Ring artifacts (RAF) in X-ray MicroCT images pose significant challenges for accurate visual interpretation and quantitative analysis. These artifacts arise from defective or non-linear detector pixel responses during data acquisition. Severe RAF can render images unusable, making artifact removal crucial. This work introduces a method for generating a synthetic dataset of images with RAF and applies two Deep Learning models UNet with encoder-decoder units and Residual Network (ResNet), to remove RAF. The models, evaluated using SSIM and MSE, showed promising results on synthetic MicroCT images. The UNet-based approach has also been tested on experimental images. Additionally, a two-phase UNet-based approach has been explored, which demonstrates better performance on experimental images for RAF correction.