Decoding pre-movement neural activity from thalamic LFPs for adaptive neurostimulation in tremor patients

Understanding the neural mechanisms underlying movement initiation is crucial for advancing movement-driven adaptive deep brain stimulation therapies for tremor disorders. We investigated the feasibility of decoding pre-movement periods of upper limb movements by machine learning using thalamic local field potentials (LFPs) and scalp electroencephalography (EEG) signals. Data were analysed from 11 patients undergoing deep brain stimulation surgery, employing machine learning models—including logistic regression, gradient-boosted decision trees, and convolutional neural networks—to distinguish rest periods from pre-movement periods. Our results demonstrate that early neural correlates can predict movement onset, achieving above-chance decoding performance starting approximately 680 ms before movement execution using thalamic LFPs and 1.09 s using EEGs. Individualized, patient-specific decoders outperformed cross-patient models, reflecting substantial inter-patient variability in neural modulatory patterns. Additionally, multiple frequency bands contributed independently to decoding performance, emphasizing the importance of incorporating a broad spectrum of frequencies rather than relying solely on single canonical bands. These findings underscore the value of personalized, multi-band machine learning approaches in capturing the neural correlates preceding movement. They support the development of adaptive neurostimulation therapies that enhance the effectiveness of clinical interventions through tailored models that account for patient-specific neural activity.