運用特徵追蹤法在三維超音波平面波影像上的運動分析

Motion Analysis In Three Dimensional Ultrasound Plane-Wave Excitation Imaging By Feature Tracking

曾令儀(Lin-Yi Tseng)

指導教授 ： 李百祺

國立臺灣大學/電機資訊學院/生醫電子與資訊學研究所/碩士(2012年)

https://doi.org/10.6342/NTU.2012.00828

摘要

心臟超音波應變影像是臨床用來評估心肌運動狀態的檢測工具，其主要的原理是應用斑點追蹤法偵測心臟超音波動態影像心肌內外膜上斑點位移前後的相對位置，再進一步推估出心肌應變量並藉以提供臨床醫生判定心臟功能狀態的資訊，但目前臨床上僅限於應用在二維超音波影像上。由於心臟是人體內博動速度最快的器官，並合併有伸長、縮短、扭轉等複雜的運動模式，要能完整獲得如此複雜的形變資訊，就必須要發展即時的三維心臟超音波應變影像(Real time 3D echocardiographic strain image )；而超音波平面波成像法擁有最快的成像速率，能夠達到即時成像的可能，另外由斑點追蹤法衍生而來的特徵追蹤法能夠解決斑點追蹤法在高維度影像資訊中遭遇龐大運算量的問題，因此本研究的目的是希望結合平面波成像法與特徵追蹤法以建立即時的三維心臟超音波應變影像。
在本研究中，我們模擬了三維平面波動態影像，並進行斑點追蹤與特徵追蹤的分析比較。結果發現，平面波影像較雙向聚焦影像在橫向位移上有較大的追蹤誤差，若將平面波影像在極座標上進行追蹤會有較佳的追蹤結果；特徵追蹤法較傳統斑點追蹤法有極佳的運算效能，但追蹤誤差較大；應用在旋轉的影像上時，特徵追蹤法與斑點追蹤法都有良好的追蹤正確率，但特徵追蹤的誤差仍較大；在特徵追蹤演算法中，當篩選特徵斑點的域值越大或內核(kernel)設定範圍越大時，特徵追蹤的追蹤誤差會越低，但所篩選的特徵斑點數目也會越少。根據上述各項進行的分析結果，我們認為結合平面波成像法與特徵追蹤法來建立即時三維心臟超音波應變影像是可行的。最後，我們嘗試著將特徵追蹤法應用於臨床三個月大嬰兒的三維心臟超音波影像分析上，由於受限於臨床超音波影像品質，心臟內外膜上篩選的特徵斑點過少，所得到的三維心臟應變影像無法表示整體心臟運動形態，因此未來還需要針對特徵追蹤法在臨床影像上的應用作進一步探討和分析。

關鍵字

心臟超音波應變影像 ； 斑點追蹤 ； 特徵追蹤 ； 高速超音波成像 ； 超音波平面波成像

並列摘要

Echocardiographic strain imaging is a clinical tool to assess the myocardial motion. Speckle tracking is typically applied to detect displacement of speckles on endocardium and epicardium. Furthermore, it can estimate the myocardial strain to help clinicians to evaluate cardiac functions, but its clinical applications are mainly limited to two-dimensions. As the heart is the fastest-moving organ, and is associated with elongation, shortening, torsion movement patterns, it is necessary to develop real-time three-dimensional strain echocardiography image for the acquisition of complete information of such a complex deformation. In view of the fact that plane-wave excitation imaging has the highest frame rate which makes it possible to achieve real-time three-dimensional imaging, and that the other feature tracking method derived from the speckle tracking method is able to solve the problem that three-dimensional speckle tracking is too computationally intensive for practical use, the aim of this study is to combine the plane-wave excitation imaging method and the feature tracking method to construct three-dimensional echocardiographic strain images. In this study, we simulate three-dimensional plane-wave excitation (PWE) images with object motion on which speckle tracking and feature tracking methods are applied and their efficacies are compared. The results show that PWE images result in greater tracking errors in lateral displacements when compared with two-way focused images. In addition, better tracking results can be obtained if the speckle tracking algorithm is implemented in polar coordinates. Furthermore, although the feature tracking method is more computationally efficient than the traditional speckle tracking method, its tracking error is relatively large. On the other hand, when applied on the rotated images, the feature tracking method and the speckle tracking method both have good tracking accuracy, but the error in the feature tracking case is still larger. In the feature tracking algorithm, the greater the threshold or the larger kernel size is set, the higher the tracking accuracy will be, but the number of feature patterns will also decrease. According to the results above, we believe that it is feasible to combine the plane-wave excitation imaging with feature tracking to constitute three-dimensional echocardiographic strain imaging. Finally, we apply feature tracking to clinical three-dimensional echocardiographic data of a three-month-old baby. Due to limited image quality, the feature patterns of endocardium and epicardium are not representative of the overall movement of the heart. Future researches will deal with the problems of the feature tracking method in clinical applications.

並列關鍵字

echocardiographic strain imaging ； speckle tracking ； feature tracking ； high frame rate imaging ； plane-wave excitation imaging

參考文獻

[2] A. A. Morsy and O. T. von Ramm, “FLASH Correlation: A New Method for 3-D Ultrasound Tissue Motion Tracking and Blood Velocity Estimation,” IEEE Trans. Ultrason., Ferroelect., Freq. Contr., vol. 46, no. 3, pp. 728-736,1999.

[3] G. R. Bashford, Student Member, IEEE and O. T. von Ramm, “Ultrasoud Three-Dimensional Velocity Measurements by Feature Tracking,” IEEE Trans. Ultrason., Ferroelect., Freq. Contr., vol. 43, no. 3, pp. 376-384, 1996.

[4] G. R. Bashford, Senior Member, IEEE and O. T. von Ramm, “Direct Comparison of Feature Tracking and Autocorrelation for Velocity Estimation,” IEEE Trans. Ultrason., Ferroelect., Freq. Contr., vol.54, no. 4, pp. 757-767, 2007.

[5] Johnny Kuo and O. T. von Ramm, “Three-Dimensional Motion Measurements Using Feature Tracking,” IEEE Trans. Ultrason., Ferroelect., Freq. Contr., vol.55, no. 4, pp. 800-810.

[6] R. G. Dantas, E. T. Costa and S. Leeman, “Ultrasound Speckle and Equivalent Scatterers ,“ Ultrasonics, vol. 43, pp. 405-420, 2005.