Project description

State of the art implantable pulse generators (IPGs) or cardiac pacemakers and implantable cardio defibrillators (ICDs) include real-time sensing capabilities reflecting the state of the heart. This real-time sensing is performed by the front-end of the pacemaker, also known as the `sense amplifier.' Since in IPGs and ICDs the information retrieved by the front-end is reduced to a single event only signaling the occurrence of cardiac depolarization (corresponding with a heart chamber contraction), morphological attributes of the electrogram are completely suppressed and therefore not taken into account. Hence, today's pacemakers or ICD's cannot discriminate between, e.g., lethal tachyarrhythmias and supraventricular tachycardia that may occur as a result of healthy exercise. This results in a large number of unnecessary therapies delivered to a patient's heart and thereby reduced battery and thus pacemaker life.

Recently, research and clinical studies report details on how morphological aspects of the electrogram relate to various pathological states of the heart and on how the wavelet transform (WT) can contribute efficiently to analyzing these. Furthermore, it turns out that the wavelet transform can significantly reduce the amount of computations necessary to perform the morphological analysis and thus requires substantially less power than conventional analysis techniques.

From a clinical viewpoint the QT-interval is a measure of interest. If the QT-interval has an abnormal length, there is a risk of developing ventricular arrythmias. The QT-interval is also of interest for the research on ventricular repolerization. As an application the BioSens aims at determing the QT-interval in cardiac signals.

As in pacemakers and implantable cardio defibrillators the recorded electrograms need to be analyzed in real time, mathematical modeling of their most characteristic features is required and thus a sufficiently large set of empirical data of cardiac signals needs to be available, encompassing the whole range of relevant signals, i.e. pathological and normal, and spanning the entire spectrum of relevant morphological features. This data is provided by the industry partners of the project. Further, these features need to be related to the identified cardiac states, such as pathologies and anomalies. Using the wavelet transform, the multifractal spectrum of a signal can be derived, providing a solid basis for a robust classification of the incoming electrogram. However wavelet can be designed for a given signal and application. These optimized wavelets can be used to detect a certain morphology in the signal. Approaches for designing these optimal wavelets will have to be developed.

To date, the wavelet transform has always been implemented in a digital fashion. Unfortunately, the relatively high power consumption associated with the required analog-to-digital converter and the limited available energy budget exclude the implementation of the wavelet transform in pacemakers and ICDs by means of digital signal processing. Recently, we have succeeded in implementing, for the first time, the wavelet transform in the analog domain by employing dynamic translinear circuits, which paves the way to the implementation of a low-power WT-based signal- processing platform for real-time sensing of cardiac signals in pacemakers and ICDs.

In this project, three major novel scientific activities are foreseen, being:

• The mathematical modeling of cardiac signals and pathologies
• Design of optimal wavelets
• The design of WT-based algorithms for intelligent sensing and feature extraction
• The development of low-power analog integrated circuits that implement the required wavelet transform and artifact detection, taking into account the limitations imposed by an implantable device, such as a low, non-symmetrical supply voltage and limited energy budget