Artificial Intelligence-Based Image Enhancement and Reconstruction in Computed Tomography Imaging

Abstract

Implementation of artificial intelligence (AI), deep learning (DL), and neural networks (NN) to medical CT imaging holds great promise to improve many challenging computation tasks. The progress made with modern technology has significantly aided in the development of AI technology, and its reduction in cost has made it readily available to the public at large. Applications of NNs to CT images have demonstrated its ability to assist the diagnostic performance of a trained physician through a variety of techniques and approaches. These AI-based computational tasks can be grouped into two broad categories: image enhancement and image reconstruction. Image enhancement with AI is the process of using the existing image to extract information or reduce erroneous information found within the image. Using AI to perform image reconstruction is a method to transform a series of images into another more comprehensive image. Here, we explore different applications of AI, DL, and NNs that address each of these two categories, various subcategories, and the specific approaches to solving their problem at hand.KeywordsArtificial intelligenceDeep learningNeural networksCTImage reconstructionImage enhancementSegmentationMaterial decompositionDenoisingImage quality assessmentImage-domainProjection-domainHybrid-domainAI-driven iterative regularization