基于胶囊网络和注意力的小样本脑肿瘤分类方法

he classification of brain tumors holds significant research value in the field of medical diagnostics. However, due to the high acquisition cost of medical imaging data and the limited number of samples, the traditional convolutional neural networks (CNNs) encounter limitations in extracting target features from small sample datasets, resulting in suboptimal accuracy. To address this issue, this study proposes a classification model that integrates an attention mechanism and capsule networks. The model leverages multidimensional vector neurons within the capsule network to comprehensively represent target features. Simultaneously, recognizing the insufficient information on target features in small sample scenarios, an attention mechanism is introduced to enhance the learning efficiency of the neural network. By learning the importance of different features, this mechanism guides the classification network to focus on features that significantly contribute to the classification results, while diminishing the impact of less influential features, thereby enhancing the model\'s classification accuracy. Experimental results conducted on the selected brain tumor dataset indicate that this algorithm, in comparison to traditional capsule network algorithms, achieves a higher level of accuracy.