Machine learning and artificial intelligence tools are on the precipice of becoming a popular method for clinical diagnosis and disease prediction. Their ability to solve specific problems given sets of constraints are unmatched and they provide novel solutions to a plethora of different tasks. This thesis will examine a deep learning image classifiers capability of successfully classifying male and female MR images into their respective counterparts. The results of these models are visualized using a Grad-Cam to help gain insight on the sexual dimorphisms present within the human brain. The models showed sex dimorphisms exist in previously known areas like the frontal, temporal and precuneus gyri along with the cerebellum and thalamus. But other regions such as the cingulate, postcentral, and fusiform gyri illustrated differences not commonly mentioned in literature. This paper deviates away from the traditional statistical approaches of neuroimaging and analysis techniques and provides a new method to draw conclusions on individual volumes.
Member of collection