This paper proposes a residual encoder-decoder convolutional neural network (RED-CNN) for low-dose CT (LDCT) imaging. The RED-CNN combines autoencoder, deconvolutional network, and shortcut connections to enhance image quality while reducing noise and preserving structural details. The network is trained on patch-based data and achieves competitive performance compared to state-of-the-art methods in both simulated and clinical cases. The method is evaluated in terms of noise suppression, structural preservation, and lesion detection, showing favorable results. The RED-CNN is implemented using a symmetric architecture with 10 layers, including 5 convolutional and 5 deconvolutional layers. The network uses shortcut connections and residual compensation to improve training efficiency and performance. The model is trained on simulated and clinical data, demonstrating its effectiveness in improving image quality for LDCT. The proposed method outperforms several existing techniques in terms of quantitative metrics such as RMSE, PSNR, and SSIM. The RED-CNN is also efficient in terms of computational cost and can process images of arbitrary sizes. The study concludes that deep learning has great potential for LDCT imaging, with the proposed RED-CNN offering a promising solution for noise suppression, structural preservation, and lesion detection.This paper proposes a residual encoder-decoder convolutional neural network (RED-CNN) for low-dose CT (LDCT) imaging. The RED-CNN combines autoencoder, deconvolutional network, and shortcut connections to enhance image quality while reducing noise and preserving structural details. The network is trained on patch-based data and achieves competitive performance compared to state-of-the-art methods in both simulated and clinical cases. The method is evaluated in terms of noise suppression, structural preservation, and lesion detection, showing favorable results. The RED-CNN is implemented using a symmetric architecture with 10 layers, including 5 convolutional and 5 deconvolutional layers. The network uses shortcut connections and residual compensation to improve training efficiency and performance. The model is trained on simulated and clinical data, demonstrating its effectiveness in improving image quality for LDCT. The proposed method outperforms several existing techniques in terms of quantitative metrics such as RMSE, PSNR, and SSIM. The RED-CNN is also efficient in terms of computational cost and can process images of arbitrary sizes. The study concludes that deep learning has great potential for LDCT imaging, with the proposed RED-CNN offering a promising solution for noise suppression, structural preservation, and lesion detection.