摘要
膀胱癌是泌尿系统最常见的恶性肿瘤,也是目前花费最高的癌症之一。肿瘤的分割在疾病的辅助诊断、治疗规划中具有重要意义。传统的肿瘤分割需要消耗大量的劳动力。随着技术的不断发展,具有预处理少,准确率高等优势的卷积神经网络作为一种图像处理高效的技术,已经在图像分割领域取得了不错的成绩。目前医学图像分割领域得到较好反响的是U-Net网络,但该网络训练深度低,准确率较低。论文提出了一种改进后的Res-U-Net网络模型,参考残差网络构造,与残差结构结合,相比于原始的U-Net模型,Dice系数平均提高了9%。
Bladder cancer is the most common malignant tumor in the urinary system and one of the most expensive cancers.The segmentation of tumors is of great significance in the auxiliary diagnosis and treatment planning of diseases.Traditional tumor segmentation requires a large amount of labor.With the continuous development of technology,convolutional neural networks with the advantages of less preprocessing and high accuracy have achieved good results in the field of image segmentation as an efficient image processing technology.At present,the U-Net network is well received in the field of medical image segmentation,but the net⁃work has low training depth and low accuracy.This paper proposes an improved Res-U-Net network model,which is based on the residual residual network structure and combined with the residual structure.Compared with the original U-Net model,the Dice co⁃efficient is increased by 9%.
作者
曹心姿
梁秋源
李瑞新
蔡兆信
潘家辉
CAO Xinzi;LIANG Qiuyuan;LI Ruixin;CAI Zhaoxin;PAN Jiahui(School of Software,South China Normal University,Foshan 528225)
出处
《计算机与数字工程》
2021年第7期1442-1447,共6页
Computer & Digital Engineering
