是通過連接量子力學和機器學習方法構建,因此所採用的醫學影像的所有信息將會全域映射到特定的能量空間。這意味著在測量目標影像的相似性時,會將所有像素信息納入估計過程。圖像尺寸越大,計算複雜度越高。此外,目前的資料密度泛函方法亦無法直接應用於醫學圖像的多腫瘤檢測的情況,造成輔助檢測和其他演算法估計是無可避免的。因此為了克服上述窘境並加強資料密度泛函方法之優勢,在此為期三年的計劃中我們期待透過納入新穎的自動化技術來增加資料密度泛函方法於大尺度醫學影像中精確識別,定位和三維視覺化三方面的功效,並將此基礎研究擴展到實際應用。需強調的是,儘管我們的初步研究成功地實現了腦腫瘤自動檢測技術及其三維視覺化方法,但針對微小之腦部組織如丘腦底核、相應的鄰近區域和其他腦組織的精確自動識別、定位和三維視覺化,仍是臨床研究和手術導引中的棘手問題。 ;The objective of this three-year proposal is to develop an automatic methodology for precisely identification, localization, and 3-dimensional visualization of tiny multi-lesion from large-scale medical imageries using the data density functional method with low computational complexity. The research consequences will benefit planning and guidance of modern surgery, clinical investigations, rehabilitation, and so forth. Based on our previous research work, we had successfully connected the quantum mechanics and the technique of artificial intelligence then created an automatic methodology called data density functional method. Also based on the research consequences from our MOST proposal this year, three contemporary techniques, the Lucas-Kanade optic flow, the pixel connectivity, and the generalized gradient approximation, had been integrated into our proposed method. The hybridized method eventually resolves the issues from image film alignment and labeling of pixels, the common problems in medical imaging setting. Then 3-dimensional brain tumor morphologies were successfully reconstructed from sets of magnetic resonance (MR) imagery automatically. Relevant automatic algorithms had also been manipulated on the platforms of MATLAB and Python programming environments for specific applications and problem solutions. A series of SCI journal papers and conference papers were published in interdisciplinary fields including machine vision, rehabilitation assistance, gait analysis, dynamic object tracking, and so forth. However, the research consequences and the relevant automatic algorithms also pointed out an inevitable inconvenience from the proposed method. The proposed method was constructed by connecting the quantum mechanics and machine learning methods. All information of the employed medical imagery was globally mapped into a specific energy space to estimate the relevant energy functionals. This means that while measuring the similarity of an employed medical imagery of interest, whole of pixel information should be taken into the estimation processes. The larger the image size, the higher the computational complexity. The computational complexity of the proposed method is about O(n^2), where n represents the image dimension. Additionally, the proposed method could not be directly applied on the cases of multi-tumor detection of a medical imagery. Auxiliary detection and additional algorithmic estimates are inevitable. To conquer the aforementioned predicaments and reinforce the data density functional method, thus, in this three-year proposal we would like to extend those fundamental studies to pragmatic applications by providing a new automatic methodology for precisely identification, localization, and 3-dimensional visualization of multi-lesion from large-scale medical imageries with low computational complexity. It should be emphasized that even though the preliminary researches were successfully achieved on the problem of automatic brain tumor detection as well as its 3-dimensional visualization, the precisely automatic identification, localization, and 3-dimensional visualization of the target tiny tissues, subthalamic nucleus, the corresponding neighboring regions, and the other brain tissues are still tough problems in clinical investigations and surgery guidance.
    Relation: 財團法人國家實驗研究院科技政策研究與資訊中心
    Appears in Collections:[生醫科學與工程學系] 研究計畫

    Files in This Item:

    File Description SizeFormat
    index.html0KbHTML65View/Open


    All items in NCUIR are protected by copyright, with all rights reserved.

    社群 sharing

    ::: Copyright National Central University. | 國立中央大學圖書館版權所有 | 收藏本站 | 設為首頁 | 最佳瀏覽畫面: 1024*768 | 建站日期:8-24-2009 :::
    DSpace Software Copyright © 2002-2004  MIT &  Hewlett-Packard  /   Enhanced by   NTU Library IR team Copyright ©   - Feedback  - 隱私權政策聲明