Classification of fMRI-based cognitive states

Stephen LaConte
Baylor College of Medicine

Multivoxel pattern analysis uses supervised learning approaches to obtain classifiers that can take neuroimaging data and predict the sensory/behavioral conditions corresponding to the time the images were acquired. In addition to how predictive these classifiers are, their structure can provide insights about the data they model. An important goal is to know your data and understand the analysis being performed. This talk provides some general approaches, but also uses the support vector machine and linear discriminant analysis as two examples of classification methods to discuss topics such as data representation, impact of preprocessing, limitations imposed by the hemodynamic response, the relationship between class labels and the stimulus paradigm, and interpretations that can be extracted from multivariate models.

Audio (MP3 File, Podcast Ready)

Back to Summer School: Mathematics in Brain Imaging