Neuroimaging aims to provide helpful insight on both the anatomy and functionality of neuronal activity within the brain. Patterns in brain functionality from one group of people to another often have the power to identify abnormalities associated with a variety of conditions, such as mood disorders, Alzheimer’s, Parkinson’s, and many more. The volume of data collected from neuroimaging scans is often overwhelming and can be difficult to isolate patterns amongst the noise; thus, the opportunity for machine learning presents itself. Machine learning in neuroimaging attempts to identify patterns in neuronal activity and reliably classify patients with varying conditions. Trained models have proven to perform at a high level of accuracy on neuroimaging data and consequently, studies in this area have increased exponentially in the last decade. 

This paper consists of a case study on the binary classification of fMRI data from epileptic dogs and normal healthy dogs. The entire modeling process, from data pre-processing to final model results are outlined and discussed. A literature review was also conducted to provide insight into modern day applications of machine learning in neuroimaging. Topics such as binary classification in OCD patients, multiclass diagnosis, and an overall evaluation of modeling approaches on fMRI data were discussed.