Use of machine learning and Poincaré density grid in the diagnosis of sinus node dysfunction caused by sinoatrial conduction block in dogs

BACKGROUND

Sinus node dysfunction because of abnormal impulse generation or sinoatrial conduction block causes bradycardia that can be difficult to differentiate from high parasympathetic/low sympathetic modulation (HP/LSM).

HYPOTHESIS

Beat-to-beat relationships of sinus node dysfunction are quantifiably distinguishable by Poincaré plots, machine learning, and 3-dimensional density grid analysis. Moreover, computer modeling establishes sinoatrial conduction block as a mechanism.

ANIMALS

Three groups of dogs were studied with a diagnosis of: (1) balanced autonomic modulation (n = 26), (2) HP/LSM (n = 26), and (3) sinus node dysfunction (n = 21).

METHODS

Heart rate parameters and Poincaré plot data were determined [median (25%-75%)]. Recordings were randomly assigned to training or testing. Supervised machine learning of the training data was evaluated with the testing data. The computer model included impulse rate, exit block probability, and HP/LSM.

RESULTS

Confusion matrices illustrated the effectiveness in diagnosing by both machine learning and Poincaré density grid. Sinus pauses >2 s differentiated (P < .0001) HP/LSM (2340; 583-3947 s) from sinus node dysfunction (8503; 7078-10 050 s), but average heart rate did not. The shortest linear intervals were longer with sinus node dysfunction (315; 278-323 ms) vs HP/LSM (260; 251-292 ms; P = .008), but the longest linear intervals were shorter with sinus node dysfunction (620; 565-698 ms) vs HP/LSM (843; 799-888 ms; P < .0001).

CONCLUSIONS

Number and duration of pauses, not heart rate, differentiated sinus node dysfunction from HP/LSM. Machine learning and Poincaré density grid can accurately identify sinus node dysfunction. Computer modeling supports sinoatrial conduction block as a mechanism of sinus node dysfunction.

Cardiovascular abnormalities in dogs with acute pancreatitis

BACKGROUND

The prevalence and clinical importance of cardiac abnormalities in dogs with acute pancreatitis (AP) is unknown.

ANIMALS

Twelve dogs with AP and 60 archived serum samples from dogs with suspected AP.

METHODS

Two-phase study.

PHASE I

Analysis of archived serum samples from dogs with clinical signs of AP and high Spec cPL concentrations. High sensitivity troponin I (TnIH) and N-terminal pro-B-type natriuretic peptide (NT-proBNP) concentrations were measured in achieved serum samples.

PHASE II

Prospective observational study. Dogs with AP underwent echocardiography and Holter monitoring. Serum cardiac troponin I (cTnI) and plasma NT-proBNP concentrations were measured. Previously described disease severity indices were calculated for each dog.

RESULTS

Phase I: 41 of 60 dogs suspected of having AP had abnormally high TnIH concentrations and 13 of 60 had abnormally high serum NT-proBNP concentrations. Higher TnIH concentrations were observed in dogs with Spec cPL concentration >2000 μg/L as compared to those with concentrations of 1000-2000 μg/L.

PHASE II

11 of 12 dogs diagnosed with pancreatitis had abnormal cTnI concentrations (median: 0.384 ng/mL, range: 0.041-2.966 ng/mL, RI: ≤0.06 ng/mL) and 7 of 12 dogs had plasma NT-proBNP concentrations above the reference interval (median: 971 pmol/L, range: 250-2215 pmol/L, RI: ≤900 pmol/L). Supraventricular and ventricular ectopic beats occurred in 3 dogs. Echocardiographic abnormalities were detected in 5 dogs. Cardiovascular variables were not associated with indices of disease severity.

CONCLUSIONS AND CLINICAL IMPORTANCE

Myocardial injury is common in dogs with AP, but clinical consequences appeared to be uncommon in our small cohort. Cardiac biomarkers should be interpreted with caution in dogs with AP.

ECG Heartbeat Classification Based on an Improved ResNet-18 Model

Based on a convolutional neural network (CNN) approach, this article proposes an improved ResNet-18 model for heartbeat classification of electrocardiogram (ECG) signals through appropriate model training and parameter adjustment. Due to the unique residual structure of the model, the utilized CNN layered structure can be deepened in order to achieve better classification performance. The results of applying the proposed model to the MIT-BIH arrhythmia database demonstrate that the model achieves higher accuracy (96.50%) compared to other state-of-the-art classification models, while specifically for the ventricular ectopic heartbeat class, its sensitivity is 93.83% and the precision is 97.44%.

Optimizing single-chamber pacing in dogs Part 1: Rate determinations, rate interventions and hysteresis

Determining ideal pacing rates to meet physiological needs and optimizing programming to prevent unnecessary right ventricular pacing in dogs requires an understanding of heart rate profiles and applicable pacing technology. The heart rate and rhythm of the dog is complex necessitating investigation of rate requirements of activity and circadian influences. Overlaying this information are a multiplicity of other factors such as age, breed, temperament, cardiovascular disease and underlining rhythm disorders that contribute to the difficulty in making general conclusions. However, all such information permits better implementation of programming options with the goal of better outcomes. In this review (Part 1 of a two-part review) instantaneous heart rate, rolling average heart rate, simple average heart rate, heart rate tachograms, RR interval tachograms (2D, 3D and dynamic), and Poincaré plots (2D, 3D and dynamic) are discussed as they apply to decisions in the determination and examination of pacing rates for dogs programmed in the VVI pacing mode (Ventricular paced, Ventricular sensed, Inhibited pacing). The applicable pacing operations available for three pacemaker companies are reviewed (Abbott, Biotronik/Dextronix, and Medtronic). The programmable options considered include: slowest pacing rate without additional features to extend the pacing interval, sleep/rest rate preferences, hysteresis to lengthen pacing interval following intrinsic beats, and intermittent increases in pacing following abrupt loss of intrinsic rhythm. Recommendations are suggested for follow-up of individual dogs with examination of pacing statistics and Holter monitoring.

Optimizing single-chamber pacing in dogs. Part 2: Rate adaptive pacing

Proper programming of pacemakers for dogs in the rate adaptive mode requires an understanding of the rate requirements for each individual and the interplay of programmable features. The specific advantages and disadvantages of the rate adaptive mode should be considered on a case by case basis. Fundamentally, two components are linked in the implementation of rate adaptive pacing: (1) sensing the need for a change in rate and (2) responding with the appropriate alteration in pacing rate. The programming interaction of these two components are interdependent and affected by the rates programmed. These features may be adjusted manually or automatically. In this review (Part 2 of a two-part review) the considerations required to program each aspect that optimizes the pacing rate profile are reviewed. These include the lower rate, upper sensor rate, activities of daily life rate, sensor threshold, acceleration and deceleration, slope, activities of daily life zone, exertion zone, automatic versus manual adjustments and closed loop stimulation. The programming features of pacemakers manufactured by three companies are summarized (Abbott, formerly St. Jude; Biotronik/Dextronix; Medtronic). Means of assessing the success of pacemaker programing is examined through examples of pacemaker data, Holter analysis, Poincaré plots and tachograms. Finally, the questions and considerations for rate adaptive pacing in dogs that demand investigation are proposed.

Instantaneous and averaged heart rate profiles: Developing strategies for programming pacing rates in dogs

Pacemakers use heart rate histograms (% beats) and sensor indicated rate histograms (% time) to illustrate rate distributions. When programmed to the rate adaptive modes, these data are used to determine the appropriateness of rate response to activity. These histograms are generated from instantaneous heart rate calculations. In humans, such data are compared to known histographic rate profiles. Such rate profiles during 24 h in the dog are not available. Moreover, data representation differ between Holter monitoring and pacemakers making comparisons challenging. The rate distribution in dogs >7-years of age was determined over 24 h using instantaneous and rolling average heart rate. Such data could serve as a guide to programming pacing rates for dogs. Sinus arrhythmia resulted in dissimilar heart rate profiles depending on the method of determining rate. The long intervals of sinus arrhythmia resulted in median values for the percent of time with an instantaneous heart rate of <50 beats/min (bpm) of 15%, whereas a rolling average heart rate of <50 bpm was 0.2%. Based on the cumulative time of the rolling average rate, dogs spent 26.3% of the day between 70-90 bpm with rates <65 bpm and >90 bpm approximating 30% for each. Rates >160 bpm were uncommon (<1%). However, high variability existed between dogs. This study demonstrated the shortcomings of both instantaneous and averaging methods to evaluate heart rate profiles in the dog and that both methods should be incorporated when making pacing rate decisions during programming.

Comparison of heart rate obtained from shorter duration Holter recordings to 24-hour mean heart rate in dogs with atrial fibrillation

The objective of this study was to evaluate the accuracy of short duration electrocardiographic (ECG) recordings extracted from ambulatory continuous ECG (Holter) to assess 24-hour mean heart rate in dogs with atrial fibrillation. In this retrospective study, Holter recordings obtained from 20 dogs with atrial fibrillation were selected for analysis. Ten out of 20 dogs were receiving drugs to control heart rate at the time of Holter evaluation. From the Holter recordings, heart rate averages were calculated for various sample durations (five-minutes, 30 minutes, one-hour, two-hours, and three-hours) for each dog. Percentage of these shorter duration ECG obtained HR averages that fell within ±10%, ±15% and ± 20% of 24-hour mean heart rate was determined for each sample duration and for each dog. Seventy five percent of heart rate averages obtained from three-hour ECG recordings fell within ±10% of 24-hour mean HR. All the heart rate averages obtained from two-hour ECG recordings fell within ±20% of 24-hour mean heart rate. Based on the results of this study it can be concluded that the duration of the ECG recording affects the prediction accuracy for 24-hour Holter mean HR. Only two and three hours of Holter recordings provided all heart rate averages within ±20% of 24-hour mean heart rate. No significant differences were noted in the prediction accuracy of shorter duration ECG recordings based on rate control therapy status. Further prospective studies are needed to assess the accuracy of HR obtained at home using various ECG recording devices to predict 24-hour mean heart rate in dogs with atrial fibrillation.

Comparison of single- versus seven-day Holter analysis for the identification of dilated cardiomyopathy predictive criteria in apparently healthy Doberman Pinscher dogs

INTRODUCTION

The primary objective of this study was to test whether seven-day Holter recording improves the sensitivity of detecting dilated cardiomyopathy (DCM) predictive criteria (DCM_p) compared with 24-h Holter recording in asymptomatic Doberman Pinscher (DP) dogs.

ANIMALS

Twenty-eight asymptomatic DP dogs with normal echocardiographic examinations.

METHODS

Dogs with normal echocardiographic examinations underwent seven-day Holter monitoring. The presence of ≥50 ventricular premature complexes and or ≥ one couplet/one triplet/one episode of ventricular tachycardia per 24-h period was considered positive for DCM_p.

RESULTS

Five dogs were positive on the first day, and an additional six dogs tested positive from day two to seven of the Holter recording. The number of dogs positive for DCM_p detected by four days was significantly different (p = 0.031) compared with the first-day Holter recording.

CONCLUSIONS

Seven-day Holter recording detected significantly more dogs with DCM_p compared with the first-day Holter recording. Follow-up studies are warranted to evaluate the long-term accuracy of multiple-day Holter analysis in predicting the development of DCM in DP dogs.