一种基于宽度学习系统变体结构的肺炎检测方法

A pneumonia detection method based on the variant structure of broad learning system

肺炎作为常见的呼吸系统疾病,准确、快速地诊断对患者的健康恢复至关重要。随着医疗技术的革新和人工智能的发展,计算机辅助诊断在医学领域的应用日益广泛。深度学习在肺炎检测领域取得了显著的成果,但其庞大的参数数量和复杂的网络结构导致训练时间长、计算资源消耗大等局限性。为了解决上述问题,提出了一种基于宽度学习系统变体结构的肺炎检测方法。该方法在原始宽度学习系统的基础上,引入了级联金字塔结构;同时,利用预训练的EfficientNet网络作为前置特征提取器;此外,还提出了适用于该模型的增量学习算法,包括增加额外的增强节点、特征节点和训练样本,以进一步优化模型性能;最后,在公开的肺炎胸部X射线数据集上进行了对比实验。实验结果表明,该方法实现了92.83%的准确率,AUC值高达98.86%,与众多深度卷积神经网络相比,具有相似的精度,同时大幅缩短了模型的训练时间。

Pneumonia as a common respiratory disease,its accurate and rapid diagnosis is crucial to the health of patients.With the innovation of medical technology and the development of artificial intelligence,computer-aided diagnosis has been increasingly used in the medical field.Deep learning has achieved remarkable results in the field of pneumonia detection,but its large number of parameters and complex network structure lead to limitations such as long training time and high consumption of computational resources.To solve the above problems,a pneumonia detection method based on the variant structure of broad learning system is proposed in this paper.The method introduces the cascade pyramid structure on the basis of original broad learning system.Meanwhile,the pre-trained EfficientNet network is utilised as the front feature extractor.In addition,the incremental learning algorithms applicable to the model are proposed in this paper,including adding additional enhancement nodes,feature nodes and training samples to further optimise the model performance.Finally,comparative experiments are conducted on the publicly available dataset of chest X-rays for pneumonia.The experimental results show that the method in this paper achieves 92.83%accuracy and 98.86%AUC value,which are comparable to many deep convolutional neural networks,while the training time of the model is significantly shortened.