標題: 環場掃描式膠囊內視鏡之設計與驗證
Implementation and verification of radial imaging capsule endoscope (RICE)
作者: 鄭偉德
Jeng, Wei-De
歐陽盟
Mang, Ou-Yang
電控工程研究所
關鍵字: 環場掃描式膠囊內視鏡;錐狀鏡;環場掃描;小腸出血;Radial imaging capsule endoscope;Cone mirror;Radial imaging;Small bowel bleeding
公開日期: 2015
摘要: 本研究提出新一代膠囊內視鏡『環場掃描式膠囊內視鏡』,利用環場掃描之特色來改善傳統前看式膠囊內視鏡之缺點。環場掃描式膠囊內視鏡透過利用一個錐狀鏡、環形透明觀景窗以及光學聚焦系統來實現環場掃描之特色。透過光學軟體模擬,有效改善弧史(Sagittal)以及子午(Tangential)平面之像差。最終完成之設計達到焦數為4.2,視角為65.08度,以及第四到第六視場之光點大小均小於17 um,與傳統前看式比起來光點大小改善達50 %,並且弧史平面之第五視場其MTF在空間頻率為100 lp/mm時,從原本7 %提升到 36 %。另外在照明系統之設計部分,由於環場掃描式膠囊內視鏡會受到雜散光影響降低照明均勻度。但透過本研究之優化設計後,例如改變LED之位置,切掉部分錐狀鏡之錐尖部分,甚至透過多層鍍膜,照明均勻度從原本12 %提升到69 %。在影像處理方面,本研究提出三個方法,意即平均絕對誤差、均方誤差、以及皮爾森相關係數法來進行影像相似度之比對,以進行影像之縫合。縫合結果可以發現皮爾森相關係數法之演算法最為強健,不會受到影像某部分之亮度有劇烈變化而受到影響。並且透過皮爾森相關係數法,其縫合之結果之原本影像之訊雜比高達80.69 dB。本研究甚至透過臨床活體實驗進行驗證,將所拍攝之小豬腸道內影像進行縫合,發現此演算法可運用於複雜之腸道環境。最後,針對環場掃描式膠囊內視鏡之色差部分,提出色彩校正之演算法,避免因為色差造成誤診。由於環場掃描式膠囊內視鏡之光學系統會造成色差,為了解決此問題本研究提出係數法與複數多項式映射法來將失真之色彩點映射到正確之色彩點。透過係數法與複數多項式映射法將原本高達21.45之色差(CIE DE2000)改善到1.53,透過複數多項式映射法可改善至1.32,此色差值已經低於人眼可辨識之色差1.5。
本研究提出環場掃描式膠囊內視鏡,透過全景拍攝之設計,可以輕易地擷取更多小腸內部影像資訊,避免傳統前看式膠囊內視鏡容易因視角有限造成死角問題。此設計對於小腸出血診斷非常重要,對於診斷率之提高有一定之幫助。
In this study, the radial imaging capsule endoscope (RICE) system was proposed. It differs from the conventional front imaging capsule endoscope (FICE) system. RICE uses a cone mirror, a radial window shell and a focus optical module to implement a radial imaging system. In minimizing the sagittal and tangential aberrations, the optical module of the RICE achieved an F-Number of 4.2, a viewing angle of 65.08°, and the spot sizes from 4th to 6th fields are less than 17 um. For the illumination system, this study proposed an optimal design to improve the uniformity of the illumination. The illumination uniformity in the RICE was increased from its original value of 0.128 up to 0.69. For the image processing, three algorithms, mean absolute error (MAE), mean square error (MSE), Pearson correlation coefficient, were used to stitch the images together. The Pearson correlation coefficient method was the most effective algorithm because it yields the highest peak signal to noise ratio (PSNR) of higher than 80.69 dB compare to original image. Furthermore, this algorism was verified by a clinical animal experiment. Finally, color calibration was carried out. Before color calibration, the color difference in RICE was 21.45 (CIE DE2000). Through the proposed conformal mapping, the color difference was substantially reduced to 1.32, and the color difference is imperceptible for human eye because it is lower than 1.5.
In summary, this study developed a prototype of RICE, which yielded panoramic view, largely increases the capability to access the small bowel without losing any information. This is very important for diagnosis of obscure small bowel bleeding, hence, improve the diagnostic yield.
URI: http://140.113.39.130/cdrfb3/record/nctu/#GT079912812
http://hdl.handle.net/11536/126565
Appears in Collections:Thesis