Deep Learning Is Effective for Classifying Normal versus Age-Related Macular Degeneration OCT Images

返回OCT图像分类

阅读笔记:

https://blog.csdn.net/qq_42950884/article/details/103214402

https://blog.csdn.net/JianJuly/article/details/80311767

论文地址:

https://arxiv.org/ftp/arxiv/papers/1612/1612.04891.pdf

实验结果:

数据量:260w张OCT图像、52690张正常人的OCT图像、48312张老年性黄斑患者OCT图像
图像角度: we achieved an area under the ROC curve of 92.78% with an accuracy of 87.63% 
黄斑角度: we achieved an area under the ROC curve of 93.83% with an accuracy of 88.98%
病人角度:we achieved an area under the ROC curve of 97.45% with an accuracy of 93.45%

occlusion test: 由20x20的方框遍历整个图片,遮挡遍历位置,一旦遮挡后图像的“概率”(此处不解)变低,则说明是深度学习所感兴趣的地方

遮挡测试(occlusion test) 
使用遮挡测试来识别图像中对分类结果贡献最大的区域: 
Tips:又一篇论文使用了遮挡测试,可见遮挡测试对于结果的可解释性非常之重要 
使用为20×20像素的黑框在原图中移动,当输出概率降低得最多时,表明此时黑框覆盖的区域对分类结果的贡献度最高

遮挡测试参考论文:

Visualizing and Understanding Convolutional Networks

https://blog.csdn.net/tonyshengtan/article/details/52528600

SCI推出相关论文:

Fully Automated Detection and Quantification of Macular Fluid in OCT Using Deep Learning

Identifying Medical Diagnoses and Treatable Diseases by Image-Based Deep Learning​​​​​​​

 

 

【其它笔记参考:引用1】
【论文地址:引用2】

你可能感兴趣的:(深度学习,机器学习,机器学习,深度学习,人工智能,计算机视觉)