目的 探讨利用计算机辅助的人工智能方法评估肝纤维化时包膜线的几何特征,进而定量评价肝纤维化程度的可行性.方法 建立肝纤维化兔模型59只,分别于第6周、第12周、第18周采集肝纤维化兔的肝包膜二维高频声像图供分析,随后获取其肝组织病理切片,依据纤维化程度分为S0-S1组、S2组、S3组、S4组.利用计算机采用人工监督和梯度优化相结合的方法,获取肝包膜线及其形状控制点,提取形状控制点的夹角均值、夹角方差以及包膜线的连续线段数3个参数,分别评估肝包膜轮廓线的平滑性和连续性.结果 随着肝纤维化程度的加重,夹角均值和方差均逐渐增大,其中S0-S1组与其他三组间比较差异均有统计学意义(P<0.05);长度和达到成像切面总宽度80％ 时所需要的线段数量亦逐渐增多,S0-S1组与S4组比较(P=0.025)、S2组与S4组比较(P=0.004)差异均有统计学意义.结论 肝包膜的几何特征分析可以提示肝纤维化的存在,有望为肝纤维化的早期诊断提供新的定量评估方法.%Objective To explore the feasibility of quantitative evaluation in fibrosis by geometrical characteristics of hepatic capsule based on computer technology . Methods Fifty-nine rabbits models of fibrosis were established and then were divided into four groups according to the results of pathologic examination:S0-S1 ,S2 ,S3 and S4 groups . Two-dimensional high-frequency ultrasound images of hepatic capsule of the rabbits were collected at 6th week ,12th week and 18th week . All the images were collected to obtain the contour line of the capsule and its shape control points by combining manual monitoring and gradient optimization methods . Subsequently ,the mean value and variance of angles between adjacent seed points and the number of continuous segments of the contour line ,were extracted to evaluate the smoothness and continuity of the contour line of the hepatic capsule . Results With aggravating fibrosis ,the mean value and variance of the angles were gradually increased ,and the differences between S0-S1 and other 3 groups were all statistically significant( P <0 .05) . Similarly ,the number of segments required to make the sum of lengths of continuous segments reach up to 80％ of the total width of the imaging section was gradually increased . The differences between S0-S1 and S4 group( P =0 .025) ,S2 and S4 group( P =0 .004) were also statistically significant . Conclusions Analysis of geometrical characteristics of the hepatic capsule partially indicates the presence of early-stage fibrosis ,and is expected to provide a quantitative method for non-invasive assessment of fibrosis .
展开▼
