Skip to content

Minicourse EDDA-AL01

Cedalion: A Python-based Framework that integrates multimodal fNIRS & DOT Data Analysis with Machine Learning Approaches

Alexander von Lühmann

with Eike Middell, Meryem Yücel, Laura Carlton, David Boas

About this course

Join us for a workshop on Cedalion, a new Python-based toolbox designed for advanced multimodal fNIRS analysis and Diffuse Optical Tomography (DOT) that can be collaboratively expanded by the community. The workshop aims to familiarize researchers with Cedalion’s growing functionality that aims to integrate core features of HOMER3 and AtlasViewer for fNIRS data analysis and image reconstruction while tapping into the rich Python ecosystem of modern tools for data analysis, machine learning, and multimodal integration.

Cedalion provides user-extensible data structures that allow for easy exchange with frameworks such as PyTorch, SciKit-Learn and Pandas to simplify the integration of conventional fNIRS data processing streams and machine learning workflows. Furthermore, the toolbox relies on the BIDS and SNIRF standards to support the integration and data exchange with other toolboxes designed for the preprocessing and analysis of physiological recordings, behavioral measures and other neuroimaging modalities, further facilitating multimodal analysis.

Participants will gain theoretical knowledge and hands-on experience in utilizing Cedalion for both channel- and image-space analysis, encompassing preprocessing, General Linear Model (GLM) analysis, optode co-registration, and DOT image reconstruction. By the end of the workshop, participants will also have learned how to construct custom script-based analysis pipelines and how to leverage Cedalion for single-trial analysis and machine learning on fNIRS data.

Learning Outcomes

  • How to use Cedalion for channel- and image-space analysis.
  • Preprocessing, GLM, optode co-registration, image reconstruction.
  • Creating custom script-based analysis pipelines. 

Course Plan

Level: Introductory

Pre-Requisites: course requires software installation (click here for more information)
Course Duration: 3 hours

Delivery Plan

30 min: Introduction and Vision, Initial Set Up
60 min: Channel-Space Analysis: Preprocessing and GLM
15 min: Break
60 min: Image-Space Analysis: Head models, co-registration and image reconstruction
15 min: Outlook and Demo – Single Trial Analysis and Machine Learning

Why enrol on this course?

Cedalion is a modular and open-source Python package that combines much of the key functionality of HOMER2/3 and AtlasViewer with widely available and powerful packages for data-driven analysis and machine learning. It uses BIDS and SNIRF standards to interface with other software. The goal of this workshop is to educate participants about Cedalion’s features and to teach them how to create their own scripting-based fNIRS/DOT preprocessing and analysis pipelines.