Thursday, April 23, 15:30 - 18:00, Location: Show and Tell Area B
O. Gergaud, F. Porée, A. Kachenoura, G. Carrault and A.I. Hernandez
Patients treated with Cardiac Implantable Devices (CIDs), such as Bi-ventricular Pacemakers (BivP) and Defibrillators (BivD) used in the cardiac resynchronization therapy require regular hospital visits to perform patient’s follow-up, to monitor whether the CID is working optimally and, eventually, to modify the pacing parameters. Current developments have for objective to propose accurate methods for remote follow-up of these patients so as to reduce the health care costs. Since the physician considers the surface ElectroCardioGram (ECG) as the reference signal for the analysis of the cardiac activity and since the CIDs only provide intracardiac ElectroGrams (EGM), it is important to propose an innovative platform able to synthesize a standard set of 12- lead ECG from EGM data in order to provide a less expensive and less time-consuming setup for monitoring the patient’s cardiac electrical activity. This demonstration is related to the paper entitled “Comparison of four estimators of the 3D cardiac electrical activity for surface ECG synthesis from intracardiac recordings” accepted by Icassp 09 in Bio Imaging and signal Processing session.
Purpose of the platform
Several methods dealing with the synthesis of the 12-lead ECG based upon the EGM have been proposed in literature [Kachenoura et al., Eurasip, 2008] or have been patented [EP1902750B1] by our laboratory, including linear and non-linear filtering. The proposed platform is a patient-specific approach and includes two main modules: the learning module and the reconstruction module.
The learning module: The user selects the ECG leads he would like to reconstruct and the EGM used for the reconstruction. A variable number of heartbeats is also chosen by the user. He has the possibility to select the method and its associated parameters among several implemented filters. Briefly, the ECG synthesis is based on the transfer function estimated between the three dimensional (3D) representation of the intracardiac (VGM) and the surface (VCG) electrical activity. More precisely, the platform computes:
1) the VGM and the VGM,
2) the transfer function (TF) between VGM and VCG. This TF is based either on a linear filter (estimated by LMS or RLS) or on a non-linear filter (Feed-Forward neural network or Dynamic Time Delay artificial neural networks).
The reconstruction module: the user applies the estimated filter on an other selected time interval (different from the learning one). Here, the platform computes the reconstructed ECG using only the EGM. The adequacy of the reconstruction is verified by superimposing estimated ECG and real ECG.
Conclusion
Evaluation of the platform has been performed on a specific database collected during the implantation of the device. Fifteen patients have been stored and can be used for the demonstration. The proposed platform offers the possibility to choose the patient, the reconstruction method and its parameters, the number of EGM leads, the number of ECG leads, the number of heartbeats for the learning step and the number of heartbeats for the reconstruction step. All these freedom degrees create an interactive demonstration. This innovative platform is now ready to perform the follow-up of the patient in clinical application. In this last situation, the learning module will be used in the operating room during the implantation. The platform will then be associated with an ECG and EGM acquisition system and the TF between the ECG and the EGM will be learnt. The reconstruction module will be used i) at home to improve by telemedicine the follow-up of the patient or ii) at the hospital during the cardiac device control. In this case, only EGM will be required. They will be collected in practice via a telemetric head reading the memory of the device and transmitting the stored EGM to the platform in charge of the ECG reconstruction.