Machine Learning in Psychiatry

I am interested in using tools from machine learning to approach clinical and scientific problems in psychiatry and behavioral neurology. These tools include supervised learning, pattern classification, regression, unsupervised learning and clustering, information theoretic methods and Bayesian methods. Many problems in psychiatry involve pattern recognition with large datasets in high dimensions, and modern artificial intelligence algorithms are especially well-equipped to deal with these datasets. Furthermore, novel techniques may be developed based on the complexity of specific datasets and goals of specific analytical problems. Specific models can guide clinical decision-making as well as point to relevant variables in disease pathogenesis and venues of therapy.

Diseases of Cognition, Memory, Language, and Higher Cortical Functions

I am interested neurological and psychiatric diseases such as stroke, epilepsy, Alzheimer's disease, mood disorders and schizophrenia, that affect either a localized region in the brain or a combination of regions.   I want to use psychophysical and neurophysiological methods to probe the anatomical and functional properties of these cortical networks.  I want to use techniques such as fMRI, EEG/ERP, MEG, TMS, deep cortical electrodes and other imaging modalities to study these networks.  Genetically specific disorders such as Williams syndrome and Lesch-Nyhan syndrome may be excellent models to tease out how genes affect networks and behavioral phenotypes.  Studies of cortical networks may yield important medical applications that can aid in disease classification, prediction and prognostication, as well as possible therapeutics.

Computational Neuroscience of Neurological and Psychiatric Disorders

I am interested in applying theoretical and computational approaches in analyzing complex nervous system disorders such as substance abuse and depression.  Models at network or higher levels may reveal new insights into the disease's etiology, progression and possible management options.  

Models of Olfactory Discrimination in Drosophila

With Drs. Larry Abbott and Richard Axel, I worked on a computational model of olfactory discrimination in Drosophila during my Ph.D.  This model is able to explain how fruit flies can process and discriminate innately important odors, such as pheromones and food odors, while simultaneously learning new associations.  It hypothesizes two design principles in two different anatomical regions in the fly.  It also proposes a previously unrecognized functional role for the antennal lobe, the Drosophila homologue of the olfactory bulb.

Theoretical Models of Olfactory Discrimination in Drosophila, Ph.D. Dissertation, Dec. 2008, Columbia University