INSTITUTE OF PHYSICS PUBLISHING JOURNAL OF NEURAL ENGINEERING J. Neural Eng. 1 (2004) 165–173 PII: S1741-2560(04)83976-4 Adaptive autoregressive identification with spectral power decomposition for studying movement-related activity in scalp EEG signals and basal ganglia local field potentials Guglielmo Foffani 1,2 , Anna M Bianchi 1 , Alberto Priori 3 and Giuseppe Baselli 1 1 Department of Biomedical Engineering, Politecnico di Milano, 20133 Milano, Italy 2 School of Biomedical Engineering, Science and Health Systems, Drexel University, Philadelphia, PA 19104, USA 3 Department of Neurological Sciences, Universit` a di Milano, IRCCS Ospedale Maggiore di Milano, 20122 Milano, Italy E-mail: guglielmo.foffani@polimi.it or guglielmo.foffani@drexel.edu Received 28 July 2004 Accepted for publication 25 August 2004 Published 10 September 2004 Online at stacks.iop.org/JNE/1/165 doi:10.1088/1741-2560/1/3/006 Abstract We propose a method that combines adaptive autoregressive (AAR) identification and spectral power decomposition for the study of movement-related spectral changes in scalp EEG signals and basal ganglia local field potentials (LFPs). This approach introduces the concept of movement-related poles, allowing one to study not only the classical event-related desynchronizations (ERD) and synchronizations (ERS), which correspond to modulations of power, but also event-related modulations of frequency. We applied the method to analyze movement-related EEG signals and LFPs contemporarily recorded from the sensorimotor cortex, the globus pallidus internus (GPi) and the subthalamic nucleus (STN) in a patient with Parkinson’s disease who underwent stereotactic neurosurgery for the implant of deep brain stimulation (DBS) electrodes. In the AAR identification we compared the whale and the exponential forgetting factors, showing that the whale forgetting provides a better disturbance rejection and it is therefore more suitable to investigate movement-related brain activity. Movement-related power modulations were consistent with previous studies. In addition, movement-related frequency modulations were observed from both scalp EEG signals and basal ganglia LFPs. The method therefore represents an effective approach to the study of movement-related brain activity. 1. Introduction The complex scenario of event-related spectral changes induced on scalp EEG signals by the execution of simple movements has been studied for decades [1]. The advent of deep brain stimulation (DBS), an innovative neurosurgical procedure to treat Parkinson’s disease, recently offered the opportunity to record local field potentials (LFPs), i.e. deep EEG signals, from the basal ganglia in living humans [2–4]. The study of movement-related spectral changes, therefore, has been deepened into the globus pallidus and subthalamic nucleus, primary DBS targets [5–10]. 1741-2560/04/030165+09$30.00 © 2004 IOP Publishing Ltd Printed in the UK 165