Developing predictive imaging biomarkers using whole-brain classifiers: Application to the ABIDE I dataset
Abstract
Within clinical neuroimaging communities there is considerable optimism that functional magnetic resonance imaging (fMRI) will provide much needed objective biomarkers for diagnosing and tracking the severity of psychiatric and neurodevelopmental disorders (Castellanos et al. 2013). Training classifiers to predict disease state and severity that are robust not only to the considerable heterogeneity present in these disorders, but also to variation in systems and protocols used to collect fMRI data, require very large and diverse training datasets. The Autism Brain Imaging Dataset Exchange (ABIDE) is addressing this need for autism spectrum disorders (ASD) by aggregating data collected from imaging studies collected at 17 different sites (Di Martino et al. 2013). To learn more about applying machine learning methods to develop fMRI-based biomarkers of disease, the goal of our Neurohackweek 2016 project was to build a modular, open-source analysis tool for training and testing whole-brain classifiers to predict clinical diagnoses. To do so we leveraged existing machine-learning technologies implemented in the Python programming language (scikit-learn Pedregosa et al. 2011) to create a simple, but flexible command-line program and tested our software using the ABIDE I preprocessed dataset. The prototype completed during Neurohackweek uses a logistic regression based classifier, but was designed to be easily adapted to other classifier models.
