基於血管分析的3D彩色超音波之乳房腫瘤化學治療評估

Abstract

腫瘤血管是腫瘤生長、惡化以及轉移的一個重要因子。一般而言，腫瘤需要更多的血管以獲取足夠的養分供其生長與存活。近年來，都卜勒 (Doppler) 超音波成像技術已經成功的應用於血流的偵測，對於腫瘤血管的相關研究也在乳房病變的診斷以及前置性輔助治療成效的評估中扮演了重要的角色。本篇論文中採用了三維 (3-D) 都卜勒超音波成像技術來評估血管資訊。傳統的血管指標 (vascularity index, VI) 評估方法已經廣泛的應用於血管資訊的評估。然而，肇因於化學治療前後所掃描的3-D影像間無可避免的位移誤差，這個方法無法在監測化學治療結果時提供足夠的再現性。因此，本篇論文提出了一個新的血管資訊評估方法以降低影像位移所造成的誤差。首先，先對血管影像採用雜訊移除演算法以移除因心臟收縮、呼吸以及身體移動所造成的雜訊。接著，利用一個固定的門檻值將較不顯著的血管移除並同時保留主要的血管。為了修正進行評估的兩個3-D影像間的位移誤差，影像間先利用影像套合的技術進行套合。在套合兩個影像之後，兩個影像之間的交集區域就可以被擷取出來並計算這個交集區域中血管密度的變化率以作為監測指標。在我們的實驗中，利用血管密度來分類有反應與沒有反應的準確率是100% (9/9)。此結果與傳統利用腫瘤尺寸變化率作為判斷基準的結果一致。此外，利用化學治療三週後的血管密度所作的早期預測準確率則是達到了89% (8/9)。

The tumor vascularity is a critical factor for growth, invasion, and metastasis. Generally, tumors need more blood vessels to obtain sufficient nutrients for growing and surviving. Recently, the Doppler ultrasound (US) is applied for detecting the blood flow successfully, and the studies of tumor vascularity have played important roles to diagnose diseases of breast and evaluate the therapeutic response of neoadjuvant therapy. In this paper, the three-dimensional (3-D) power Doppler US is adopted for evaluating the vascularity. The conventional vascularity index (VI) method has been broadly used to evaluate the vascularity. However, this method could not provide enough reproducibility for monitoring the chemotherapy response due to the unavoidable displacement between the 3-D volumes scanned before and after the chemotherapy. Hence, a new vascularity evaluation method is proposed to reduce the inaccuracy caused by volume misalignment in this paper. First, the noise reduction algorithm for the vessel image is performed to reduce the noise caused by the systole, breath, and body movement. Second, a predefined threshold value is applied to filter out the less significant vessels and main vessels can be preserved. To revise the misalignment between two 3-D volumes for evaluation, volumes are registered by the image registration technique. After registering the two volumes, the intersection of these two volumes is found and the change ratio of vessel densities in the intersected volume is calculated as the monitor index. In our experiment, the correctness of classifying responder and non-responder by using the vessel density is 100% (9/9). This result is consisted with the conventional evaluation method based on the change of tumor sizes. Furthermore, the accuracy of early prediction by evaluating the vessel densities calculated about 3 weeks after the first chemotherapy is up to 89% (8/9).