收录:
摘要:
Pneumonia diagnosis based on CT scans is crucial for the effective treatment. Existing deep leanring-based methods mainly focus on the global struction of the whole lung organs, while ignore the information of local detailed lesions. This can easily lead to errors in pneumonia decisions and a decline in classification accuracy. Actully, the diagnosis process of specialists in practice involves basically two steps glancing the whole lung organs to capture global information (global view) and gazing at local regions for observing detailed lesion (local view). To mimic this behaviour, we propose a multi-view information fusion network for pneumonia diagnosis from full sequence CTs. First, we design a multi-view information fusion network to extract spatial features from lung CT slices from global and local perspectives. Then, a recursive neural network (RNN) is utilized to solve the problem of dependency between slices and the continuity of lesions. Extensive experiments on real-world datasets are conducted and the results demonstrate the effectiveness of our proposed method. © 2023 IEEE.
关键词:
通讯作者信息:
电子邮件地址:
来源 :
年份: 2023
页码: 1648-1652
语种: 英文
归属院系: