Low-Dose CT Image Reconstruction using Vector Quantized Convolutional Autoencoder with Perceptual Loss

Computed Tomography (CT) has become a useful screening procedure to identify disease or injury within various regions of the human body. The human beings' health issues caused by CT radiation have attracted the interest of the researchers and academic community. Reducing the radiation dose is the solution, but the CT image generated with low-dose radiation results in excessive noise due to lower intensity and fewer angle measurements. Low-dose CT scan images reduce image quality and thus affect a doctor's diagnosis. Deep learning methods have become increasingly popular in recent years, many models have been proposed for Low-Dose CT image reconstruction. Low-Dose CT Image Reconstruction is an active area of modern medical imaging research. Deep learning-based medical image reconstruction methods will be helpful to reduce noise without compromising image quality. Therefore, this paper introduces a novel CT image reconstruction method based on the vector quantization technique utilized in the convolutional autoencoder network. The quality of the results is evaluated based on the perceptual loss function. Experimental evaluations are conducted on the LoDoPaB-CT benchmark dataset. Its result showed that the proposed network obtained better performance metric values and better noise elimination results, in terms of quantitative and visual evaluation, respectively.