利用3-D彩色超音波的乳房腫瘤化學治療反應評估

Abstract

乳房X光片、超音波、以及磁振造影等技術，常被用來檢視乳癌對於化學治療的反應效果。在本篇論文中，是採用三維的都普勒(Doppler)超音波影像來評估血管對於化學治療的反應。傳統的血管評估方法為血管指標(vascularity index, VI)，但是此方法在本實驗中無法為血管影像提供良好的再現性(reproducibility)，這主要是由於血管影像會因為病人的心跳以及呼吸等影響，而在影像中產生雜訊。因此，我們在本篇論文中提出了一個基於血管長度的評估方法，此方法可以去除影像中不必要的雜訊。首先，對於血管影像使用三維的細化演算法，以取出血管的主要骨架，並且計算血管的長度。為了要去除雜訊造成的影響，我們提出一個新的雜訊移除演算法。再此雜訊移除演算法中，我們先定出一個半徑門檻值。對於所有的血管，只有半徑大於此門檻值的才會計算其長度。在相對值雜訊移除演算法中，百分之二十和百分之十所得到的再現性分別為0.075和0.078。而在固定值演算法中，五像素和三像素所得到的結果分別為0.119和0.097。相對值演算法會比固定值演算法得到更好的結果。另外，我們則是利用血管長度的變化量，來評估一個病人對於接受化學治療後的反應。我們發現，利用百分之二十所得到的結果比百分之十的結果還要差。因此我們認為，百分之十的相對值演算法比較適合對化學治療的反應作評估。

Several imaging modalities such as mammography, ultrasound (US), and magnetic resonance imaging (MRI) have been applied to evaluating the therapeutic response of neoadjuvant therapy for breast cancer. In this paper, the three-dimensional (3-D) Doppler US is adopted for evaluating the vascularity. The conventional vascularity index (VI) method is usually used to evaluate the vascularity. From our experiments, this method could not have enough reproducibility for monitoring the chemotherapy response due to the artificial vessel noise by breath and heartbeat. Hence, a new vascularity evaluation method based on the vessel length is proposed to reduce the vessel noise in this paper. At first, the 3-D thinning algorithm for the vessel image is performed to extract the vessel skeleton and then the vessel lengths could be found from the skeleton as the vascularity. Nevertheless, this proposed vessel length information still does not have enough reproducibility due to the vessel noise. In order to improve the vessel reproducibility, a vessel noise reduction method is developed to analyze the vessel lengths. In the proposed noise reduction method, only the lengths of vessels larger than a threshold are counted. The threshold could be a fixed value or proportion to the maximum vessel radius. In the relative noise reduction method, the average reproducibilities of 20% and 10% reductions are 0.075 and 0.078 in our cases. The reproducibility of fixed threshold reduction by 5-pixel and 3-pixel are 0.119 and 0.097 respectively. The relative threshold reductions have better performance than the fixed threshold reductions. Furthermore, the response during the chemotherapy for a patient could also be evaluated by the changes of vessel lengths. The therapy evaluation result of 20% reduction is worse than that of 10% reduction. Therefore, we conclude that the 10% noise reduction method is a better evaluation method for the response of chemotherapy.