Convolutional Neural Networks (CNNs) show considerable potential in removing artifacts and noise from low-dose CT (LDCT) images. However, conventional CNNs often struggle to restore fine details effectively, leading to overly smoothed outputs and poor noise reduction performance. To overcome these limitations, this study introduces a dual convolutional network for LDCT image denoising, comprising a main network and a sub-network dedicated to extracting global and local features, respectively. These features are then fused to enrich the representation of complex structures and details. Furthermore, adaptive edge extraction operators are incorporated to enhance structural details, while large kernel attention (LKA) modules are employed to facilitate adaptive self-attention and model long-range dependencies. Extensive experimental results validate that the proposed dual convolutional denoising approach not only effectively suppresses noise in LDCT images but also significantly improves the preservation of fine details.