Abstract: Objective Computer-Aided Diagnosis of pneumonia based on deep learning is a major research focus. However, during featureextraction, the positionalfeatures of lesions are often not sufficiently captured, making it difficult to effectively focus on affected regions. Method To address the above problems, this paper proposes a Cross-Dimensional Feature-Enhancement Former(CDFEF) for the computer -aided diagnosis of pneumonia from X-ray images. The main contributions are as follows:Firstly, a Cross -Dimension Location-Aware Module(CDLAM) is designed to effectivelyextract space and channel-wise information. Secondly, aLocal-Global Former Module(LGFM) is designed to enable the network focus more effectivelyon diseased regionsby using a parallel architecture of spatial window attention and channel grouped attention. Result In order to verify the effectiveness of the proposed model, comparative experiments are conducted on two chest X-ray datasets. The results show that the accuracy, F1 value, recall rate, accuracy and specificity of the proposed model on Dataset 1 are 97. 48%, 95. 14%, 95. 15%, 95. 12% and 98. 37%, respectively. The accuracy, F1 value, recall rate, accuracy and specificity on Dataset 2 are 97. 14%, 95. 13%, 95. 46%, 95. 22% and 97. 37%, respectively. Conclusion The CrossDimension Feature-Enhancement Former proposed in this paper can effectively improve the accuracy of pneumoniaidentification in chest X-ray images, which holds important clinical relevance for Computer-Aided Diagnosis.