磁振影像雖圖像清晰，但對於心血管系統動態器官的診斷仍處於研究階段，原因為資訊量過大而導致手動方式分析效率不彰，醫療資訊處理自動化為解決上述問題之工具。左心室收縮時於短軸面方向所造成體積及形狀誤差較小，因此核磁共振於左心室之取樣一般以短軸面為準，並以心室之內膜為資料判讀之基準。動態規劃邊界檢測演算法於內膜邊界檢測時會產生彎曲不平之現象，而Snake 演算法雖可提供平滑化邊界，但須輸入初始邊界。本論文將動態規劃及Snake二者結合，相互提供互補效果，發展一無需初始值之邊界檢測暨平滑化系統。統計分析結果皆顯示經由Snake平滑化演算法處理後之左心室內膜邊界與手動圈選平均之左心室內膜邊界之差異，顯著小於未經平滑化處理的邊界。

　　　　　An effective computer aided tool is essential for on-line implementation of Magnetic Resonance Image (MRI) based computer aided diagnosis for dynamic organs such as the cardiovascular system. In short axis left ventricular MR images, the borders of the endocardium are the major parameters used to evaluate the cardiovascular functions such as ejection fraction, end systolic volume and end diastolic volume. The most widely used dynamic programming-based automatic border detection algorithm generates zigzag border lines, which cause measurement errors. In this paper, the automatic border detection output is post-processed by a snake algorithm to smooth over the zigzag phenomena. One drawback of the snake algorithm is the requirement of an initial solution, which can be remedied by using dynamic programming as the pre-processor of the snake algorithm. Statistic results show that the snake algorithm can significantly reduce the errors of border lines.