Higher-order Sonification of the Human Brain

Area of Science:

• Neuroimaging
• Data Sonification
• Computational Neuroscience

Background:

• Sonification, translating data into sound, is emerging for science communication and accessibility.
• Current sonification methods are limited to one-dimensional data, hindering multi-dimensional analysis.
• Analyzing complex datasets like brain MRI requires advanced techniques for effective sonification.

Purpose of the Study:

• To develop and validate a sonification technique for multi-dimensional structural MRI data.
• To assess the accuracy of brain age estimation using sonified bispectral data.
• To explore the potential of sonification in differentiating age groups within a brain imaging dataset.

Main Methods:

• Structural MRI data from the OASIS-3 dataset were analyzed using Fourier space statistical measures (power spectrum, bispectrum).
• Reduced bispectra were computed to capture higher-order spatial correlations in 3D voxel intensity distributions.
• Sonification was applied to ensemble-averaged bispectral outputs for five distinct age groups.

Main Results:

• A brain age regression model achieved a mean absolute error of 4.7 years.
• Sonification of the data revealed clear auditory differentiations between age groups.
• Reconstruction of bispectral data from sonified signals showed minimal information loss.

Conclusions:

• The developed sonification approach effectively encodes multi-dimensional brain MRI data into an auditory format.
• This method offers a robust and quantitative way to analyze complex datasets, improving accessibility for visually impaired individuals.
• Sonification of bispectral features provides a promising avenue for exploring neuroimaging data and identifying age-related brain changes.