Combining 12-Lead ECG Information for a Beat Detection Algorithm

Abstract

This paper proposes a beat detection algorithm specially tailored to be used with 12 channel ECG records. The algorithm first obtains beat positions on each channel, and then combines this information to get an improved estimate. The detection process involves two stages: 1) single-channel detection: implemented by improving one of the most popular methods (Pan-Tompkins) developed to detect beat positions; and 2) multichannel detection: an algorithm that combines the information of the beat positions obtained in each of the 12 channels. In this way, our results clearly improve those obtained with the single-channel detection method, discarding detection errors, false positives, and duplicated beats. Besides, our single-channel method significantly reduces the temporal error when estimating the positions of QRS complexes. In the multichannel detection, the assessment of our algorithm against one-channel based approaches shows a significant improvement in detection outcome (Se = 99.86%, P+ = 99.98%, RMS RR Interval Error = 25.98 ms, F-Score = 0.9992), making it a good starting point for automatic diagnosis of heart conditions.

A Multi-Lead Fusion Method for the Accurate Delineation of QRS Complex Location in 12 Lead ECG Signal

“…In detected by [16], [27] are detected by the proposed algorithm. The comparative results of average Sensitivity and Positive Predicitvity with the reported results in literature [24], [19], [27], [37], [28], [38], [29], [39] are shown in Table I. Hence, the proposed method shows satisfactory performance on both the databases.…”

“…In [23] In a work done by Ledezma et al [19], six different single lead QRS detector and optimum data fusion algorithm was used and evaluated on INCART database with good sensitivity but a supervised training period of few minutes was required for the calculation of weight coefficients associated with each lead. The work reported by Mondelo et al [28] detected the QRS complex location by arranging all available QRS complex location in ascending order and then by defining two limits. The method first detects the actual QRS locations and later the false positives, which make the algorithm not suitable for real medical applications.…”

A unique cardiac electrocardiographic 3D model. Toward interpretable AI diagnosis

“…Single-lead detections provided in Step 2 are combined to obtain a common reference annotation set. Following the ideas given in Mondelo et al., 29 the single detections are increasingly sorted to build groups of at least four nearby detections, in a way that the groups are far enough from each others. The median of the nearby detections within groups, called , are computed.…”

“…One may refer to Mondelo et al. 29 for details. From now on, only the leads on the set will be considered.…”

Multiple Physiological Signals Fusion Techniques for Improving Heartbeat Detection: A Review

“…ECG recordings usually contain noise, baseline wander, and artifacts caused by patient movement or even sensor disconnections, which may result in intervals with total signal loss. Multi-lead ECG approaches for heartbeat detection have also been presented [15,16,17,18,19,20,21], though these also tend to fail with ECG data corrupted with high levels of noise; ECG noise (especially caused by patient movements and electrical interference) tends to be present in all the available leads.…”