CAJ | 학술논문

目的探讨数字图像处理技术在超声内镜鉴别诊断自身免疫性胰腺炎与慢性胰腺炎中的应用价值。方法回顾2005年5月至2013年1月确诊的100例慢性胰腺炎和81例自身免疫性胰腺炎患者的内镜超声图像，选择具有病变典型表现的内镜超声图像和勾画的感兴趣区域，通过计算机提取胰腺分类系统中的9大类105维特征，采用类间距法和顺序前进法筛选纹理特征的最优特征组合，通过支撑向量机建立分类模型，使用2种不同的样本集划分方法对慢性胰腺炎和自身免疫性胰腺炎病例进行自动分类，统计分类的准确率、敏感度、特异度、阳性预测值和阴性预测值。结果最优纹理特征组合包括5大类25维特征，此时准确分类率达最大（90．08％）。181例病例采用均匀划分样本集法和留一法随机划分为训练集和测试集，共进行200次随机检验，均匀划分样本集法最终分类准确率为（86．04±3．15）％、敏感度为（83．66±6．57）％、特异度为（88．54±4．37）％、阳性预测值为（85．96±4．44）％、阴性预测值为（87．12±4．39）％。结论计算机辅助的图像分析技术具有客观性、无创性，能够提高内镜超声识别自身免疫性胰腺炎的准确率，为自身免疫性胰腺炎诊断提供了一个新的有价值的研究方向。
목적탐토수자도상처리기술재초성내경감별진단자신면역성이선염여만성이선염중적응용개치。방법회고2005년5월지2013년1월학진적100례만성이선염화81례자신면역성이선염환자적내경초성도상，선택구유병변전형표현적내경초성도상화구화적감흥취구역，통과계산궤제취이선분류계통중적9대류105유특정，채용류간거법화순서전진법사선문리특정적최우특정조합，통과지탱향량궤건립분류모형，사용2충불동적양본집화분방법대만성이선염화자신면역성이선염병례진행자동분류，통계분류적준학솔、민감도、특이도、양성예측치화음성예측치。결과최우문리특정조합포괄5대류25유특정，차시준학분류솔체최대（90．08％）。181례병례채용균균화분양본집법화류일법수궤화분위훈련집화측시집，공진행200차수궤검험，균균화분양본집법최종분류준학솔위（86．04±3．15）％、민감도위（83．66±6．57）％、특이도위（88．54±4．37）％、양성예측치위（85．96±4．44）％、음성예측치위（87．12±4．39）％。결론계산궤보조적도상분석기술구유객관성、무창성，능구제고내경초성식별자신면역성이선염적준학솔，위자신면역성이선염진단제공료일개신적유개치적연구방향。
Objective To explore the feasibility of using digital imaging processing (DIP)to extract EUS image parameters for the differential diagnosis of autoimmune pancreatitis (AIP)and chronic pancreati-tis (CP).Methods A total of 81 patients with AIP and 100 patients with CP diagnosed from May 2005 to January 2013 were recruited to this study.A total of 105 parameters of 9 categories were extracted from the region of interest by using computer-based techniques.Then the distance between class algorithm and se-quential forward selection (SFS)algorithm were used for a better combination of features.A support vector machine (SVM)predictive model was built,trained,and validated.Results Overall,25 parameters of 5 categories were selected as a better combination of features when the incidence of accurate category was max (90.08%).A total of 181 sample sets were randomly divided into a training set and a testing set by using two different algorithms and 200 random tests were performed.The average accuracy,sensitivity,specificity, the positive and negative predictive values of AIP based on the half-and-half method were (86.04 ± 3.15)%,(83.66 ±6.57)%,(88.54 ±4.37)%,(85.96 ±4.44)% and (87.12 ±4.39)%,respective-ly.Conclusion Computer-aided diagnosis of EUS images is objective and non-invasive,which can improve the accuracy in differentiating AIP from CP.This technology provides a new valuable diagnostic tool for the clinical determination of AIP.