孔徑信號處理技術與其在超音波影像之應用

Abstract

本論文主要研究超音波陣列系統的孔徑信號(aperture domain data)相關之成像方法與信號處理技術。一般系統廣泛運用的波束形成技術為延遲-加總法，此方法可藉由調整各頻道的時間延遲和權重而使得聲波波束可以電子式聚焦到特定深度及可以任意地調控與位移波束，並調整橫向解析度及波束形狀而使得掃描深度內皆為動態聚焦。在波束加總前的各頻道接收信號亦稱為孔徑信號。在傳統的系統裡，孔徑信號會因為資料量過於龐大而在波束加總後被捨棄，但延遲-加總法僅能獲得波束方向上的空間資訊反而使得超音波影像在臨床上的應用受到限制。本論文中將探討向量流速估計與相位偏移校正方法等兩種孔徑信號處理技術¬。 在第一部份流速估計的應用上，使用所提出的使用孔徑信號之二維流速估測方式來改善傳統的流速估計方法只能量測平行波束方向的流速分量。在這方法中一個沿著陣列方向之時間偏移量變曲線被建立並近似成為一次多項式來求得軸向與橫向速度分量。我們經由模擬和實驗來驗證方法的可行性，結果顯示所提出的方法能改善向量流速估計之誤差且其結果比傳統流速估計法來的更佳，並且可以實現即時二維血流量測。 在相位偏移校正的應用上，我們使用臨床乳房影像驗證一基於接收孔徑信號之同調性的旁瓣抑制法。傳統灰階超音波在乳房病灶偵測上常會因為對比解析度不足而使其效果被限制。由臨床實驗結果顯示相較於傳統基於相關性之方法，所提出的權重方法能顯著的改善乳房影像品質。 在本論文的第三部分，基於同調性的旁瓣抑制法也被延伸應用到高速超音波影像上，在這方法中使用一高精確度之Capon估計法來量測同調能量，並使用僅八次平面波激發與合成發射孔徑方法來達到高速成像。模擬和實驗結果皆顯示所提出之方法都能對對比度與病灶清晰度等影像品質有所改善。結果顯示這些基於同調性的方法能有效改善臨床上的病灶偵測，因我們所提出之方法無須任何聚焦誤差的假設便能有效降低旁瓣貢獻。本論文開發多種影像方法並有效提升流速與對比解析度有助於提升臨床診斷。論文最後亦將探討相關技術之延伸應用。

The purpose of this dissertation is to investigate various processing techniques for ultrasound image formation and signal processing based on aperture domain data for ultrasonic system using arrays. Conventionally, an array system utilizes the widely adopted delay-and-sum method to focus acoustic beams electrically at specific depths with arbitrary steering and shifting by the delay and weighting of each array element. This method can adjust lateral resolution and beam-shapes and therefore provides dynamic focusing throughout the scan depth. The data recorded from individual array channels prior to beam summation are referred to aperture domain data and are often discarded after beam summation due to a large data size. However, the delay-and-sum method only preserves the spatial information along the beam direction and therefore limits the clinical applications. In this thesis, two specific tasks of aperture domain data processing including vector velocity estimation and phase-aberration (i.e., focusing errors resulting from sound-velocity inhomogeneities) correction are investigated. The first topic in this dissertation is the vector flow estimation. A conventional scanner can only estimate the flow velocity parallel to the beam axis. The proposed flow estimation technique uses aperture domain data for 2D flow-velocity estimation. A time-shift profile along the array direction is constructed and approximated by a first-order polynomial to determine the axial and lateral velocity components. The efficacy of the vector velocity estimation method is verified by simulations and experiments. The results demonstrate that the accuracy of the proposed method is comparable to existing vector velocity estimation method and real-time two-dimensional velocity vector estimation is feasible. For phase-aberration correction, a sidelobe-reduction technique based on the coherence of the receive aperture domain data is tested with clinical breast data. The performance in lesion detection using B-mode ultrasound is often limited by poor contrast resolution. Experimental results demonstrate that the proposed weighting method is feasible in breast imaging and rivals the conventional correlation-based method with significant image quality improvement. In the third part of the dissertation, the coherence-based sidelobe-reduction technique is also extended to high-frame-rate adaptive imaging with a high accuracy Capon estimator to estimate the coherent energy. The high frame rate image is formed using plane-wave excitation and a synthetic transmit aperture method using only 8 firings. Significant improvement in contrast and lesion definition is demonstrated through the simulations and breast imaging experiments. The results demonstrate that these coherence-based methods are feasible to improve lesion detection in clinics since these techniques can effectively reduce sidelobe contributions without any assumption regarding the source of focusing errors. In summary, advanced imaging techniques were developed in this thesis to improve velocity and contrast resolution and thus increase diagnostic confidence in clinics. Potential extended application of these methods will also be described.