Abstract:
Magnetic resonance imaging (MRI) data contains enormous amounts of information about brain structure and function. To address the challenges of neuroimaging data analysis -- the high dimensionality of the data, the scale of modern neuroimaging datasets, and its inherent complexity, we use a range of data science (DS) and artificial Intelligence (AI) methods. In our hands, these methods enable scalable research pipelines, uncover complex relationships in data, and extract new kinds of biological information from existing data. I will demonstrate these applications of DS/AI with a few projects that my group has executed in the last few years. One of the challenges of these methods is their lack of transparency and interpretability, and I will also show how we address these challenges using methods for interpretation of AI models. Finally, new AI models and methods promise to represent generalizable knowledge about brain structure and function that can be applied to many different questions. I will discuss ongoing and future work that aims to deliver on this promise.
Ariel Rokem received a Bachelors and Masters degree in Biology and Cognitive Psychology at the Hebrew University of Jerusalem (2002 and 2005). He then received a PhD in neuroscience from UC Berkeley (2010) and additional postdoctoral training in computational neuroimaging at Stanford (2011 - 2015). He was a Senior Data Scientist at the University of Washington eScience Institute (2015-2020), before joining the faculty of the Department of Psychology in 2020. His group (https://neuroinformatics.uw.edu/) develops computational tools to study the biological basis of brain function and applies them to a variety of research questions.
