Metal Artifacts Reduction in Dental CT Scans

Artifacts caused by the presence of metals in computed tomography scans impact their readability and can cause problems when making decisions for medical professionals. In recent years, deep learning-based methods have seen considerable success in solving this problem, compared to older hand-crafted solutions. In this work, two supervised neural network models (Autoencoder, U-net) are implemented, along with a better way to solve the problem of creating a synthetic dataset, as otherwise in this case it is naturally impossible to obtain. The results in evaluation metrics (PSNR, SSIM) achieved are on par with those of state-of-the-art solutions while reducing the need for prerequisites that are complicated to prepare. This generalized solution enables a broader and easier application without needing a specific controlled environment