FPGA-based Platform For Real-Time Processing of Electrophysiological Signals: The QRS Segmentation Issue
Résumé
This demo presents an FPGA-based platform dedicated to the parallel processing of four electrophysiological (EP) signals - EEG – electroencephalography, EMG – electromyography, ECG – electrocardiography and EOG – electrooculography - for clinical diagnosis. The demonstrator comprises a custom daughter card for simultaneous signal acquisition and a Cyclone IV FPGA-based platform for signal processing. A specific HMI has been developed for the visualization of the signals and the extracted features. In this demo, we will focus on the real-time extraction of specific ECG-parameters acquired from volunteers. To achieve this, we will present a segmentation algorithm of cardiac intervals based on the automatic retrieval of QRS complexes from ECG signals. The proposed algorithm was implemented on FPGA using System Level Design methodology. The algorithm and stimuli were first implemented in MatlabTM and then the various processing blocks were either synthesized with the MatlabTM HDL coder tool or hard-coded in VHDL. The FPGA-based algorithm correctly detects 91.85% of the QRS complexes, with a very limited false positive rate (only 5%). These results are commensurate with the most recent state-of-the-art off-line algorithms. On real data, the demonstrator correctly detects 97.37% of the QRS complexes with a sensitivity of 93.77%