智慧型演算法用於乳房病變之超音波影像辨識

Intelligent-Based Algorithm in Classifying Breast Nodules of Ultrasound Image

侯玉翎(Yu-Ling Hou)

指導教授 ： 林志民

元智大學/電機通訊學院/電機工程學系/碩士(2007年)

https://doi.org/10.6838/YZU.2007.00005

摘要

好的電腦輔助診斷系統可以幫助經驗不足的醫生避免誤判和在不漏掉惡性腫瘤的情形下減少良性組織的切片檢查。已經有很多用在醫學影像上的電腦輔助診斷系統被提出。然而，仍有很多人對這些系統抱著不信任的態度。無論有多少人感到懷疑，電腦補助診斷系統仍然持續發展中，並且被應用在很多臨床上的醫學影像領域。在本論文中，我們將針對超音波電腦補助診斷系統進行研究。我們致力於找出一個高效率的分類器。我們比較類神經網路和小腦模型的效率，並且提出結合類神經網路和小腦模型的網路架構。最後，可以發現多階層網路效率優於類神經網路與小腦模型。
最後，因為在女性癌症死亡原因中，乳癌的排名逐漸往前提升，因此一套用來幫助醫生能夠早期偵測癌症的電腦補助診斷系統變得越來越重要。最後的實驗結果顯示，本研究所使用的這些腫瘤特徵與分類架構，能有效地應用在超音波影像的腫瘤判斷上，且得到不錯的乳房腫瘤診斷之正確率可達96%，高於醫師人為判斷的正確率(約88%)，這可幫助臨床醫師使之能做出正確的判斷，更能節省醫師診斷所花費的時間，以達到早期發現早期治療的目標。

關鍵字

超音波影像，斑點雜訊，類神經網路，小腦模型

並列摘要

A well-designed computer-aided diagnosis (CAD) system can assist physicians to avoid misdiagnosis and reduce the number of benign lesion biopsies with missing cancers. There have been many successful applications of CAD strategy reported for ultrasound image. However, many people were still skeptical about these CAD systems in the past. No matter how skeptical these people may have been, CAD is still undergoing great development and utilization within the field of medical imaging as its potentials are well demonstrated in many clinic areas. In this thesis, we will explore the ultrasound CAD system and aim for a high performance classifier with intelligent algorithm. We will compare the performance of the neural network (NN) and cerebellar model articulation computer (CMAC). We also merge NN and CMAC to a new structure called intelligent hierarchical network (HN). The new intelligent HN understands the advantages of NN and CMAC. Finally, the performance of HN is proven to be superior to NN and CMAC.
Lastly, since the mortality rate of breast cancer in women is gradually increasing, CAD system helps in the early diagnosis of cancer by doctors using CAD system. The simulations demonstrate that the proposed CAD system can differentiate breast nodules with relatively high accuracy (96%); it is higher than the operator’s diagnosis (88%).So that it can help operators to avoid misdiagnosis.

並列關鍵字

ultrasound (US) image, speckle, neural network (NN), cerebellar model articulation computer (CMAC)

參考文獻

[2] S. Joo, Y. S. Yang, W. K. Moon and H. C. Kim, ‘‘Computer-aided diagnosis of solid breast nodules: use of artificial neural network based on multiple sonographic features, ’’ IEEE Trans. on Med. Imag., vol. 23, no. 10, pp. 1292–1300, 2004.

[3] J. W. Goodman, “Some fundamental properties of speckle,” J. Opt. Soc. Amer. ,vol. 66, no. 11, pp. 1145–1150, 1976.

[4] C. B. Burckhardt, “Speckle in ultrasound B-mode scans,” IEEE Trans. on Sonics Ultrason., vol. 25, no. 1, pp. 1–6, 1978.

[6] J. G. Abbott and F. L. Thurstone, “Acoustic speckle: Theory and experimental analysis,” Ultrason. Imag., vol. 1, no. 4, pp. 303–324, 1979.

[7] J. G. Thomas, R. A. Peters, and P. Jeanty, “Automatic segmentation of ultrasound images using morphological operatots,” IEEE Trans. on Med. Imag., vol. 10, no. 2, pp. 180–186, 1991.