結合相位展開技術實現基於區塊的彩色都卜勒去混疊

Abstract

彩色都卜勒作為一種可視化工具，對於診斷心臟和血管相關疾病，例如瓣膜周圍逆流和狹窄的問題，至關重要。然而，根據奈奎斯特準則，當血流速度超過系統能顯示的最大速度(即奈奎斯特速度)時，會產生混疊效應，導致速度資料的顯示與真實情況不相符。因此，本研究致力於開發一套去混疊演算法，並驗證在不同情境下的有效性，以解決彩色都卜勒的混疊問題。我們首先回顧了現有的去混疊方法，包括基於區塊分割之演算法、二維相位展開演算法和多種脈衝重複頻率交錯方法。分別介紹了它們的運作原理，並通過實例說明應用效果。由於這些方法各自存在著一些限制，導致難以達到理想效果，因此本研究提出了一套新的去混疊方法來克服這些不足。此技術的特點包含在區塊分割階段導入閾值設置，在速度校正時以區塊為單位並結合相位展開的概念，藉此解決分割不足的問題同時降低資料失真的發生。除此之外，我們通過取樣的方式獲得多種脈衝重複頻率的資料，隨後進行去混疊處理，期望藉由比較不同脈衝重複頻率設置下的處理結果，找出最佳的去混疊結果，從而提升訊號還原的正確性。為了評估演算法的成效，我們使用多種類型的資料進行測試，資料來源包括蒙地卡羅模擬、血管仿體實驗、活體實驗及臨床彩色都卜勒影像。實驗結果顯示，本研究提出的演算法在各種類型的資料中，正規化均方根誤差均優於其他演算法，顯示出其具有較高的適應性和通用性，能有效應對各種不同類型的資料。在臨床上，提供穩定可靠的去混疊結果，能協助醫生找出發生速度峰值的確切位置，有助於疾病診斷與後續治療。此外，本研究還為彩色都卜勒影像的去混疊設計了一個能與使用者互動的圖形介面，提高了該演算法在實際應用中的便捷性。

Color Doppler is an imaging tool for diagnosing heart and vascular diseases such as valvular regurgitation and stenosis. However, according to Nyquist criterion, when blood flow velocities exceed Nyquist velocity, aliasing occurs. This results in inaccuracies of velocity display compared to actual conditions. Therefore, this study aims to resolve aliasing problems in Color Doppler data by developing a dealiasing algorithm and validating the effectiveness across different scenarios. We start with reviewing existing dealiasing methods, including the segment-based dealiasing algorithm, the 2D phase unwrapping algorithm, and the staggered multi-PRF method. We introduce their underlying principles and demonstrate their applications. Each approach has some limitations, and this research proposes a new method to overcome these limitations. The technical highlights include introducing threshold setting in the segmentation stage and reconstructing the velocity field based on segment units, incorporating the concept of phase unwrapping during the process to alleviate the under-segmentation problem and mitigate data distortion. In addition, the acquisition of data with multiple PRFs is achieved through simple data decimation, followed by the dealiasing process. By comparing the results from different PRF settings, the optimal outcome is determined, thus enhancing the accuracy. To evaluate the algorithms, we applied them to various data collected from Monte Carlo simulations, flow phantom experiments, in vivo experiments, and clinical Color Doppler images. The experimental results demonstrate that the proposed algorithm achieves lower normalized root-mean-square error (NRMSE) than other algorithms across different data, indicating high adaptability and versatility. Providing the reliable dealiasing result can help doctors identify the specific position causing the peak velocity, thereby assisting in diagnosis and subsequent treatment. Furthermore, this study develops a user-interactive graphical interface, enhancing practical usability in real-world applications.