Comparative Study of Artifact Removal Algorithms for Enhancing Lung CT Image Quality in Early Cancer Detection

Artifact removal is a critical preprocessing step in lung computed tomography (CT) imaging, as image artifacts negatively affect the accuracy of early lung cancer detection. Various algorithms have been developed to suppress noise, reduce motion distortions, and eliminate reconstruction errors that obscure lung nodules. This study presents a comparative evaluation of commonly used artifact removal algorithms, including Median Filtering, Non-Local Means (NLM), Bilateral Filtering, Wavelet Transform Denoising, and Deep Learning–based Convolutional Neural Networks (CNNs). The comparison is based on quantitative metrics such as Peak Signal-to-Noise Ratio (PSNR), Structural Similarity Index (SSIM), and computational time, as well as qualitative improvement in nodule visibility. Results show that traditional filters such as Median and Bilateral methods reduce noise efficiently but introduce smoothing that may obscure small nodules. Wavelet-based denoising preserves structural details but is computationally expensive. CNN-based artifact removal provides superior performance with the highest PSNR and SSIM values while maintaining diagnostic features essential for early tumor identification. The study concludes that deep learning techniques offer the most reliable balance between noise reduction and structural preservation for clinical lung cancer screening.