透過您的圖書館登入
IP:44.195.30.216
  • 學位論文

利用頻譜分析胃腸道系統影像之輔助診斷系統

A Quantify Image Analysis for Gastrointestinal System Using Spectrum

指導教授 : 蘇振隆

摘要


在診斷消化道相關疾病時,都是利用內視鏡來進行病灶的觀察採樣。但是在診斷時由於內視鏡探入消化道內的不適,以及病人的呼吸或是動作都會造成影像的模糊,本論文作出一套合適的影像分析系統對醫生的診斷作出更有利的幫助,同時減少使用內視鏡的時間,降低診療時程。 在臨床醫學診斷上,消化道內視鏡檢查仍以人類視覺之影像進行腸胃道相關疾病分析。本研究分析上消化道內視鏡影像在可見光上之頻譜分布,利用影像特徵分類。將內視鏡影像轉換為不同色彩空間,比較不同色彩表示法對於可見光頻譜的轉換效果。經由影像分析產生一組新的影像特徵分類時,大量無意義的特徵參數將會影響分類結果優劣,將所有的特徵參數進行一次數據正規化,再利用逐步區別分析將對影像分類有意義的特徵選出來,提升支持向量機的分類準確度。最後得到的分類結果可以提供醫師診斷時一個可信任的參考分類。 有效的原始影像共計54張。隨機取樣的方式將影像分成27張訓練組和27張測試組,每一組當中包含10張無疾病和17張有疾病的影像,經過特徵的篩選後進行分類,訓練組的部分正確率為100%,在測試組的分類正確率為92.6%,靈敏度為89.5%,專一性為100%,kappa值為0.92。另外將本分析運用在FICE所拍攝的其他頻道影像上,channel 2和channel 8所得出的測試組結果Kappa值分別為0.68和0.79,因此本系統在其他頻帶的分類結果上面雖然不如原始影像來的理想,但最後得到的分類結果都可以提供醫師診斷時可信任的參考分類。

並列摘要


In diagnosing of gastrointestinal related diseases, the endoscopy is always used to observe samples. However, during diagnosis, patient will feel discomfort when the endoscopy gets into the digestive. The quality of endoscopic images is also affected by patient's breathing and movement. In this paper, a suitable image analysis system was established for helping physicians to do diagnosis and reducing the time of using endoscopy. Image of gastrointestinal endoscopy is the most common material for analysis gastrointestinal related diseases, and the most diagnosis is done by human vision in clinical. This system analyzed the visible light spectral distribution of upper gastrointestinal endoscopy images, and did the classification based on the image features. The endoscopic images were transformed into different color spaces, and the transformed results of visible light spectral distribution between different color spaces were compared. After image analysis, the meaningless parameters affected the accuracy of classification. Therefore, all of the features were normalized, and then the best combination of significant features was chosen by Stepwise Discriminant Analysis to improve the prediction accuracy of Support Vector Machines. The results of this system could provide a convincing reference for the diagnosis of physicians. In this study, 54 of total valid original images were used. The images were divided into training and test groups, each group had 10 disease-free images and 17 images of disease. The classification results by significant features were 100% for training group, and 92.6% for test group. The sensitivity was 89.5%, the specificity was 100%, and the kappa value was 0.92. This analysis method was also used on the other FICE channels, such as channel 2 and channel 8, and the kappa value was 0.68 and 0.79 respectively. Although the results from the other channels were not good as from original images, the analysis results could still be a reliable reference for helping diagnosis to physicians.

參考文獻


[19] 胡祐莊,“以膠囊內視鏡影像的色彩與紋理特徵應用於潰瘍之偵測”, 中原大學醫學工程研究所碩士論文,民國99 年。
[13] Henry Fuchs , Mark A. Livingston , Ramesh Raskar , Andrei State , Jessica R. Crawford , Paul Rademacher , Samuel H. Drake , Anthony A. Meyer. Augmented reality visualization for laparoscopic surgery. Medical Image Computing and Computer-Assisted Interventation XMICCAI
[1] H. Machida, Y. Sano, Y. Hamaoto, M. Muto, T. Kozu, H.Tajiri, and S. Yoshida. Narrow-band imaging in the diagnosis of colorectal mucosal lesions: a pilot study. Endoscopy, 36(12):1094-1098. 2004.
[2] Sharma P, Bansal A, Mathur S, Wani S, Cherian R, McGregor D, Higbee A, Hall S, Weston A. The utility of a novel narrow band imaging endoscopy system in patients with Barrett's esophagus. Gastrointest Endosc. 64(2):167-175. 2006.
[3] Uraoka T, Saito Y, Ikematsu H, Yamamoto K, Sano Y. Sano’s capillary pattern classification for narrow-band imaging of early colorectal lesions. Dig Endosc. 23 Suppl 1:112-115. 2011 .

被引用紀錄


陳昱丞(2016)。以超解析技術輔助NBI內視鏡影像強化之探討〔碩士論文,中原大學〕。華藝線上圖書館。https://doi.org/10.6840%2fcycu201600825
潘昭延(2015)。高動態範圍技術於增強內視鏡影像之應用〔碩士論文,中原大學〕。華藝線上圖書館。https://doi.org/10.6840%2fcycu201500939
張孝慈(2015)。上消化道診斷之超解析度放大技術之開發〔碩士論文,中原大學〕。華藝線上圖書館。https://doi.org/10.6840%2fCYCU.2015.00142

延伸閱讀