Session S41.4

A Novel Single-Channel Real-Time Event Monitoring Software for Extremely Hardware-Limited ECG Devices

R Petrovic, A Khawaja*, J Steininger, TK Zywietz

Institute for Biosignal-Processing
Munich, Germany

Cardiac monitoring continuously provides instant assessment of the patients’ heart rhythm. A major drawback of many existing ECG monitoring methods is the high incidence of false alarms. In this paper a reliable single-lead real-time cardiac monitoring algorithm for extremely limited hardware platforms is presented. The algorithm performs robust detection and classification for QRS complexes as well as a trusted recognition method for certain cardiac events including life-threatening arrhythmias. The ECG analyzing software is developed for clinical and homecare applications, validated in real environments and implemented in a portable battery-powered device environments with up to seven-day operating time and low-power consumption (8 MHz MSP430F169 microcontroller from Texas Instruments, 200 Hz A/D converter and very restricted memory resources of 2 KB). Accordingly, the design and the implementation of the corresponding analyzing software were quite challenging. In order to meet these requirements, a particular QRS detector was developed based on the method of spacial-velocity slope detection. Moreover, several special morphology-based similarity measures were carried out and optimized allowing high precision of beat classification and event detection. The whole ECG software implementation requires only 537 Bytes of RAM memory and round 20 KB of ROM flash memory. The QRS detector achieved a sensitivity (Se) of 99.48% and a positive predictivity (P+) of 99.67 % after analyzing the first channel of all MIT-BIH Arrhythmia Database records, while a Se of 99.50% and a P+ of 99.33% were attained after considering the first lead of all records provided by AHA Arrhythmia Database. However, the Se and P+ of PVC detection for the first channel of both databases are greater than 92% and 76% respectively. Furthermore, a so-called post-classification method was developed to enhance the detection specificity of many cardiac events including Bigeminy, Couplet, Triplet and Ventricular Tachycardia. A sensitive and well-trusted monitoring algorithm for cardiac events with a very low rate of false alarms was accomplished, even with limited hardware resources and potential motion artifacts from ambulatory ECG devices.

(Abstract Control Number: 166)