超音波影像信號對雜訊特徵值與斑點大小的相關性:實驗上的研究

Abstract

超音波影像中存在許多班點(Speckle)，此即為影像上紋路(Texture)主要的來源。由此我們可以推論出，班點大小將決定其解析度。此外肝臟組織十分複雜，以目前的影像分割技術很難自動地決定醫生有興趣的區域（Region Of Interest），或是取樣範圍（Sampling Windows size），尤其在肝炎的影像分割上，影像分割很容易受到肝硬化的影響而無法有效地分割出腫瘤（Tumor）。所以如何決定最適切的取樣範圍的大小來決定紋裡的特徵值問題都跟班點大小或是數量(number)有關。 瀰漫性肝疾病是台灣死亡原因最高之一，本論文中，依照物理統計理論來觀察針對瀰漫性肝炎組織特徵值與斑點大小間的關係是最主要的目的，因此特別製作類似的仿體來模擬病理與實驗的結果來分析，期望在此基礎上能進入真正活體病理上的實驗。

In the ultrasound imaging process, the texture statistics are dominated by speckle formation. And it may be possible to deduce the image resolution is mainly limited to the speckle size. Besides, liver parenchyma is too complicated to decide the region of interest or sampling window size. Especially, it is hardly to capture tumor from liver cirrhosis. Therefore, how to optimize the sampling window size and extract the texture features is involving in speckle size or numbers. Diffuse parenchymal liver disease is one of the top death causes in Taiwan. According to physical statistical theorem, observing the correlation between the feature signal to noise ratio of diffuse parenchymal liver disease and speckle size is the mainly purpose in my paper. For this reason, it is required to analyze the pathological and experiment of fabricating phantoms and on this basis we expect to analyze in vivo pathological experimenting.