目前醫師正在嘗試著利用小量的鼻咽組織來進行診斷分類,由於人類視覺的判斷會過於主觀,並不能清楚了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方式。因此利用電腦來提供一個精確的、量化的數據對於醫師的診斷是很有幫助的。除此之外,醫師可以藉由電腦輔助診斷系統先篩選出鼻咽癌的高危險以便對後續的治療與追蹤將助益良多。在鼻咽癌組織切片的診斷裡可以分成4類,分別是正常組織 Normal、鱗狀組織 Squamous Metaphases(SM)、小細胞癌 Small Cell Carcinoma (SCC)、未分化癌 Undifferentiated Carcinoma (UnCa)。在這個系統中我們將利用Snake Model將細胞核切割出來,再依據醫師所給的各項標準後,依據4個分類的特性取出特徵,並建立起分類樹,以此對應到前述的4個分類。我們的初步實驗結果可以達到97.3%的正確性,來提供臨床醫師在鼻咽癌診斷之輔助。
Currently doctors are trying to use a small amount of tissue for diagnostic classification of nasopharyngeal, as the human visual judgments are too subjective, can not clearly distinguish small differences in tissue biopsy. Therefore it is necessary to provide a quantitative data for a higher accuracy of the diagnosis. The diagnosis of nasopharyngeal carcinoma in tissue section can be divided into four categories, Normal, Squamous Metaphases (SM), Small Cell Carcinoma (SCC), and Undifferentiated Carcinoma (UnCa). In this system we adopted Snake Model for the segmentation of the nucleus, and then extracted features for the classification, according to standards given by the physician. We then established a decision tree for the classification. Our experimental results show that our method can achieve 97.3% of accuracy.