Multi-lead Wavelet-based ECG Delineation on a Wearable Embedded Sensor Platform F Rinc ´ on 2,1 , N Boichat 1 , V Barbero 2 , N Khaled 3 , D Atienza 1 1 Embedded Systems Laboratory (ESL)- Ecole Polytechnique F´ ed´ erale de Lausanne, Switzerland 2 DACYA-Complutense Univ. of Madrid, Spain 3 Dept. of Signal Theory and Communications (TSC)-Univ. Carlos III of Madrid, Spain Abstract This work is dedicated to the sensible optimization and porting of a multi-lead (ML) wavelet-transform (WT)- based electrocardiogram (ECG) wave delineator to a state-of-the-art commercial wearable embedded sensor platform with limited processing and storage resources. The original offline algorithm was recently proposed and validated in the literature, as an extension of an earlier well-established single-lead (SL) WT-based ECG delin- eator. Several ML ECG delineation approaches, including SL selection according to various criteria and lead combi- nation into a single root-mean-squared (RMS) curve, are carefully optimized for real-time operation on a state-of- the-art commercial wearable embedded sensor platform. Furthermore, these ML ECG delineation approaches are contrasted in terms of their delineation accuracy, complex- ity and memory usage, as well as suitability for ambulatory real-time operation. Finally, the robustness and stability of the ML ECG delineation approaches are benchmarked with respect to a validated SL implementation. 1. Introduction A significant amount of research effort has been devoted to the automated analysis of electrocardiogram (ECG) sig- nals, and in particular to the underlying detection of the major ECG characteristic waves, namely the QRS com- plex, P and T waves, so-called ECG delineation [1]. As a result, several automatic delineation methods working on a single ECG lead can be found in the literature [1, 2]. In practice, however, multiple leads are simultaneously ac- quired both in traditional clinical settings (the standard 12 leads) and in emerging ambulatory ECG monitoring (the 3-lead configuration). The deployment of delineation approaches able to exploit the multiple leads, i.e., multi- lead (ML) delineation, can potentially improve the accu- racy, stability and resilience to artifacts of the character- istic waves measurements, compared to single-lead (SL) This work was supported by the Spanish Government Research Grant TIN2008-00508. delineation [3, 4]. This is particularly relevant for the herein considered ambulatory/wearable ECG application, the quality of the measurements by the different lead may not be known a-priori (ECG electrodes installed by the pa- tient himself) and/or may vary due various artifacts intro- duced during daily usage. The present work investigates two approaches to deploy- ing ML delineation based a single-lead ECG delineator, namely, lead selection and multi-lead combination into a single root-mean-squared (RMS) curve. The considered single-lead delineator is based on a state-of-the-art wavelet transform-based delineator [2, 5], which exploits the time- scale description of this transform to provide a robust, effi- cient and reliable automated analysis of the multiresolution waves of the ECG signal. More specifically, we use a mod- ified version of this early offline algorithm [2], which we have previously optimized for real-time implementation on a state-of-the-art commercial embedded wearable sensor platforms (EWSNs) with limited processing and storage resources Shimmer TM [7]. The detailed description and validation of our optimized algorithm can be found in [6]. In addition to comparatively evaluating the delineation accuracy of the aforementioned multi-lead ECG delin- eation approaches, this work describes their optimization and porting to the Shimmer TM platform. In particular, we report on the complexity and memory usage of these multi- lead approaches, and thus assess their suitability for ambu- latory real-time operation. The rest of the paper is organized as follows. In Sec- tion 2, we introduce the investigated ML delineation meth- ods. In Section 3 we describe the applied optimization techniques to port the ML delineation methods onto an EWSN. Then, in Section 4 we present the experimental results that validate the quality of the delineation. Finally, in Section 5 we draw the main conclusions of this work. 2. Multi-lead delineation methods Two main approaches to ML delineation methods are considered in this work. The first approach is lead selec- tion, which simply selects among the available multiple ISSN 0276-6574 289 Computers in Cardiology 2009;36:289-292.