Brian 

The paper that I will be presenting on utilizes the preexisting correlation between neuromelanin, a chemical byproduct of dopamine production and psychosis. My lab tested Neuromelanin using Neuromelanin magnetic resonance imaging to obtain Neuromelanin contrast data to train a machine learning model capable of making predictions for patients at clinical high risk of psychosis or with Schizophrenia Spectrum Disorders. Traditional imaging methods used to study dopamine related conditions, such as PET scans, can cost hospitals upwards of two million dollars annually, making them impractical. Additionally, they cannot be used long term due to radiation exposure. Psychosis is often characterized by a disconnect from reality and has been linked to excess presynaptic dopamine production; when the neuron sending the chemical signal produces more dopamine than the receiving neuron can process. The study trained a support vector regression model on NM-MRI data from the substantia nigra, a region of the brainstem that is responsible for most of the brain’s dopamine production. The purpose of this model was to predict psychosis severity scores across two different patient groups. The model successfully predicted symptom severity in both the clinical high risk and the Schizophrenia patients, with NM-MRI showing a significant relationship to dopamine related psychosis markers. These findings suggest that NM-MRI signal in the substantia nigra tracks dopaminergic dysfunction in a way that is clinically relevant. Their work also demonstrates that NM-MRI combined with machine learning could provide a scalable, affordable tool for early identification and monitoring of psychosis.