Adaptive Aggregated Attention Network for Pulmonary Nodule Classification

Lung cancer has one of the highest mortality rates in world and threatens people’s health. Timely accurate diagnosis can greatly reduce number deaths. Therefore, an system is extremely important. The existing methods have achieved significant performances on lung diagnosis, but they are insufficient fine-grained representations. In this paper, we propose a novel attentive method to differentiate malignant benign pulmonary nodules. Firstly, residual attention network (RAN) squeeze-and-excitation (SEN) were utilized extract spatial contextual features. Secondly, multi-scale (MSAN) was proposed capture features automatically, MSAN integrated advantages mechanism mechanism, which very important for capturing salient Finally, gradient boosting machine (GBM) algorithm used We conducted series experiments Image Database Consortium image collection (LIDC-IDRI) database, achieving accuracy 91.9%, sensitivity 91.3%, false positive rate 8.0%, F1-score 91.0%. experimental results demonstrate that our outperforms state-of-the-art with respect accuracy, rate, F1-Score.