Learning efficient, explainable and discriminative representations for pulmonary nodules classification

作者：

Highlights：

• First, to our best knowledge, this is the first attempt that uses NAS for pulmonary nodules classification.

• Second, we analyse the reasoning process of the network, which is in conformity with physicians’ diagnosis.

• Third, we employ A-Softmax loss to train the network for learning discriminative representations.

• Forth, our model is highly comparable with previous SOTA method by using less than 1/40 parameters. The related code and models have been released at: https://github.com/fei-hdu/NAS-Lung.

摘要

•First, to our best knowledge, this is the first attempt that uses NAS for pulmonary nodules classification.•Second, we analyse the reasoning process of the network, which is in conformity with physicians’ diagnosis.•Third, we employ A-Softmax loss to train the network for learning discriminative representations.•Forth, our model is highly comparable with previous SOTA method by using less than 1/40 parameters. The related code and models have been released at: https://github.com/fei-hdu/NAS-Lung.

论文关键词：Pulmonary nodule classification,Convolutional neural network,Neural architecture search,Computer-aided diagnoses,Convolutional block attention module

论文评审过程：Received 1 July 2020, Revised 20 November 2020, Accepted 22 November 2020, Available online 14 January 2021, Version of Record 20 January 2021.

论文官网地址：https://doi.org/10.1016/j.patcog.2021.107825