Adaptive Cardiac Resynchronization Therapy Device: A Simulation Report RAMI ROM,* JACOB EREL,† MICHAEL GLIKSON,‡ KOBI ROSENBLUM,£ RAN GINOSAR,§ and DAVID L. HAYES¶ From the *AI Semi Ltd, Granot, †Meir Hospital, Kfar-Saba, ‡Sheba Medical Centre and Tel Aviv University, £Neurobiology and Ethology, Haifa University, Haifa, §Electrical Engineering, Technion, Haifa, Israel, and ¶Mayo Clinic, Rochester, New York, USA ROM, R., ET AL.: Adaptive Cardiac Resynchronization Therapy Device: A Simulation Report. We re- port the results of a simulation of an adaptive cardiac resynchronization therapy (CRT) device performing biventricular pacing in which the atrioventricular (AV) delay and interventricular (VV) interval parameters are changed dynamically in response to data provided by the simulated IEGMs and simulated hemody- namic sensors. A learning module, an artificial neural network, performs the adaptive part of the algorithm supervised by an algorithmic deterministic module, internally or externally from the implanted CRT or CRT-D. The simulated cardiac output obtained with the adaptive CRT device is considerably higher (30%) especially with higher heart rates than in the nonadaptive CRT mode and is likely to be translated into improvement in quality of life of patients with congestive heart failure. (PACE 2005; 28:1168–1173) cardiac resynchronization therapy, neural network Introduction Cardiac Resynchronization Therapy (CRT) is currently an established therapy for patients with congestive systolic heart failure and intraventric- ular electrical or mechanical conduction delays. It is based on synchronized pacing of the atrium and the two ventricles. 1 The resynchronization task de- mands exact timing of the cardiac chambers so that the overall stroke volume is maximized for any given heart rate (HR). Optimal timing of activation of the atrium and the right and left ventricles is one of the key factors in determination of the car- diac output. The timing parameters that are pro- grammable in a CRT device that determines the pacing intervals are the atrioventricular (AV) de- lay and interventricular (VV) interval. Clearly, op- timizing resynchronization is patient dependent as well as time and activity dependent. Intuitively, the best combination of pacing time intervals that restores optimal synchrony will change consider- ably during normal daily activities. CRT devices must be individually optimized, “fine-tuned” to provide optimal benefits. In addition to being time consuming and expensive, echo-guided AV and VV interval programming is limited to a static, sedentary activities. The impact of HR, body po- sition, medications, and many other variables on these programmable variable is unknown. Deter- mining these programmable variables should ide- ally be automatic and adaptive to the patient’s activities. Address for reprints: Dr. Rami Rom, AI Semi Ltd, M.P. Hefer, Granot 38100, Israel. Fax: +972-4-632-1392; e-mail: rami@ AISemi.com Received November 29, 2004; revised March 3, 2005; accepted June 26, 2005. The next generation of CRT devices should have online adaptive capabilities determined by hemodynamic performance. There are no clinical trial data to date sys- tematically assessing the effect of optimized ver- sus nonoptimized programming of CRT devices. However, the difficulty and importance of optimal programming of CRT devices is often noted. 3,4,5 Optimization of a CRT device by continuous hemodynamic monitoring using the Medtronic Chronicle TM device has recently been reported. 3 In this case report, the authors demonstrated that continuous hemodynamic monitoring provided useful information for optimization of the AV de- lay. This article takes this to the next level by using a simulation where both the AV delay and the VV interval are changed dynamically by a simulated adaptive CRT device according to hemodynamic sensors. Hence, the hypothesis of the adaptive CRT device presented here is that dynamic optimiza- tion of the AV delay and VV interval according to feedback from hemodynamic sensors, and man- aged by a combined controller and neural network processor, may significantly improve heart failure patients’ quality of life and general well-being. Methods We use a closed simulation environment that simulates the intrinsic cardiac electrical and hemodynamic behaviors and an adaptive CRT pacemaker. The simulated adaptive CRT pace- maker is designed to process the simulated heart module electrical and hemodynamic inputs and to deliver optimized CRT pacing to the simu- lated heart. The goal of the simulated systems is to demonstrate the importance of dynamically 1168 November 2005 PACE, Vol. 28