The group is working.

Modeling Music Auditory Network Under Alzheimer’s-Like Perturbation

Summer 2026 · BU RISE

Using a 379-region virtual brain model, we investigated why familiar music can remain recognizable even as Alzheimer’s disease disrupts other forms of memory. We compared semantic and episodic music-processing networks to identify neural pathways that may remain resilient.

Abstract

Alzheimer’s disease (AD), characterized by gradual loss of memory and cognitive function, affects over 30 million people worldwide. Many patients are unable to recognize familiar concepts, people, and places, but recognize familiar music. Specifically, experiments have shown that despite deficits in episodic musical memory (e.g., the context in which music was heard), semantic musical memory (e.g., the melody) remains unimpaired; however, the neurological mechanisms underlying this phenomenon remain poorly understood. We hypothesize that changes in functional connectivity (FC) of music processing pathways induce differences in semantic and episodic musical memory capacity, with the episodic musical network regressing across healthy control (HC), mild cognitive impairment (MCI)-like, and AD-like groups, and the semantic musical network remaining comparatively stable across these same groups.. Plaque formation due to AD alters FC, particularly in brain regions associated with episodic memory, like the posterior cingulate cortex, postero-medial cortex, and precuneus. Regions associated with semantic memory–the anterior temporal, inferior, and supero-medial prefrontal cortices–develop AD later. To assess the changes in these networks as AD-like perturbation increases, we use the Virtual Brain (TVB) to simulate a 379-region brain network. Following a previous study, we transform a surrogate amyloid map to produce AD-like perturbation conditions in TVB. We stimulated primary auditory cortices bilaterally with a 100-ms constant-amplitude pulse, a 2 Hz signal, and a 5 Hz signal, maintaining total input power. We subtracted the unstimulated output, then normalized FC, transfer gain, and response latency per network. We reported the differences in FC, gain, and latency between the networks using interaction analysis. Across 20 paired initializations, we tested the semantic-versus-episodic interaction contrast against the null hypothesis using a two-sided one-sample t-test. This work has the potential to identify circuits that remain viable targets for music-based interventions in AD patients, providing potential avenues for novel treatment strategies.

Poster