| 中文摘要 |
免疫風濕科在檢查自身免疫疾病的主要依據是抗核抗體 （antinuclearautoantibodies, ANA），醫師會藉由人類表皮細胞癌 （human epithelioma type 2,HEp-2） 之細胞培養，透過免疫螢光顯影技術來辨識ANA，以預測病人的疾病。目前辨別樣式是仰賴專家觀察螢光顯微鏡下的載玻片來完成，此方式需要具有高度專業之技術人員來操作且耗時，因此自動化分析免疫螢光顯影影像有其必要性。目前研究自動化分類ANA 螢光影像流程的技術眾多，本研究從眾多方法中找出分類結果正確率較高之組合方法，進而協助醫師進行疾病之診斷。實驗結果顯示使用Canny邊緣檢測，經紋理分析得到特徵，由支援向量機遞迴特徵消去法進行特徵選擇，最後結合SVM 分類方法具有最高的正確率（97.00%）。 |
| 英文摘要 |
Experts of immunology and rheumatology department inspect autoimmunedisease by recognizing antinuclear autoantibody (ANA) patterns.Physicians diagnose patients’ disease by inspecting HEp-2 cells foridentification of ANA. Up to now, identification of ANA is completed byinspecting slides using a fluorescent microscope. This manual procedurerequires highly specialized technicians and is very time-consuming.Therefore, it is necessary to have methods for classifying ANA imagesautomatically. This research attempts to find the cascaded method withthe highest accuracy from among different processing methods. Thesemethods include edge detection, feature selection, and classification. Theexperimental results show that the optimal cascade methods are the com- bination of Canny and Support Vector Machines (SVM). The accuracyrate is about 97.00%. |
本站僅提供期刊文獻檢索。
