BackgroundSerial electrocardiography aims to contribute to electrocardiogram (ECG) diagnosis by comparing the ECG under consideration with a previously made ECG in the same individual. Here, we present a novel algorithm to construct dedicated deep-learning neural networks (NNs) that are specialized in detecting newly emerging or aggravating existing cardiac pathology in serial ECGs.MethodsWe developed a novel deep-learning method for serial ECG analysis and tested its performance in detection of heart failure in post-infarction patients, and in the detection of ischemia in patients who underwent elective percutaneous coronary intervention. Core of the method is the repeated structuring and learning procedure that, when fed with 13 serial ECG difference features (intra-individual differences in: QRS duration; QT interval; QRS maximum; T-wave maximum; QRS integral; T-wave integral; QRS complexity; T-wave complexity; ventricular gradient; QRS-T spatial angle; heart rate; J-point amplitude; and T-wave symmetry), dynamically creates a NN of at most three hidden layers. An optimization process reduces the possibility of obtaining an inefficient NN due to adverse initialization.ResultsApplication of our method to the two clinical ECG databases yielded 3-layer NN architectures, both showing high testing performances (areas under the receiver operating curves were 84% and 83%, respectively).ConclusionsOur method was successful in two different clinical serial ECG applications. Further studies will investigate if other problem-specific NNs can successfully be constructed, and even if it will be possible to construct a universal NN to detect any pathologic ECG change.Electronic supplementary materialThe online version of this article (10.1186/s12938-019-0630-9) contains supplementary material, which is available to authorized users.

Background:Indirect fetal electrocardiography is preferable to direct fetal electrocardiography because of being noninvasive and is applicable also during the end of pregnancy, besides labor. Still, the former is strongly affected by noise so that even R-peak detection (which is essential for fetal heart-rate evaluations and subsequent processing procedures) is challenging. Some fetal studies have applied the Pan-Tompkins’ algorithm that, however, was originally designed for adult applications. Thus, this work evaluated the Pan-Tompkins’ algorithm suitability for fetal applications, and proposed fetal adjustments and optimizations to improve it. Method: Both Pan-Tompkins’ algorithm and its improved version were applied to the “Abdominal and Direct Fetal Electrocardiogram Database” and to the “Noninvasive Fetal Electrocardiography Database” of Physionet. R-peak detection accuracy was quantified by computation of positive-predictive value, sensitivity and F1 score. Results:When applied to “Abdominal and Direct Fetal Electrocardiogram Database”, the accuracy of the improved fetal Pan-Tompkins’ algorithm was significantly higher than the standard (positive-predictive value: 0.94 vs. 0.79; sensitivity: 0.95 vs. 0.80; F1 score: 0.94 vs. 0.79; P<0.05 in all cases) on indirect fetal electrocardiograms, whereas both methods performed similarly on direct fetal electrocardiograms (positive-predictive value, sensitivity and F1 score all close to 1). Improved fetal Pan-Tompkins’ algorithm was found to be superior to the standard also when applied to “Noninvasive Fetal Electrocardiography Database” (positive-predictive value: 0.68 vs. 0.55, P<0.05; sensitivity: 0.56 vs. 0.46, P=0.23; F1 score: 0.60 vs. 0.47, P=0.11).Conclusion:In indirect fetal electrocardiographic applications, improved fetal Pan-Tompkins’ algorithm is to be preferred over the standard, since it provides higher R-peak detection accuracy for heart-rate evaluations and subsequent processing.

Atrial fibrillation (AF) is a common cardiac disorder that can cause severe complications. AF diagnosis is typically based on the electrocardiogram (ECG) evaluation in hospitals or in clinical facilities. The aim of the present work is to propose a new artificial neural network for reliable AF identification in ECGs acquired through portable devices. A supervised fully connected artificial neural network (RSL_ANN), receiving 19 ECG features (11 morphological, 4 on F waves and 4 on heart-rate variability (HRV)) in input and discriminating between AF and non-AF classes in output, was created using the repeated structuring and learning (RSL) procedure. RSL_ANN was created and tested on 8028 (training: 4493; validation: 1125; testing: 2410) annotated ECGs belonging to the “AF Classification from a Short Single Lead ECG Recording” database and acquired with the portable KARDIA device by AliveCor. RSL_ANN performance was evaluated in terms of area under the curve (AUC) and confidence intervals (CIs) of the received operating characteristic. RSL_ANN performance was very good and very similar in training, validation and testing datasets. AUC was 91.1% (CI: 89.1–93.0%), 90.2% (CI: 86.2–94.3%) and 90.8% (CI: 88.1–93.5%) for the training, validation and testing datasets, respectively. Thus, RSL_ANN is a promising tool for reliable identification of AF in ECGs acquired by portable devices.

Background:Fetal well-being evaluation may be accomplished by monitoring cardiac activity through fetal electrocardiography. Direct fetal electrocardiography (acquired through scalp electrodes) is the gold standard but its invasiveness limits its clinical applicability. Instead, clinical use of indirect fetal electrocardiography (acquired through abdominal electrodes) is limited by its poor signal quality.Objective:Aim of this study was to evaluate the suitability of the Segmented-Beat Modulation Method to denoise indirect fetal electrocardiograms in order to achieve a signal-quality at least comparable to the direct ones.Method:Direct and indirect recordings, simultaneously acquired from 5 pregnant women during labor, were filtered with the Segmented-Beat Modulation Method and correlated in order to assess their morphological correspondence. Signal-to-noise ratio was used to quantify their quality.Results:Amplitude was higher in direct than indirect fetal electrocardiograms (median:104 µV vs. 22 µV; P=7.66·10-4), whereas noise was comparable (median:70 µV vs. 49 µV, P=0.45). Moreover, fetal electrocardiogram amplitude was significantly higher than affecting noise in direct recording (P=3.17·10-2) and significantly in indirect recording (P=1.90·10-3). Consequently, signal-to-noise ratio was initially higher for direct than indirect recordings (median:3.3 dB vs. -2.3 dB; P=3.90·10-3), but became lower after denoising of indirect ones (median:9.6 dB; P=9.84·10-4). Eventually, direct and indirect recordings were highly correlated (median: ρ=0.78; P<10-208), indicating that the two electrocardiograms were morphologically equivalent.Conclusion:Segmented-Beat Modulation Method is particularly useful for denoising of indirect fetal electrocardiogram and may contribute to the spread of this noninvasive technique in the clinical practice.