Abstract— We present a novel closed-loop subthalamic nucleus (STN) deep brain stimulation (DBS) scheme for Parkinson’s disease (PD) and obsessive-compulsive disorder (OCD). The algorithm is designed to effectuate real-time, adaptive stimulation employing the outcome of the 0-1 test for chaos as a state-specific biomarker. In case of a null outcome, the system identifies optimal patterns of stimulation desynchronizing pathologic neuronal activity with minimal energy consumption, on grounds of a stochastic dynamical model and an appropriately formulated cost function. Simulations are performed utilizing microelectrode recordings (MERs) acquired during 8 and 2 DBS surgical interventions for PD and OCD, respectively. I. INTRODUCTION It is becoming increasingly evident that standard continuous open-loop deep brain stimulation (DBS) cannot be optimally adjusted to the long-term post-operative dynamic nature of movement and neuropsychiatric disorder symptoms, which, depending on the effect of medication, may fluctuate over various time-scales. On these grounds, the concepts of „closed-loop‟, „feedback controlled‟, or „adaptive‟ stimulation are interchangeably emerging as powerful „on-demand‟ alternatives, constituting the most revolutionary scenario in the modern era of DBS [1,2]. In a closed-loop DBS system, delivery of stimulation is modulated utilizing specific biomarkers that capture the patient‟s clinical state in „real-time‟ [3]. In a pilot clinical study, feedback controlled stimulation based on local-field potential (LFP) beta power has been proven to be clinically more effective than standard stimulation [4]. Nevertheless, specific drawbacks, including the relatively low complexity dynamics captured by this signal, point to the necessity for more sensitive biomarker approaches [5, 6]. In this context, S D Karamintziou, G L Tsirogiannis, N G Deligiannis, G E Polychronaki and K S Nikita are with National Technical University of Athens, Athens, Greece(emails:skaram@biosim.ntua.gr ,georgioschi@gmail.com ,artificial21 00@hotmail.com , layiapol@gmail.com, knikita@ece.ntua.gr ,). B Piallat and O David are with Grenoble University and Grenoble Institute of Neurosciences (INSERM-U836), Grenoble, France (e-mails: brigitte.piallat@ujf-grenoble.fr , Olivier.David@inserm.fr ). S Chabardès and M Polosan are with INSERM-U836 and University Hospital of Grenoble, Grenoble, France (e-mails: SChabardes@chu- grenoble.fr , MPolosan@chu-grenoble.fr ). P G Stathis is with Mediterraneo Hospital, Glyfada, Athens, Greece (e- mail: stathis.pantelis@gmail.com ). G A Tagaris is with „G. Gennimatas‟ General Hospital of Athens, Athens, Greece (e-mail: tagaris@otenet.gr ) E J Boviatsis is with Attikon University Hospital, Athens, Greece (e- mail:eboviatsis@gmail.com ). D E Sakas is with „Evangelismos‟ General Hospital, Athens, Greece (e- mail: sakasde@med.uoa.gr ). model-based control systems provide a powerful alternative in the determination of optimal feedback control parameters [5]. Design of these systems has been primarily oriented to identification of novel stimulation waveforms on grounds of computational modeling or stochastic phase models of neural population activity [7-9]. Optimization of DBS settings is expected to generate maximal clinical benefit at the lowest possible power consumption, resulting in reduction of stimulation-related complications and longer battery life. In this paper, we present a novel configuration of a closed-loop subthalamic nucleus (STN)-DBS system for Parkinson‟s disease (PD) and obsessive-compulsive disorder (OCD) (Fig. 1), in light of evidence that disruption of neural synchronization in this nucleus may be related to the therapeutic mechanism of action of stimulation [10,11]. The system operates according to an online real-time algorithm that integrates a sensitive biomarker approach, the 0-1 test for chaos [12], and an improved modification of a stochastic dynamical phase model [13]. Principally, the 0-1 test for chaos is a binary test for definitely distinguishing chaotic from regular dynamics in deterministic dynamical systems, and may therefore be superior to the „dubious‟ LFP beta power as a neurophysiological biomarker of the pathological state. Herein, we validate the outcome of this test as a control parameter in the proposed closed-loop DBS scheme and its potential to support „on-demand‟ stimulation. In case of a null outcome, optimal stimulation patterns may be determined on grounds of the stochastic phase model [13] and an appropriately designed cost function for minimum energy desynchronization of the pathologic neuronal activity. Preliminary results are obtained on the basis of microelectrode recordings (MERs) acquired during 8 and 2 DBS surgical interventions for PD and OCD, respectively [13, 14]. II. PATIENTS AND METHODS A. Patients During an 1-year period, 8 patients with idiopathic PD underwent STN-DBS at the Department of Neurosurgery, at Evangelismos General Hospital of Athens. Informed consent was provided by each patient. Stereotactic surgery and patients‟ clinical characteristics have been described in detail elsewhere [13]. A commercially available MER system (Leadpoint TM Neural Activity Monitoring System, Medtronic Inc., Minneapolis, MN) was used intraoperatively for data acquisition. Secondly, during an 1- year period, 2 patients with treatment-refractory OCD underwent bilateral STN-DBS at the Grenoble University Hospital. Informed consent was provided by each patient, while strict ethical Design of a Novel Closed-Loop Deep Brain Stimulation System for Parkinson’s Disease and Obsessive-Compulsive Disorder S D Karamintziou, B Piallat, S Chabardès, M Polosan, O David, G L Tsirogiannis, N G Deligiannis, P G Stathis, G A Tagaris, E J Boviatsis, D E Sakas, G E Polychronaki, and K S Nikita 7th Annual International IEEE EMBS Conference on Neural Engineering Montpellier, France, 22 - 24 April, 2015 978-1-4673-6389-1/15/$31.00 ©2015 IEEE 860