Simulation-Driven fNIRS using NIRFASTer: Modeling Hemodynamic Responses, Realistic Noise, and Diffuse Optical Tomography
Instructors: Jiaming Cao, Hamid Dehghani
About this course
This course introduces the principles of functional near-infrared spectroscopy (fNIRS) through computational modeling and simulation, with a focus on generating realistic optical neuroimaging datasets. The training combines theory and hands-on exercises to illustrate how physiological signals arise and how they can be recovered using computational models.
Participants will build forward models of realistic head geometries to simulate light transport using NIRFASTerFF. Within these models, localized brain activations are introduced by defining hemodynamic responses based on canonical and parameterized hemodynamic response functions (HRFs). By modulating changes in oxy- and deoxy-hemoglobin concentrations over time, participants generate controlled ground-truth activation patterns and examine how HRF shape, amplitude, and timing influence measured fNIRS signals.
To approximate real experimental conditions, these models will incorporate physiologically and instrumentally added noise, including systemic oscillations (e.g., cardiac and respiratory signals), low-frequency drift, motion artifacts, and detector noise. This approach enables attendees to explore how signal-to-noise ratio, channel geometry, and preprocessing choices affect data quality and interpretation.
What will you learn?
- Basic theories behind forward data generation, including photon transport and parameterized HRF
- How to generate forward models from segmented brain atlases and given optode montages using NIRFASTerFF
- How to simulate realistic fNIRS signals given a desired activation location and HRF. Participants will also learn how to simulate various types of – physiological and instrumental noise
- How to save the simulated results in .snirf format for processing and sharing.
Course details
Course duration
3 hours
Prerequisites and/or other notes
Participants should bring their own laptops and have Python and the NIRFASTerFF package installed prior to the minicourse. Environments that support Jupyter notebooks (e.g. Jupyter or VSCode) are required. The NITFASTerFF package can be downloaded here: https://github.com/milabuob/nirfaster-FF All notebooks and tutorials will be made available.
Delivery plan
- Lecture: Introduction to tissue optics & course overview (15min)
- Lecture: Introduction to the NIRFAST-family softwares (15min)
- Hands-on: Head model generation (30min)
- Lecture: Modeling of focal activation and noise modeling (15min)
- Hands-on: fNIRS data simulation in realistic head models (45min)
- Lecture: Basics of parameter recovery and image reconstruction (15min)
- Hands-on: Simple reconstruction and its visualization; saving data in BIDS-compatible formats (30min)
- Q&A: 15min
Why should you enroll in this course?
- To understand the basic principles of light transport in tissue
- To understand the basic operations of the NIRFASTerFF
- To generate physiologically-relevant forward data using computational models
- To perform simple fNIRS/DOT parameter recovery and visualize the results