基于深度学习对前列腺及前列腺癌MRI分割的研究进展  

Progress in MRI segmentation of prostate and prostate cancer based on deep learning

作　　者：朱静逸 李海庆[2] 金倞 耿道颖 ZHU Jingyi;LI Haiqing;JIN Liang;GENG Daoying(Academy for Engineering&Technology,Fudan University,Shanghai 200433,China;Huashan Hospital,Fudan University;Shanghai Medical College of Fudan University.)

机构地区：[1]复旦大学工程与应用技术研究院,上海200433 [2]复旦大学附属华山医院 [3]复旦大学上海医学院

出　　处：《国际医学放射学杂志》2023年第3期322-325,332,共5页International Journal of Medical Radiology

摘　　要：前列腺癌的早期诊断有助于提高治愈率以及病人生存率。随着人工智能(AI)技术的逐渐成熟,计算机辅助诊断前列腺癌的技术也在迅速发展,然而将其常规用于临床仍然存在许多挑战。对常用的前列腺MRI公开数据集进行了系统梳理,并对基于卷积神经网络(CNN)、U-Net、分辨率增强方法、对抗生成(GAN)和Transformer技术的前列腺及前列腺癌病灶分割方法最新进展予以综述。Early diagnosis of prostate cancer can improve the cure and survival rates of patients.With the gradual maturity of artificial intelligence technology,the technology of computer-aided diagnosis of prostate cancer is also developing rapidly,however,there are still many challenges in its routine clinical application.We systematically sorted out the commonly used prostate MRI public data sets and summarized the latest progress in segmentation of prostate and prostate cancer lesion based on convolution neural network(CNN),U-net,resolution enhancement,antagonism generation(GAN)and Transformer technology.

关 键 词：前列腺癌 深度学习 磁共振成像 计算机辅助诊断 图像分割算法 

分 类 号：R445.2[医药卫生—影像医学与核医学] R737.25[医药卫生—诊断学]