文献类型：期刊文献

英文题名：PPNet: Pyramid pooling based network for polyp segmentation

作者：Hu, Keli[1,2,3];Chen, Wenping[1];Sun, YuanZe[1];Hu, Xiaozhao[4];Zhou, Qianwei[5];Zheng, Zirui[1]

机构：[1]Shaoxing Univ, Dept Comp Sci & Engn, Shaoxing 312000, Peoples R China;[2]Hangzhou Med Coll, Zhejiang Prov Peoples Hosp, Affiliated Peoples Hosp, Canc Ctr, Hangzhou 310014, Peoples R China;[3]Peking Univ, Informat Technol R&D Innovat Ctr, Shaoxing 312000, Peoples R China;[4]Shaoxing Peoples Hosp, Shaoxing 312000, Peoples R China;[5]Zhejiang Univ Technol, Coll Comp Sci & Technol, Hangzhou 310023, Peoples R China

年份：2023

卷号：160

外文期刊名：COMPUTERS IN BIOLOGY AND MEDICINE

收录：SCI-EXPANDED(收录号：WOS:001008587300001)、、EI(收录号：20232014092478)、Scopus(收录号：2-s2.0-85159149656)、WOS

基金：This work was supported in part by the Humanities and Social Sciences Youth Foundation of Ministry of Education of China under Grant 21YJCZH039, Zhejiang Provincial Postdoctoral Science Foundation under Grant ZJ2022066, the Natural Science Foundation of Zhejiang Province under Grant LY20F020011, TY22F025548, the National Natural Science Foundation of China under Grant 61603258, 61802347, 62271448, 61772018, and in part by the Key scientific research project of Shaoxing University under Grant 2020LG1004.

语种：英文

外文关键词：Colorectal polyp; Transformer; Polyp segmentation; Pyramid pooling

外文摘要：Colonoscopy is the gold standard method for investigating the gastrointestinal tract. Localizing the polyps in colonoscopy images plays a vital role when doing a colonoscopy screening, and it is also quite important for the following treatment, e.g., polyp resection. Many deep learning-based methods have been applied for solving the polyp segmentation issue. However, precisely polyp segmentation is still an open issue. Considering the effectiveness of the Pyramid Pooling Transformer (P2T) in modeling long-range dependencies and capturing robust contextual features, as well as the power of pyramid pooling in extracting features, we propose a pyramid pooling based network for polyp segmentation, namely PPNet. We first adopt the P2T as the encoder for extracting more powerful features. Next, a pyramid feature fusion module (PFFM) combining the channel attention scheme is utilized for learning a global contextual feature, in order to guide the information transition in the decoder branch. Aiming to enhance the effectiveness of PPNet on feature extraction during the decoder stage layer by layer, we introduce the memory-keeping pyramid pooling module (MPPM) into each side branch of the encoder, and transmit the corresponding feature to each lower-level side branch. Experimental results conducted on five public colorectal polyp segmentation datasets are given and discussed. Our method performs better compared with several state-of-the-art polyp extraction networks, which demonstrate the effectiveness of the mechanism of pyramid pooling for colorectal polyp segmentation.