透過您的圖書館登入
IP:3.237.44.242
  • 學位論文

自動化盲腸辨識系統

Auto-Recognition System of Cecum

指導教授 : 陳中平

摘要


在此篇論文中,我們提出了一個可以自動辨識盲腸的系統。此系統是以盲腸的特徵來當作分類器,對於輸入的照片盡可能地給予正確判斷。 隨著經濟的繁榮、科技的進步以及飲食西化,消化系統的疾病日益增多,其中又以大腸癌的成長率為最。自民國96年起,大腸癌已成為國人癌症發生人數第一位,每年有超過一萬人被診斷罹患大腸癌。健保署公布民國103年前十名癌症醫療支出排行榜,第一名便是大腸癌,足可見大腸癌之重要性。第0、1期的大腸癌經過治療後,5年存活率幾乎可以達到9成以上,可是大腸癌早期並無症狀,所以定期的檢查可以減少我們罹患大腸癌的風險。 除了糞便淺血檢查,人們還可以利用大腸鏡篩檢來做進一步的檢測。由此可知大腸鏡品質攸關篩檢的整體效益,因此必須確保鏡檢的品質。如何使檢查的流程更加順利且品質更完善,需要國家政策與實務的共同合作。本論文著重於大腸鏡品質中的盲腸到達率(CIR)。經由研究發現,盲腸到達率高的醫生其病人得到大腸癌的機率相對低,可見盲腸到達率是非常重要的數據。因此,我們提出了盲腸辨識的演算法,利用電腦的分析和判斷,來檢視每個大腸鏡篩檢後的照片,藉此計算每個醫生的盲腸到達率,以達到第三方公正的判斷。我們以自己建立的系統測試了醫生給的650張盲腸與非盲腸的照片,實驗結果顯示,我們提出的演算法確實能做一個初步的判讀並且通過了新的盲腸照片之測試。 關鍵字:盲腸、大腸癌、盲腸到達率、影像辨識、影像分割、交叉驗證、直線偵測、分類器

並列摘要


In this thesis, we proposed a system which can classify the cecum image or not automatically. Our system use the feature of cecum to recognize the input images. With the advance of technology, economic prosperity and the trend of western food, the diseases of digestive system are on the rise, especially the colorectal cancer. Since A.D. 2007, colorectal cancer had become top one cancer people who got in Taiwan. Over 10,000 people are diagnosed to have colon cancer every year. According to government statistics, the colorectal cancer is the most expensive cancer in Taiwan in 2014. If the colorectal cancer has been found and cured in Tis and T1, almost 90% patients can survive after five years. There is no symptom in the early stage of colorectal cancer, so examination regularly can reduce the risk of colorectal cancer. Besides fecal occult blood test, the doctor can also check more detail by colonoscopy. From the above, colonoscopy quality is very important, we should keep it in high sensitivity. To make high quality and faster workflow, a lot of things still need to work hard. In this thesis, we focus on the cecal intubation rate (CIR) which is an important quality of colonoscopy. According to the study [9], it is a positive correlation between CIR and colorectal cancer. Thus, we proposed the cecum recognition method. We test 650 images include cecum images and not cecum images. In our experiment results, our method can provide a classification and have a good performance with another new 100 cecum images which was selected strictly by NTU hospital. Keywords: cecum, colorectal cancer, cecal intubation rate, pattern recognition, image segmentation, cross validation, line detection, classification

參考文獻


[6] 衛生福利部中央健康保險署,"103年各類癌症健保前10大醫療支出統計", http://www.nhi.gov.tw/
[1] "World Cancer Research Fund International, Worldwide Data." http://www.wcrf.org/int/cancer-facts-figures/worldwide-data
[2] "International Agency for Research on Cancer, Simple Maps." http://globocan.iarc.fr/Pages/Map.aspx
[7] Robert L. Barclay, Joseph J. Vicari, Andrea S. Doughty et al. "Colonoscopic withdrawal times and adenoma detection during screening colonoscopy," N Engl J Med, vol.355, No.24, pp.2533-2541, Dec. 2006.
[8] Michal F. Kaminski, Jaroslaw Regula, Ewa Kraszewska et al. "Quality indicators for colonoscopy and the risk of interval cancer," N Engl J Med, vol.362, No.19, pp.1795-1803, May. 2010.

延伸閱讀


國際替代計量