冠狀動脈自動化偵測技術之改進方法

Abstract

冠狀動脈疾病(Coronary Arterial Disease, CAD)是現代人常面臨的疾病之一，同時也是主要死因之一。通常發生在高齡者，然而高油脂的飲食與缺發運動的生活習慣，罹患CAD的年齡層逐漸下降。由於電腦斷層掃描技術的進步，快速且為非侵入式的冠狀動脈血冠造影是CAD主要的預防檢測方法。判讀照影結果必須具有專業知識，而大量的影像需到耗費龐大心力檢視。因此希望能提出改進自動化冠狀動脈擷取技術幫助診斷。 擷取冠狀動脈的主要方法之一是先從心臟中取得可能的冠狀動脈點，然後再進行連結。取得的點越精確，就越能得到完整的冠狀動脈。本論文利用冠狀動脈在心臟的分布走向，找出冠狀動脈環繞心臟的平面並作為模板篩選冠狀動脈候選點，以降低自動化擷取過程中雜訊的干擾，可以減少計算量並減少錯誤的發生。

Coronary arterial disease (CAD) is one of today's major causes of death. Although it is usually considered a disease for older people, put lifestyles of high-fat diets and lack of exercises have lowered the ages of people getting this disease. Given the progress of computed tomography techniques, this fast and non-intrusive technique has an important tool for detecting and diagnosing CAD. However, it takes specialized knowledge and is time-consuming to analyze the CT data. As a result, automatic coronary artery extraction techniques have become very useful. One major approach for extracting coronary arteries is to first identify potential candidate points in the CT images, and then link them to form the coronary trees. If we can obtain very accurate candidate points, we can get very complete coronary arteries. In this thesis, our objective is to utilize the typical spatial positioning of coronary arteries within the heart, focusing on the two planes that typically contain the major coronary arteries. By detecting these planes first, we can filter the candidate points to remove noise points, reduce amount of computation, and lower the chances of false detections.