Robust-tedana: An automated denoising pipeline for multi-echo fMRI data

Area of Science:

• Neuroimaging
• Biomedical Engineering
• Data Science

Background:

• Multi-echo fMRI improves neural signal separation from noise.
• Current denoising methods like tedana can be inconsistent, requiring manual intervention.
• This limits scalability for large-scale and automated processing.

Purpose of the Study:

• Introduce Robust-tedana, an optimized automated denoising pipeline for multi-echo fMRI data.
• Enhance the reliability and consistency of denoising across single-subject and group analyses.
• Facilitate advanced analysis of Multi-Band Multi-Echo (MBME) fMRI data.

Main Methods:

• Incorporation of Marchenko-Pastur Principal Component Analysis (MPPCA) for thermal noise reduction.
• Utilization of robust independent component analysis for stable signal decomposition.
• Implementation of a modified component classification process for improved accuracy.

Main Results:

• Robust-tedana demonstrated consistent performance at both single-subject and group levels.
• The pipeline mitigated erroneous attenuation of genuine task activation.
• Increased magnitude of group-wise effects was observed compared to conventional methods.

Conclusions:

• Robust-tedana provides a reliable and automated solution for multi-echo fMRI denoising.
• It overcomes limitations of existing methods, enabling large-scale and clinical research.
• Facilitates advanced analysis of MBME fMRI data, including individual clinical assessments.