本文共 1338 字,大约阅读时间需要 4 分钟。
Less is More: Simultaneous View Classifification and Landmark Detection for Abdominal Ultrasound Images
Author:Zhoubing Xu,..., Shaohua Zhou
idea:提出了一个多任务框架:1)超声视图分类,2)腹部长短抽的landmark detection
- 网络框架:
- Encoder:ResNet50。The first convolutional layer is modifified to take a single channel input. While the fifirst two residual blocks are shared across all tasks for low level feature extraction, two copies of the 3rd and 4th residual blocks are used, one for view classifification, and the other for landmark detection.
- On each level of skip connection between encoder and decoder, we append Global Convolutional Network (GCN) [9] and boundary refifinement modules to capture larger receptive fields.
- 四个Loss:
- Loss_CE:Cross Entropy Loss
- Loss_L2:Regression Loss,热力图回归,求距离
- Loss_cc:Landmark Location Error。采用热力图加权的方式得到predicted landmark,S_i代表横坐标,T_i代表纵坐标。
- Loss_AD:判断一个batch中pred和True是否是来自同一张图
Fully convolutional regression network for accurate detection of measurement points (DLMIA 2017)
Author:Michal Sofka, Fausto Milletari, Jimmy Jia, and Alex Rothberg
解决问题:Accurate automatic detection of measurement points in ultrasound video sequences
方法:1)用Fully Convolutional Neural Network (FCN) regress the point locations;2)a Long Short-Term memory cells which processes several previous frames in order to refine the estimate in the current frame.
转载地址:http://ufeqf.baihongyu.com/