基于边缘分割与改进CNN的CT影像预诊断技术  

CT image pre⁃diagnosis technology based on edge segmentation and improved CNN

作　　者：董聪慧 岳晓磊 马朋朋[2] DONG Conghui;YUE Xiaolei;MA Pengpeng(Zhangjiakou Infections Disease Hospital,Zhangjiakou 075000,China;The First Affiliated Hospital of Hebei North University,Zhangjiakou 075000,China)

机构地区：[1]张家口市传染病医院,河北张家口075000 [2]河北北方学院附属第一医院,河北张家口075000

出　　处：《电子设计工程》2024年第21期146-150,共5页Electronic Design Engineering

基　　金：河北省科技厅科技支撑计划项目(182777214);2023年张家口市科技计划项目(2322030D)。

摘　　要：针对基于图像识别的智能预诊断精确度较低的问题,文中提出了一种融合边缘分割与改进CNN的CT影像预诊断算法。在Bandelet变换的基础上构建WTS-MRF模型,并采用分割递归算法对CT影像的特征区域进行处理,进而设计出基于决策输出补偿的Faster R-CNN预诊断识别算法。同时还利用了脑出血、肺结核和肾结石等典型病例影像的数据样本,通过设置对比实验验证了该算法的预诊断可靠性。相较于同类预诊断识别方法,所提算法的准确率提升了6%,CT影像的分割准确率平均值为90%,预诊断识别精确率的平均值则可达96.9%。故其性能优于同类文献对比算法,能为基于人工智能的CT影像预诊断技术发展提供一定的理论支撑。Aiming at the low accuracy of intelligent pre⁃diagnosis based on image recognition,this paper proposes a CT image pre⁃diagnosis algorithm that combines edge segmentation and improved CNN.The WTS⁃MRF model is constructed based on Bandelet transform,and the segmentation recursive algorithm is used to segment the characteristic region of CT image.The Fast R⁃CNN pre⁃diagnosis recognition algorithm based on decision output compensation is designed.Based on the image data samples of typical cases such as cerebral hemorrhage,pulmonary tuberculosis and kidney stone,the reliability of the algorithm for pre⁃diagnosis is verified by setting up comparative experiments.Compared with similar pre⁃diagnosis recognition algorithms,the accuracy of the proposed algorithm is improved by 6%,the average segmentation accuracy of CT image is 90%,and the average pre⁃diagnosis recognition accuracy is 96.9%.The performance of this algorithm is better than that of the similar literature comparison algorithm,which provides theoretical support for the development of CT image pre⁃diagnosis technology based on artificial intelligence.

关 键 词：边缘分割 CT影像预诊断 快速区域卷积神经网络 小波域树结构的马尔可夫场模型 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TN713[自动化与计算机技术—控制科学与工程]