Low-dose computed tomography (LDCT) is widely used for lung cancer screening because it significantly reduces radiation exposure. However, lowering the radiation dose increases image noise and reconstruction artifacts, which often mask small nodules or distort tumor boundaries. Deep learning techniques have emerged as powerful tools for restoring LDCT quality without sacrificing diagnostic content. This study investigates the effectiveness of convolutional neural networks (CNNs) and generative adversarial networks (GANs) for artifact reduction and noise suppression in LDCT lung images. Using a curated dataset from public repositories, LDCT images were enhanced and compared against routine-dose CT (RDCT) references. Performance was evaluated using peak signal-to-noise ratio (PSNR), structural similarity index (SSIM), and tumor visibility scores assessed by radiologists. Results show that GAN-based models achieved the highest artifact reduction performance, improving PSNR by 28.4% and SSIM by 17.6% compared to conventional denoising methods. The improved clarity of tumor edges and lesion contrast confirmed the superiority of deep learning approaches. The findings suggest that deep learning-based reconstruction can significantly enhance tumor visibility in LDCT scans, supporting more accurate early detection and diagnosis.