標題: 非侵入式口腔癌檢測儀之開發
Development of noninvasive apparatus for oral cancer diagnosis
作者: 楊智翔
Yang,Chin-Siang
歐陽盟
Ou-Yang,Mang
電控工程研究所
關鍵字: 口腔癌;oral cancer
公開日期: 2012
摘要: 超頻譜影像應用於生醫檢測已有許多相關文獻以及臨床結果佐證,而目前用於癌症檢測的超頻譜影像裝置大多仍是將疑似病變的區域做切片,然後將樣品做處理後,置於顯微鏡觀察,再將影像資訊透過超頻譜影像儀器分析,此流程必須對疑似病灶點做活體組織切片。本文提出一種檢測法,利用光纖內視鏡對口腔組織做檢測,將影像資訊透過內視鏡傳送出,再透過超頻譜影像儀器做分析,不必對病人做活體切片,也提高檢測的便利性。 本研究開發出兩套檢測裝置,利用光纖內視鏡結合超頻譜影像系統, 在硬體上將目測用高解析度光纖內視鏡結合延遲透鏡式超頻譜影像系統(ERL-HSI),透過光纖傳送目標物影像資訊至超頻譜影像,有較少移動元件及穩定性高的優點。為了快速檢測,本研究提出另一種簡便的手持式裝置,改裝現有的手持顯微鏡,使之能使用選定光譜光源,照射目標物產生螢光反應,再將螢光影像透過特定光譜濾光片擷取所需波段影像,最後經影像感測器取得影像資訊。 本研究並對口腔癌檢體作實際量測,在光纖內視鏡結合超頻譜系統的實驗數據中,以演算法找出判斷口腔癌的光譜指標。達到了94%的靈敏性以及93%的特異性。而在手持系統的實驗數據,利用影像處理的方法處理圖檔,並以熵(Entropy)作為判斷標準,達到了84%的靈敏性及52%的特異性。
Many clinical results and literatures revealed that the hyper-spectral imaging was widely applied to biomedical diagnosis. Nowadays, the operative steps of hyper-spectral devices for cancer detection were slicing the region of suspected lesions, and the pathologist observed the image information through the analysis of hyper-spectral instruments. The process must slice the lesions and caused the pain of patients. This thesis proposes an in vivo diagnostic method that utilizes the fiber endoscopic to capture the image of the oral tissue and then the image information is transmitted to the hyper-spectral instrument through the endoscope. The research develops two sets of diagnostic systems. The high-resolution fiber endoscopy combined with Embedding Relay Lens Hyper-spectral Image system (ERL-HSI). The system can simultaneous acquire 400nm to 1000nm spectral information and spatial image for analysis. For the clinical fast diagnosis, this study proposes a simple handheld device. The fluorescent image is excited by the UV light source and then passes through the selected spectral filter. Finally, the sensor acquires the specific band image for diagnosis. This study eventually develops the algorithm for diagnosing the oral cancer with high accuracy. In the experimental data of the fiber endoscope combined with hyper-spectral system, we find the spectral index of oral cancer using for algorithm. It reached 94% sensitivity and 93% specificity.
URI: http://140.113.39.130/cdrfb3/record/nctu/#GT079812573
http://hdl.handle.net/11536/71560
Appears in Collections:Thesis