Quantifying Neural Stability: Validation of a New Brain Stability Index

Abstract

Quantifying longitudinal changes in an individual's brain is central to the development of personalised neural biomarkers in neurology and psychiatry. However, existing approaches for characterising individual neurophysiological signatures focus on discrimination between people rather than the quantification of within-subject change. To address this, we introduce the Brain Stability Index (BSI), a whole-brain metric that quantifies the similarity between two longitudinal neurophysiological scans in a low-dimensional latent space, with reference to a normative magnetoencephalography (MEG) database. Using 276 open resting-state MEG datasets and matched synthetic data, we first characterise how finite test-retest reliability sets a noise floor on the BSI. We then demonstrate that the BSI is sensitive to graded changes in whole-brain neural change that extend beyond measurement variability. Finally, we show that Factor Analysis, by separating shared structure from feature-specific noise, makes the BSI more robust to measurement artefacts. Together, these findings establish the BSI as a robust, bounded measure of neural stability that is well suited to longitudinal monitoring in neurology and psychiatry.