Research On Segmentation Method Of Lung Tumor Image Based On Improved U-Net

Medical images play a vital role in the diagnosis of many major diseases.Lung cancer is one of the most common malignant tumors with the highest mortality rate in the world,and its tumor lesion segmentation is of great significance.The continuous development of deep learning technology has gradually made great progress in the field of computer vision,especially in the field of medical image segmentation.Traditional methods can no longer meet the increasingly large and complex medical image segmentation tasks,and in deep learning algorithms,the uniqueness of the structure of the fully convolutional neural network U-Net and its sensitivity to medical images make it unique advantages in medical image segmentation tasks.The U-Net basic network model is suitable for most medical images,such as lung CT images,but there will be many problems in the segmentation process,such as gradient disappearance,easy loss of spatial information,and low feature utilization,which makes it difficult to improve the accuracy of segmentation.Aiming at the shortcomings,this paper adopts two lung medical image segmentation methods based on improved U-Net,namely improved Deep ResUnet and U-Net3+,which can well solve the problems of gradient disappearance and low feature utilization.On the lung CT data set,the two improved algorithms are compared with the improved algorithms of other U-Net series.The segmentation results are excellent.It can completely segment the lung tumor of a continuous sequence slice of a case.The improved Deep ResUnet The Dice coefficient of the network can reach 0.9243,which is 0.0160 higher than Deep ResUnet,and the Dice coefficient of U-Net3+network can reach a higher 0.9318,which is 0.0057 higher than U-Net++,and they are also significantly faster in training speed.The improvement verifies the effectiveness of the algorithm.These two kinds of U-Net improved networks have their own characteristics.In addition,this paper also designs and implements a lung CT image tumor segmentation system,which integrates seven U-Net series segmentation algorithms and evaluation index calculation functions.The system has a huge auxiliary effect and application value for the diagnosis and treatment of early lung cancer patients,and has laid a solid foundation for the subsequent three-dimensional reconstruction of tumors.