Inverted papilloma and nasal polyp classification using a deep convolutional network integrated with an attention mechanism

Xinyao Li, Haoran Zhao, Tao Ren, Yicong Tian, Aihui Yan, Wei Li

Abstract: Highlights•A deep CNN with attention mechanism composed of DenseNet and SENet, is applied for classifying the IP and NP in CT images.•The classification performance of SE-DenseNet is evaluated, and the accuracy of our method can achieve 88.4% with an AUC of 0.87.•The visualized heatmaps obtained from the convolutional layers indicate the reliability of the method.