摘要
笔者利用200位骨癌病人的图像数据,针对多模态骨癌影像受姿势角度等因素的影响以及医学数据集有限而无法获取大量样本的情况,基于注意力机制的卷积神经网络分类算法,结合骨癌病灶特点进行残差网络和双线性融合改良,同时优化损失函数,以便在普通计算机能承受的负载基础上搭建有效且更具实用性的细粒度分类网络模型。
The paper uses the image data of bone cancer patients to study the CNN classification algorithm based on the attention mechanism. In order to solve the problem of the limited number of multimodal bone cancer data sets, we combine Resnet with bilinear fusion method and improve the loss function optimization.It aims to build an effective and lightweight Fine-grained classification network,which can be loaded on an ordinary computer.
作者
柯艺雅
周小波
KE Yiya;ZHOU Xiaobo(Department of Computer Technology,Tongji University,Shanghai 201804,China)
出处
《信息与电脑》
2021年第6期136-138,共3页
Information & Computer