- 學位論文
基於二階方法的前列腺病理影像分類
The Grading of Prostatic Cancer in Biopsy Image based on Two Stages Approach
若您是本文的作者，可授權文章由華藝線上圖書館中協助推廣。
摘要
前列腺癌在臨床上的診斷方式，主要是利用組織的病理切片來進行癌症的分類。由於人類視覺的判斷會過於主觀，並不能清楚的分辨組織切片的微小差異，所以無法達到一個高度準確的分類結果。因此利用電腦來提供精確的、量化的數據，對於輔助醫師的診斷是很有幫助的。前列腺癌的等級分類是依據格里森分級系統(Gleason grading system)，本論文提出了二階段的方法，將實驗的分類結果對應格里森分級。第一階段將病理切片影像使用骨架集(Skeleton-set, SK-set)分群，相同群聚裡的影像會有相似的骨架特徵。第二階段，我們對每一個子頻帶(sub-band)作碎形維度分析。使用支援向量機(Support Vector Machines, SVM)當作分類器，正確率測試方法為Leaving-one-out。實驗內容包括50位病例，共1000張前列腺病理切片影像。初步實驗結果可以達到92%以上的正確率，來提供臨床醫師在前列腺癌診斷之輔助。
並列摘要
Prostatic biopsies provide the information for the determined diagnosis of prostatic cancer. Computer-aid investigation of biopsies can reduce the loading of pathologists and also inter- and intra-observer variability as well. In this thesis, we proposed a two stages approach for prostatic cancer grading according to Gleason grading system. The first stage classified biopsy images into clusters based on their Skeleton-set (SK-set), so that images in the same cluster consist of the similar two-tone texture. In the second stage, we analyzed the fractal dimension of sub-bands derived from the images of prostatic biopsies. We adopted the Support Vector Machines as the classifier and used the leaving-one-out approach to estimate error rate. The present experimental results demonstrated 92.1% of accuracy for 50 medical cases and a set of 1000 pathological prostatic biopsy images.
參考文獻
[1] [Online] Fractal Applications,
[2] [Online] Kohonen Network - Background Information,
[4] [Online] Principle of Support Vector Machines,
[7] G. L. Andriole, R.L. Grubb, and S.S. Buys et al.(2009), “Mortality Results from a Randomized Prostate-Cancer Screening Trial,” The NEW ENGLAND JOURNAL of MEDICINE, Vol. 360, pp.1310-1319.
[12] B. B. Chaudhuri and N. Sarkar(1995), “Texture Segmentation using Fractal Dimension,” IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 17, No. 1, pp. 72-77.