2023/05/15 by Sean Paulsen, Paulsen, Sean, Lloyd May +3
Computer Science · Neuroscience · #FOS: Biological sciences #Hearing Loss and Rehabilitation #Music and Audio Processing #Neural dynamics and brain function #Neurons and Cognition (q-bio.NC) #Neuroscience and Music Perception #Speech and Audio Processing
paper · pdf · doi:10.48550/arxiv.2305.08987
openalex publication_date 2023/05/15 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
Stimulus decoding of functional Magnetic Resonance Imaging (fMRI) data with\nmachine learning models has provided new insights about neural representational\nspaces and task-related dynamics. However, the scarcity of labelled\n(task-related) fMRI data is a persistent obstacle, resulting in\nmodel-underfitting and poor generalization. In this work, we mitigated data\npoverty by extending a recent pattern-encoding strategy from the visual memory\ndomain to our own domain of auditory pitch tasks, which to our knowledge had\nnot been done. Specifically, extracting preliminary information about\nparticipants' neural activation dynamics from the unlabelled fMRI data resulted\nin improved downstream classifier performance when decoding heard and imagined\npitch. Our results demonstrate the benefits of leveraging unlabelled fMRI data\nagainst data poverty for decoding pitch based tasks, and yields novel\nsignificant evidence for both separate and overlapping pathways of heard and\nimagined pitch processing, deepening our understanding of auditory cognitive\nneuroscience.\n