Deep learning and state of the art infant brain MRI processing methods

Abstract

Over the past decade there has been heightened interest from federal funding partners, such as the National Institutes of Health, in the developing brain. This interest led to the funding of the Adolescent Brain Cognitive Development study in 2015, the largest long term study of brain development and child health in the United States. Following the developmental brain health momentum, the National Institute of Health and other federal partners more recently funded the HEALthy Brain and Child Development study, which aims to collect information beginning at birth through early childhood, including magnetic resonance imaging of the brain. Due to the dynamic brain development changes that occur during the first year of postnatal life, traditional magnetic resonance imaging brain processing and analysis implemented on adult, adolescent, and pediatric study samples do not work well with infants. With this backdrop in mind, alternative processing and analysis approaches are needed. With improvements in computing power, namely graphical processing units, machine and deep learning approaches have proliferated within the brain magnetic resonance imaging space. Here, within this work, deep learning approaches are explored to improve infant brain image processing, to bridge the gap between data collection and analysis.