應用GPU加速於即時超音波影像與電腦斷層影像對位

醫學放射影像成為現在醫師對於術前的規劃及治療評估時不可或缺之重要工具，放射線治療已經成為醫師治療惡性腫瘤時常備之工具，而電腦斷層影像具有影像解析度佳的特性，現今的醫師皆是利用電腦斷層影像來找出腫瘤，並進行治療規劃。超音波能提供即時資訊，成本與風險也低於其它醫學成像技術。將其電腦斷層影像與超音波進行影像對位，能提供醫師即時與精確的影像資訊，輔助判斷理想治療範圍的一個工具，對提升放射線治療有相當的助益。本文進行超音波影像與電腦斷層影像定位研究。本研究所涵蓋的內容可概分為:1.影像前處理、2.特徵擷取、3.影像對位。由於超音波影像具有破碎、多雜訊的特性，利用中值濾波和非等向性擴散當做影像前處理，改善器官內部灰階值不均與破碎的情況。對於電腦斷層影像則透過區域成長法做影像切割，切割出目標器官的三維資訊，再依超音波探頭之方位重建出對應的電腦斷層影像切片。在特徵擷取的部分，超音波影像和斷層影像分別使用Sobel 和 Canny進行邊緣提取及邊緣梯度運算。影像定位部分，利用梯度向量積將兩張影像邊緣的梯度資訊加以進行比對，找出最相似之位置。最後使用CUDA平行化處理技術加速需即時運算的演算法執行速度。本實驗所用的標的物是由CIRS公司所提供的腹部假體以及透過馬偕醫院提供的超音波探頭和電腦斷層掃描影像做為測試資料。

關鍵字

超音波影像與電腦斷層影像定位；特徵擷取；梯度向量積； CUDA

並列摘要

Biomedical image play important roles in preoperative planning and treatment evaluation. The treatment evolution based on conformal radiation therapy (RT) techniques has been extensively accepted for treatment of malignancies. The CT provided excellent information for spatial resolution, and medical doctors use computer tomography to find out the tumors and recommend their treatment plan into practice. In contrast, ultrasound can provide real-time information in treatment room because it is more harmless than other medical techniques. Combining CT and ultrasound images to perform image registration will provide doctors with quicker and more accurate location of tumor and increase the effectiveness of radiation therapy. In this thesis, the proposed method is divided into three phases: (1)image pre-processing , (2)feature extraction, (3)registration. In first phase, due to the speckle noise and distortion in US images, this thesis adopted median filter and anisotropic diffusion to firstly process the US image. Then, the region growing method is utilized to segment the target organ in CT image. In second phase, the reconstruction slice of CT is extracted by any angle according to the pose of US probe, it helps us to confirm the CT and US images are in the same orientation. The image edge gradient information is further obtained by Canny operator and Sobel algorithm for both CT and US images. In last phase, the similarity measurement based on gradient direction is employed to evaluate the similarity between the CT and the US images. Finally, the registration experimental results in phantom, produced by CIRS Ltd., verify our proposed algorithm can get an efficient and accurate registration between CT and US images.

並列關鍵字

CT and US images ； Anisotropic diffusion ； Feature extraction ； Registration

參考文獻

[2] Y. Chen, Z. Wang, W. Zhao and X. Yang," Liver Segmentation from CT Images Based on Region Growing Method," iCBBE., art. no. 5163018, 2009.

[3] 鄭元愷，醫療影像軟體開發 – 三維醫療影像物件之擷取與處理，碩士論文，國立成功大學機械工程學系，台南，2003。

[4] Z. Hu," Extraction of Any Angle Virtual Slice on 3D CT Image," IITA , art. no. 4739594, 2008, pp. 356-360.

[6] C. Steger, “Similarity Measures for Occlusion, Clutter, and Illumination Invariant Object Recognition, “ Lecture Notes in Computer Science, Vol. 2191, 2001, pp. 148-154.

[7] P. Perona and J. Malik," Scale-space and Edge Detection using Anisotropic Diffusion," IEEE Trans. Pattern Anal. Machine Intell., vol.12, no. 7, July 1990, pp. 629–639,.

國際替代計量

應用GPU加速於即時超音波影像與電腦斷層影像對位

全文下載

主題瀏覽