Characterising insomnia: A graph spectral theory approach.

Area of Science:

• Computational neuroscience
• Graph theory
• Sleep science

Background:

• Insomnia diagnosis relies on subjective reports and clinical observation.
• Objective, quantitative methods for characterizing insomnia are needed.
• Sleep onset period analysis offers a potential window into sleep disorders.

Purpose of the Study:

• To develop and validate a computational method for differentiating insomniacs from healthy controls.
• To apply graph spectral theory to sleep stage transition networks.
• To identify robust quantitative markers for insomnia.

Main Methods:

• Expert-generated hypnograms of sleep onset periods were used to construct sleep stage transition networks.
• Four similarity distances were computed based on graph spectral properties.
• Statistical analysis was performed to compare healthy and insomnia cohorts (n=16 each).
• Eigenvalue determination was central to calculating key similarity distances (d1, d4).

Main Results:

• Similarity distances derived from eigenvalues (d1 and d4) showed significant differences between groups.
• d1 successfully characterized 93% of insomniacs.
• d4 successfully characterized 87% of insomniacs.
• These measures proved reliable and robust for insomnia characterization.

Conclusions:

• Graph spectral theory provides a powerful computational framework for objective insomnia assessment.
• Eigenvalue-based similarity distances offer promising biomarkers for distinguishing insomnia.
• This approach could enhance the diagnostic accuracy and understanding of sleep disorders.