模糊分類在醫學影像分割上的應用

Application of Fuzzy Clustering Algorithms in Medical Image Segmentation

胡裕仁(Yu-Jen Hu)

指導教授 ： 楊敏生

中原大學/理學院/數學研究所/碩士(2001年)

https://doi.org/10.6840/cycu200100004

摘要

本文乃是希望從FCM及AFCM兩種分類演算法的應用，進而修正MRI醫學影像的模糊扭曲為診斷時重要的參考，達到預防勝於治療的效果。在臨床醫學上，核磁共振造影（Magenetic Resonance Image、簡稱為MRI）技術可以將人體各個部位組織，規律的製成組織切片造影，但是在攝製過程中往往由於組織膊動等因素造成MRI造影有模糊扭曲或有假影等雜訊發生，使得醫師無法作出準確的判斷，因此如果能事先將受測者可疑的病變組織照片處理、除去雜訊，再經由專科醫生來判讀，更能順利找出病因、增進檢查的時效性及可靠性。
傳統上分割MRI影像資料時，通常是利用監督式(Supervised)演算法也就是說將預先定義訓練好的資料，再利用一個區別函數，將其應用在所要分析的影像資料上，但是此法受限於預先定義好的圖樣有限，因此若拿來套用在所有影像資料可能無法獲得最佳的結果，因此本文使用非監督式分類演算法應用本文以眼部MRI影像為例，利用非監督式演算法將MRI影像造影重現，以作為眼科醫生偵測腫瘤及組織異常之輔助，本研究除使用常用的FCM分類演算法外，我們也提出使用一套新建立的分類演算法，稱為相對式模糊C均值(Alternative Fuzzy C－Means、簡稱為AFCM)分類法，經由實驗比較後我們發現AFCM有良好結果因此AFCM可被推薦來作為影像分割的一項新工具。

關鍵字

相對式模糊C均值 ； 核磁共振造影 ； 模糊C均值 ； 影像分割

並列摘要

In the clinic field, MRI medical image uncertainty is widely present in data. In particular, borders between tissues are not exactly defined and memberships in the boundary regions are intrinsically fuzzy. Therefore, computer assisted unsupervised fuzzy clustering methods turn out to be particularly suitable for handleing a decision-making process concerning segmentation of magnetic resonance images. In this paper we focus our attention on the Alternative Fuzzy Clustering Mean （AFCM） method and the resolutions of MRI segmentation uncertainty in Ophthalmology .
Cluster analysis is a tool for clustering a data set into groups of similar individuals. Image segmentation is a method to partition image pixel into similar regions. Thus, clustering algorithms would naturally be applied to enhance images in the areas of segmentation. Recently, this Fuzzy C-Means (FCM) algorithm has been more frequently used in segmenting MRI.
In the clinic oncological field, physicians depend on different clinical frameworks, different anatomical evidences, and different theoretical approaches, etc. to diagnose a patient. It is often impossible to establish rule-based systems. MRI or computer-assisted approaches may be particularly helpful in the clinical oncological field as support in the diagnosis of Retinoblastoma, an inborn onclogical disease in Ophthalmology, which usually shows the symptoms in the early childhood. For the purpose of early treatment with radiotherapy and surgery, FCM clustering algorithm should be introduced to the diagnosis of every Retinoblastoma patient. Recently, Wu and Yang proposed a new algorithm AFCM which provides more information available in medical images.
This literature on MRI segmentation techniques has paid particular attention to the differentiation of abnormal and normal tissues in Ophthalmology with the use of AFCM clustering algorithms to help reduce the medical image noise effects originating from low resolution of sensors or the structure that moves during the acquisition of data. Therefore, the unsupervised segmentation algorithms may be particularly helpful in the clinical oncological field as an aid to the diagnosis of Retinoblastoma. A newly proposed algorithm AFCM is introduced to provide more information for medical images used by Ophthalmologists. Both AFCM and FCM segmentation techniques provide useful information and good results. However, the AFCM method has better detection of abnormal tissues then that of the FCM according to a window selection. Overall the new proposed AFCM segmentation technique is recommended in the segmentation of MRI.

並列關鍵字

Alternative Fuzzy Clustering Mean ； Image Segmentation ； Fuzzy Clustering Mean ； Magenetic Resonance Image

參考文獻

[1] F. Masulli and A. Schenone, A fuzzy clustering based segmentation system as support to diagnosis in medical imaging, Artificial Intelligence in Medicine, Vol. 16, pp.129-127, 1999.

[2] J. C. Bezdek, Pattern Recognition With Fuzzy Objective Function Algorithms, Plenum, New York, 1981.

[3] J. C. Bezdek, R. Ehrlich, W. Full, The fuzzy c-means clustering algoritnm, Computers and Geosciences Vol. 10, No. 2-3, pp.191-203, 1984.

[4] J. Dunn A fuzzy relative of the ISODATA process and its use in detecting compact well-separated clusters, Cybernetics Vol.3, pp.32-57, 1974.

[5] J. Suckling, T. Sigmundsson, K. Greenwood and E. T. Bullion, A modified fuzzy algorithm for operator independent brain tissue classification of dual echo magnetic resonance images, Magnetic Resonance Imaging Vol. 17, No.7 pp.1065-1076, 1999.