A multi-scale dense residual and global attention combined image denoising network was proposed to address the issues of texture detail loss and excessive smoothness in reconstructed images caused by the lack of intrinsic connection between spatial features and denoising tasks in current low-dose computed tomography (LDCT) image denoising methods. The multi-scale dense residual blocks were introduced to extract multi-scale feature information from images, and the global attention mechanism (GAM) was used to focus on cross dimensional information between different channels of the model, while adding skip connections to further expand the range of global interactive features, and finally the multi-scale feature loss function was used to enhance image texture details and avoid the problem of image smoothness. After experimental verification, the proposed algorithm achieves 35.183 8 dB and 0.960 5 in peak signal-to-noise ratio (PSNR) and structural similarity index method (SSIM), respectively, which effectively preserves image details while removing noise, outperforming other algorithms.