A Practical Review on Medical Image Registration: From Rigid to Deep Learning Based Approaches

Abstract

The large variety of medical image modalities (e.g. Computed Tomography, Magnetic Resonance Imaging, and Positron Emission Tomography) acquired from the same body region of a patient together with recent advances in computer architectures with faster and larger CPUs and GPUs allows a new, exciting, and unexplored world for image registration area. A precise and accurate registration of images makes possible understanding the etiology of diseases, improving surgery planning and execution, detecting otherwise unnoticed health problem signals, and mapping functionalities of the brain. The goal of this paper is to present a review of the state-of-the-art in medical image registration starting from the preprocessing steps, covering the most popular methodologies of the literature and finish with the more recent advances and perspectives from the application of Deep Learning architectures.