基于优化随机森林的肝脏病变识别应用研究  

Research on the Application for Liver Lesion Recognition Based on Optimized Random Forest

作　　者：姚冰莹 杨薇 陈坚强 牛婷婷 YAO Bingying;YANG Wei;CHEN Jianqiang;NIU Tingting(Software Engineering Institute of Guangzhou,Guangzhou 510990,China;Wuhan University of Science and Technology,Wuhan 430081,China)

机构地区：[1]广州软件学院,广东广州510990 [2]武汉科技大学,湖北武汉430081

出　　处：《现代信息科技》2021年第18期90-93,共4页Modern Information Technology

基　　金：广东省高校青年创新项目(2018 KQNCX394、2018KQNCX395、2019KQNCX222)。

摘　　要：针对肝病早期的病理特征不够明显容易出现漏诊的问题以及引入噪声的肝脏CT图像容易导致疾病的误诊,提出了一种基于优化的随机森林的肝脏病变识别方法。首先通过PNet图像增强网络,实现噪声伪影抑制和细节对比度提升。将增强后的肝脏CT图像输入基于SENet与ResNeXt-152结合双通道重标定机制的网络提取二维特征并突出重要特征,使用改进的类C3D网络进行三维空间特征的提取,然后进行多维特征融合,实现肝脏CT图像的高级语义特征的提取。最后使用随机森林进行肝脏病变识别,使用3D断层扫描肝脏CT图像进行实验,达到较好的病变识别效果。Aiming at the problem that the early pathological features of liver disease are not obvious enough to cause missed diagnosis and the noise-introduced liver CT images easily lead to the misdiagnosis of diseases,a liver lesion recognition method based on optimized random forest is proposed.First,the PNet image enhancement network is used to achieve noise artifact suppression and detail contrast enhancement.The enhanced liver CT image is input into a network based on SENet and ResNeXt-152 and combined with a dual-channel recalibration mechanism to extract two-dimensional features and highlight important features,uses an improved C3D-like network to extract three-dimensional spatial features,and then perform multi-dimensional feature fusion.Realize the extraction of high-level semantic features of liver CT images.Finally,random forest is used to identify liver lesions,and 3D tomography liver CT images are used for experiments to achieve better lesion recognition results.

关 键 词：肝脏CT图像 深度学习 高级语义特征 随机森林 病变识别 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] TP391.4[自动化与计算机技术—控制科学与工程]