Multivariate pattern analysis (MVPA) for fNIRS: Advancing from “Where in the brain?” to “What’s in the brain?”

Benjamin Zinszer1,2, Anna Herbolzheimer1, Vikranth Bejjanki3, Lauren Emberson1,

1 Princeton University, 2 University of Delaware, 3 Hamilton College

Duration: 180 min

Capacity: 30

Level: Introductory

Analysis Method

Requirements: Laptop; Matlab and toolbox pre-installed

Synopsis: This educational workshop introduces multivariate pattern analysis (MVPA) methods for analyzing and interpreting neuroimaging data, with a special focus on how MVPA can be adapted for use with fNIRS data. Despite the increasing popularity of MVPA in fMRI, this approach has seen limited application in fNIRS beyond experiments with brain-computer interface technology. Over a three hour session, we will provide a tutorial on how and why researchers might implement multivariate pattern analyses with fNIRS to answer questions about cognition or development. We begin with a non-technical introduction to MVPA and classification approaches in general. We will survey how these approaches have been practically applied in the current literature, and the kinds of questions that MVPA approaches may be best suited for. Then we will lead a semi-technical discussion of how a classification experiment is designed and the resulting data prepared for analysis, including hands-on demonstration of multiple kinds of classifiers using previously published datasets and our open-source software in Matlab.

This course is intended for all fNIRS researchers, regardless of programming experience, and will include conceptual discussions, simple scripting demonstrations, and opportunities to ask technical questions.

Course organization:

  • Introduction to MVPA at the conceptual level, typical research questions, basic terms (feature, classification, training-vs-testing, etc.)
  • Overview of types of classifiers used in cognitive neuroscience (especially examples of classifiers already published in fNIRS literature) and their relative strengths/weaknesses
  • Hands-on demonstration with the code to illustrate analytic decisions and how they may affect results

Requirements: You are expected to bring a laptop with Matlab and the required code installed.