利用雷射散斑統計特性進行非侵入式眼底血流成像

Abstract

在此論文中，我們整合一套以傳統眼底鏡為基礎的雷射光學系統，建構出眼底血流成像的架構。由於雷射光源具有高同調性，光場被散射後其自干涉形成負指數型的強度分布，稱為雷射散斑圖案，此圖案可以作為天然的結構光。當散斑圖案受到移動的紅血球影響產生散射變化時，在曝光時間內會被平均掉並改變其統計特性。為了驗證此理論，我們使用兔子進行動物實驗，經由雷射光源獲得兔眼視網膜的雷射散斑影像。一系列的原始散斑影像經由演算法處理後，可以分別得到統計特性包括散斑對比(空間域)與同調時間(時間域)。為了得到最佳的影像品質與參數，三種演算法包含空間式、時間式、時間-空間複合式同時採用並互相比較。最後，我們採用時間的演算法，因高空間解析度的優點可以得到較多血管細節與流速資訊。

In this thesis, we integrated a laser-based imaging optics into a conventional fundus camera. Due to the high coherence property of the laser source, the speckle pattern formed by the scattering and self-interference can be adopted as a naturally structured light whose intensity obeys the negative exponential distribution. Modulated by the blood flow, the speckle pattern was averaged out during exposure time to change its statistical characteristics. We conducted an animal experiment to acquire a laser speckle images from the rabbit retina. A series of raw speckle images were processed to obtain the statistical properties including the speckle contrast (space domain) and the correlation time (temporal domain), respectively. To achieve the optical image quality and parameter estimation, three algorithms involving spatial, temporal, and spatiotemporal respectively were utilized and compared. Finally, we resorted to the temporal contrast scenario to reveal more detailed vessels as well as a flowgraphy.