先進體積型醫學影像壓縮系統之研究

醫學影像是醫生經常使用來判斷病症的一個重要依據，尤其是三維的體積型(Volumetric)醫學影像，能一次呈現受檢器官組織的立體構造，使醫生對於複雜的人體結構、鄰近器官的相關位置與病變組織的體積能有更準確的診斷與判定，因此體積型醫學影像在醫學上確實有很大的需求與診斷價值。體積型醫學影像資料量相當龐大。在遠距醫療(Telemedicine)的應用環境中，在有限頻寬的網路中欲傳輸如此龐大的醫療資訊，將會是一種相當大的負擔。因此本論文的目的在於發展一套適合體積型醫學影像且具有高壓縮效能的演算法，希望能減少醫學影像的資料儲存量。因為體積型醫學影像其鄰近單張影像間具有高度相似性，所以本論文提出一個以離散小波轉換(DWT)與向量量化(VQ)為基礎的體積型醫學影像壓縮演算法，希望藉由WT-VQ來充分運用影像序列中的冗餘性，以達到壓縮的目的。有別於傳統WT-VQ演算法的是，在只需設計單一碼簿的情況下，本文提出一個特殊且新穎的樹狀向量與樹狀碼簿架構來增加向量量化器的效能。為了使重建的醫學影像保有一定的品質與診斷特徵，因此我們在VQ中加入Distortion Constrained Codevector Replenishment (DCCR)機制，使失真量限制於一定的範圍之內。由實驗結果得知，在碼簿大小為1024且DCCR的失真臨界值為500時，重建的醫學影像不但在視覺品質上保有一定的診斷特徵，而且平均壓縮率101倍，單張影像壓縮率更可高達130倍。

關鍵字

向量量化；遠距醫療；離散小波轉換；體積型醫學影像； DCCR機制

並列摘要

Medical Images play important roles for doctors to diagnose diseases. In particular, the 3-D volumetric medical images can demonstrate 3-D structures of inspected organs at one time, and let doctors make a better observation or an accurate diagnosis on complex human body structures, related position of adjacent organs and volume of bad tissues. Thus, volumetric medical images indeed have great requirements and values in medical treatment. The data of volumetric medical images are quite enormous. In the application of telemedicine, it will be a heavy loading to transmit such enormous medical information in limited bandwidth. The objective of this thesis is to propose an excellent compression algorithm suitable for volumetric medical images in order to reduce the amount of these data. Since the volumetric medical images have high similarity between two adjacent pictures, we propose a compression method based on discrete wavelet transform (DWT) and vector quantization (VQ) to exploit this similarity. With this approach, we can exploit the redundancy among images in order to implement data compression. Since we only need to design one single codebook, our method is different from traditional WT-VQ, where multiple codebooks are usually required. The proposed codebook consists of a novel tree vector structure that can be used to enhance the efficiency of VQ. We include the distortion constrained codevector replenishment (DCCR) mechanism in VQ to limit our quantization errors and maintain acceptable quality and available diagnosis features for reconstructed images. Our experimental results show that when the codebook size is 1024 and the threshold of DCCR mechanism is 500, the reconstructed images not only have clear visible diagnosis features but also excellent compression ratio (CR). The average CR for an image sequence is 101 and the CR can even reach 130 for a particular image in the image sequence.

並列關鍵字

Telemedicine ； Discrete wavelet transform ； Distortion constrained codevector replenishment ； Vector quantization ； Volumetric medical image

參考文獻

[1] Y. Kim and W. A. Pearlman, “Stripe-Based SPIHT Lossy Compression of Volumetric Medical Images for Low Memory Usage and Uniform Reconstruction Quality,” Proc. of Int. Conf. on Image Processing, Vol. 3, pp. 652-655, 2000.

[2] J. H. Hu, Y. Wang, and P. T. Cahill, “Mutispectral Code Excited Linear Prediction Coding and Its Application in Magnetic Resonance Images, “ IEEE Trans. Image Processing., Vol. 6, No. 11, pp. 1555-1566, Nov. 1997.

[3] M. Midtvik and I. Hovig, “Reversible Compression of MR Images,” IEEE Trans. Med. Imag., Vol. 18, No. 9, pp. 795- 800, Sept. 1999.

[4] K. Karadimitriou and J. M. Tyler, “Centroid Method for Compressing Sets of Similar Images,” Pattern Recognition Letters, Vol. 19, no. 7, pp. 585-593, 1998.

[5] S. H. Tai, Y. G. Wu, and C. W. Lin, “An adaptive 3-D Discrete Cosine Transform Coder for Medical Image Compression,” IEEE Trans. Inform. Technol. Biomed., Vol. 4, No. 3, pp. 259- 263, Sept. 2000

國際替代計量

先進體積型醫學影像壓縮系統之研究

未授權

主題瀏覽