智慧型演算法於超音波影像輪廓圈選

Intelligent-Based Algorithm in Contour Extraction of Ultrasound Image

范偉哲(Wei-Che Fan)

指導教授 ： 林志民

共同指導教授 ： 葉秩光(Chih-Kuang Yeh)

元智大學/工程學院/電機工程學系/碩士(2006年)

https://doi.org/10.6838/YZU.2006.00278

摘要

自動化偵測腫瘤及萃取其邊界是一項具有挑戰性的任務，這是由於超音波影像上不規則的腫瘤形體以及附含有斑點雜訊的干擾所導致。在本研究中，我們提出適應性模糊邏輯濾波器演算法及支持向量機器演算法。
在適應性模糊邏輯濾波器演算法當中，適應性模糊邏輯濾波器能有效的使用於降低雜訊以及強化腫瘤邊緣特徵。而在支持向量機器演算法中將針對低對比以及微小的腫瘤情況來做分類處理，其對於影像上各自的像素點利用支持向量模型分類為腫瘤或者是斑點雜訊。接著，將使用影像處理技術如適應性二值化、通道分量標籤化以及圓盤擴張法去萃取有意義的腫瘤。實驗結果顯示經由設計的原始輪廓以及經由本論文所提出的演算法圈選後的輪廓，正確的重疊面積能夠高於92%。而在臨床影像上，結果會非常接近於有經驗的臨床醫師手動圈選結果。此外這兩種演算法基於反覆的操作設計，能夠在單張影像上成功的萃取多個腫瘤。

關鍵字

超音波影像 ； 斑點雜訊 ； 適應性模糊邏輯濾波器 ； 連通分量 ； 支持向量機器 ； 圓盤擴張法

並列摘要

Automatically detecting tumors and extracting lesion boundaries in ultrasound images are challenging tasks due to the variance in shape and the interference from speckle noise. In this study, we present an adaptive-fuzzy-logic-filter based algorithm (AFLFBA) and a support-vector-machine based algorithm (SVMBA).
In AFLFBA, the adaptive fuzzy logic filter (AFLF) was utilized to suppress speckle noise and enhance the lesion boundary features effectively. The SVMBA was proposed to process the low-contrast and also tiny lesions. Each pixel within an ultrasound image is classified as being either a lesion or speckle noise according to the SVM model. In the following, we can use image processing techniques such as adaptive thresholding, connected component, and disk expansion to extract the significant lesions. The experimental results show that the average of the true positive area overlap between the designed contour and the contour obtained by the AFLFBA and SVMBA are higher than 92%. The analyzed results of the clinical images show that the extracted lesions contour were close to the experienced clinician's manual delineation. Furthermore, based on the iterative design, these algorithms can be performed for multiple lesion extraction in a single image successfully.

並列關鍵字

Ultrasound image ； speckle noise ； adaptive fuzzy logic filter ； connected component ； support vector machine ； disk expansion

參考文獻

[3] A. Madabhushi and D. N. Metaxas, ‘‘Combining low-, high-level and empirical domain knowledge for automated segmentation of ultrasonic breast lesions,’’ IEEE Trans. Med. Imag., vol. 22, no. 2, pp. 155–169, 2003.

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

[5] C. Kotropoulos, X. Magnisalis, I. Pitas, and M. G. Strintzis, “Nonlinear ultrasonic image processing based on signal-adaptive filters and self-organizing neural networks,” IEEE Trans. Image Processing, vol. 3, no. 1, pp. 65–77, 1994.

[6] J. Feng, W. Lin, and C. Chen, “Epicardial boundary detection using fuzzy reasoning,” IEEE Trans. Med. Imag., vol. 10, no. 2, pp. 187–199, 1991.

[7] S. K. Satarehdan, and J. J. Soraghan, “Automatic cardiac LV boundary detection and tracking using hybrid fuzzy temporal and fuzzy multiscale edge detection,” IEEE Trans. Biomed. Eng., vol. 46, no. 11, pp. 1364–1378, 1999.