基于混合域注意力机制和残差网络的特纳综合征分类研究

目的:利用人脸图像,构建基于深度学习的特纳综合征(Turner syndrome,TS)分类模型,旨在提高TS诊断准确率,降低诊断开销.方法:首先,将通道域注意力机制和空间域注意力机制以及残差结构相结合,提出一种具有混合域注意力模块的残差网络,然后使用深度迁移学习技术完成模型的初始化,最后使用TS人脸数据集对网络模型进...

A Classification Method of Turner Syndrome Based on Mixed Domain Attention Mechanism and Residual Network

Objective:A deep learning based classification model is proposed for the diagnosis of Turner syndrome(TS)using facial images,which aims at improving the diagnostic accuracy and reducing costs for patients.Methods:First,the spatial and channel attention mechanisms and the residual block are combined to build a residual network with mixed domain attention blocks;then,the deep transfer learning technology is utilized for model initialization,and finally the resulting model is fine tuned using TS facial images.Results:The empirical results show that the accuracy of the proposed model reaches 0.9171,which is superior to existing TS detection methods.Conclusion:The proposed model is capable to effectively classify TS facial images,which contributes to the clinical diagnosis of TS.