Deep sparse feature selection for computer aided endoscopy diagnosis

作者：

Highlights：

• We design a new feature selection model via group sparsity, named Deep Sparse SVM.

• We design a general computer aided endoscopy diagnosis framework.

• We build a new endoscopy dataset including 10k+ images from 1284 volunteers.

• We annotate 3.8k+ images with pixel-level and frame-level groundtruth.

摘要

Highlights•We design a new feature selection model via group sparsity, named Deep Sparse SVM.•We design a general computer aided endoscopy diagnosis framework.•We build a new endoscopy dataset including 10k+ images from 1284 volunteers.•We annotate 3.8k+ images with pixel-level and frame-level groundtruth.

论文关键词：Deep sparse,Group sparsity,Feature selection,Computer aided diagnosis,Endoscopy,Image representation

论文评审过程：Received 10 April 2014, Revised 2 July 2014, Accepted 10 September 2014, Available online 20 September 2014.

论文官网地址：https://doi.org/10.1016/j.patcog.2014.09.010