摘要
医学图像分割是疾病诊断和治疗的重要组成部分,通常由经验丰富的医生或专家手动完成。随着医学成像技术的发展,医学图像的规模快速增长,给医学专家带来了大量且繁琐的工作。因此,许多研究人员提出了医学图像自动分割方法。其中,深度学习近年来已成为医学图像分割任务的首选方法。为此,提出了一种基于U-Net++的脊椎MRI图像分割方法,剪掉了U-Net++的L4阶段,简化了计算量,并改进了YOLOv3模型,用于人体椎间盘检测。实验结果表明,进行剪枝后的网络在实验时间上较原来的方法减少了一半,m AP提升到了81.49%。该方法准确率达到了良好水平,并一定程度上减小了计算量。
Medical image segmentation is an important part of disease diagnosis and treatment,which is usually completed manually by experienced doctors or experts.With the development of medical imaging technology,the scale of medical ima-ges is growing rapidly,which brings a lot of cumbersome work to medical experts.Therefore,many researchers have pro-posed automatic medical image segmentation methods.Among them,deep learning has become the preferred method for medical image segmentation in recent years.In this paper,a human spine MRI image segmentation method based on U-NET++is proposed,which cuts off the L4 stage of U-NET++and simplifies the amount of calculation.Yolov3 model is improved for human intervertebral disc detection.The experiment shows that the experimental time of the pruned network is reduced by half compared with the original method,and the map is increased to 81.49%.The accuracy of the proposed method reaches a good level,and the amount of calculation is reduced to a certain extent.
作者
吴相远
申诺
蒙玉洪
王子民
WU Xiangyuan;SHEN Nuo;MENG Yuhong;WANG Zimin(School of Computer and Information Security,Guilin University of Electronic Technology,Guilin 541004,China)
出处
《桂林电子科技大学学报》
2022年第1期36-42,共7页
Journal of Guilin University of Electronic Technology
基金
国家级大学生创新训练项目(201710595006)
广西高校图像图形智能处理重点实验室研究课题(GIIP201705)。
关键词
卷积网络
人体脊椎MRI图像
图像分割
目标检测
convolutional network
human spine MRI image
image segmentation
object detection
