Discussion
The simulation results presented here demonstrate the potential of the adaptive CRT device with dynamic optimization of AV and VV intervals for improving cardiac output beyond the current non-adaptive CRT devices.
The improvement in the simulated cardiac output shown in Figure 4, as a result of online optimization of the AV delay and VV interval according to simulated hemodynamic sensors, is the primary result demonstrated in this study. Hemodynamics are influenced by many variables, activity being one of the most significant. The model reported suggests that an online adaptive therapeutic device that stimulates the heart and auto adapts in response to the hemodynamic variation can improve hemodynamic performance considerably and deliver optimal and consistent CRT. Performance at higher heart rates is especially important in CHF patients, who tend to be tachycardic even at low exercise levels. The present simulation report demonstrates the potential improvement of cardiac output with adaptive CRT device at higher heart rates. This may therefore translate to an improvement in exercise capability and quality of life of CHF patients with a future adaptive CRT device.
The adaptive CRT device working in a closed loop with hemodynamic sensors will enable a clinician to optimize pacing intervals and to select a better lead position using a responder to the CRT online diagram shown in Figure 6. Using this diagram the clinician might be able to convert a "nonresponder" into a "responder" and a "responder" into a "better responder."[4]
The adaptive CRT device is expected to simplify and hence to shorten the time required by the clinician during a patient's follow-up. Since the device has built-in autoprogrammability capabilities the clinicians will primarily need to review stored histograms and verify the appropriateness of automatic features.
Dynamic optimization of the AV interval in response to hemodynamic sensors will also be beneficial to patients with a dual chamber pacemaker, since it is well known that phsyiologically the AV delay is rate-dependent. Since dynamic optimization enables more physiologic pacing, it is expected to be beneficial to all future pacemakers and defibrillators after clinical benefit is proven.
The adaptive CRT device shown in the block diagram in Figure 1, adds a "spiking neuron network co-processor" to the existing microcontroller. This architecture has the advantage of the combined deterministic algorithmic module (the microcontroller) and a learning module, the spiking neuron network. The microcontroller master trains "online" the learning module that is responsible for adjusting to the patient's intrinsic hemodynamics. The adaptive CRT device architecture shown in Figure 1 enables minimal changes to the existing CRT device microcontroller code that will ease the integration and regulatory efforts.
CRT device management and more specifically the management of the AV delay and VV intervals are especially suitable for neural network processing given the marked interindivudal and intraindividual variation in these parameters. Neural network are able to process the hemodynamic sensors biological signals using pattern recognition techniques and to perform a feedback control task dynamically and online.[5] Neural networks are a different paradigm for computing. They are based on parallel architecture with simple processing elements and a high degree of interconnection. Neural networks are known to have advantages over standard algorithmic processing in performing tasks such as adaptive control and pattern recognition.[5] Spiking neural networks architectures are a unique form of neural networks that are inspired by the biological nervous system.[6] The adaptive CRT device described here is based on a novel spiking neural network architecture and learning rule developed specifically for adaptive CRT devices and for closed loop therapeutic medical devices more generally.[7,8]
A primary and novel feature of the spiking neural network architecture, used here for the adaptive CRT device, is the combination of a spiking neural network architecture with the Hebbian learning rule to perform an online feedback control task combined with a pattern recognition task implemented to the hemodynamic sensor signal. The second important feature of the spiking neuron network is its ability to perform both tasks with extremely low power dissipation or current drain. Low power dissipation is a crucial point for implanted pacemakers and defibrillators devices that need to operate 57 years with a battery power source only.
In the next phase of investigations the adaptive CRT device will be tested in an acute animal experiment with implanted pressure sensors in the right and left ventricles and with sono crystals implanted in the left ventricular wall. Later, in planned chronic animal experiment investigations and clinical studies the ideal hemodynamic sensor will be defined. Whether the ideal sensor is indeed stroke volume as derived from intracardiac impedances as used in the in vitro model described,[9] intraventricular pressure sensing as performed with the Medtronic Chronicle™ device,[3] or wall motion sensing as performed with SORIN peak endocardial accelerometer (PEA™) sensor is yet to be determined.
Whereas the greatest potential use of the neural network processor is when integrated into an implanted cardiac rhythm device with continuous interval adaptation, a similar external adaptive CRT device can be developed. In the external adaptive CRT device, the spiking neuron network learning module will be utilized externally to the implanted CRT device that will telemetrically receive the IEGMs and hemodynamic sensors' data and transmit back the optimal AV and VV intervals to the implanted device. The external device will be used to optimize an implanted CRT device, to identify responders to CRT and may reduce the time and cost of the follow-up procedure for CRT patients.
Pacing Clin Electrophysiol. 2005;28(11):1168-1173. © 2005 Blackwell Publishing
Cite this: Adaptive Cardiac Resynchronization Therapy Device: A Simulation Report - Medscape - Nov 01, 2005.
