Comparison of alarm strategies for continuous 12-lead ST-segment monitoring

Several alarm detection strategies for real-time ST-segment monitoring (including detection using any single lead, any two leads, any two contiguous leads, any three leads, and STindex, computed as the sum of the absolute ST values from three quasi-orthogonal leads) have been evaluated using a 12-lead ST database generated from the Dalhousie PTCA database. The database includes 95 patients with a total of 202 balloon occlusions. For all the occluded vessels combined, the detection sensitivity using any single lead for ST-alarm detection decreases from 84.7% to 74.8% when the threshold is increased from 1 mm to 1.5 mm. For alarm detection using any two leads, two contiguous leads, and three leads, the results using a 1-mm threshold are 79.7%, 78.7%, and 77.2%, respectively. The two-lead performance results can be improved to 87.1% for any two leads and to 84.2% for any two contiguous leads if a lower threshold of 0.8 mm is used. For all the possible three-lead combinations, the STindex computed using leads III, V2, and V5 shows the highest correlation to the sum of the absolute ST values from all 12 leads. A detection sensitivity of 85.2% can be obtained using STindex with a threshold of 1.5 mm.