Applications of artificial intelligence
in wearable cardiac monitoring

Recent technological advancements have enabled machine learning to recognize voice commands like Amazon Alexa, and recognize your face like Apple Face ID. Today, these technical advances are utilized in healthcare to help identify cancer in CT scans, detect diabetic retinopathy and identify irregular rhythms in ECG. Cardiac defibrillators, insulin pumps and monitors all work wirelessly and can feed data directly into the electronic health record (EHR).

The BodyGuardian® Remote Monitoring System is one such example. Patients wear the cardiac monitor, which feeds real-time data into a cloud-based health platform physicians can access. The need for these technologies is growing as we see a greater incidence of cardiac disease and the population is aging. This will bring a greater reliance on algorithms to provide high-quality reporting in a timely manner. These factors are amplified in the case of mobile cardiac telemetry (MCT), where ECG is streamed directly to data processing centers, annotated, and may be used to quickly alert clinicians of potentially life-threatening cardiac events. For MCT to be most effective, data annotation must be highly accurate and quick.

Caution: U.S. Federal law restricts this device to sale by or on the order of a physician.



Traditional machine learning is a type of artificial intelligence that enables machines to learn statistical relationships between inputs and outputs. Uncovering these relationships through learning produces algorithms that make decisions in a way that can mirror humans. Decisions are driven by experience and capture complexity and nuance that non-learning algorithms cannot reproduce. However, machine learning relies on human insight and engineering to extract the important information from raw data, placing a fundamental limit on the ability of these algorithms to achieve human-level performance.

Deep learning is a type of artificial intelligence that is like machine learning but with a powerful new trick. Learning is used to identify the important information within data and to make decisions based on that information. This means that the performance of deep learning algorithms is not limited to what we know is important about the data. The learning processes can uncover new features that may be difficult to describe but are fundamental to human decision making. Applying learning more generally is what enables deep learning algorithms to achieve human-level performance.



Improved patient experience, improved health of populations and reduced costs are the Holy Grail of healthcare today. Using AI for clinical decision support empowers clinicians with more accurate information, faster. Enabling clinicians to deliver better care to more patients is the promise of AI. Because deep learning leverages computation but learns like humans, it can be trained through trial-and-error on thousands or millions of ECG recordings. This provides the algorithm with a lifetime of training in only a few hours. Like with human learning, more training makes for better decision making.



The algorithms that make up the Preventice cloud-based analysis platform, BeatLogicTM, detect cardiac arrhythmias and route ECG to skilled technicians to maximize speed of processing while minimizing false positives and false negatives. With the highest value monitoring services like telemetry, every beat of ECG is processed by multipole advanced algorithms to ensure accuracy and the highest quality reporting. BeatLogicTM achieves state-of-the-art performance by leveraging big data in combination multiple proprietary deep learning architectures. Algorithms are trained using a diverse expertly curated dataset consisting of ECGs from more than 9,000 patients to classify beats and detect 34 types of arrhythmia.


Every remote monitoring system is different.

  • How is the arrhythmia detected and what is the level of accuracy?
  • Has the algorithm been validated by research and peer-reviewed studies?

It is important to understand that developing a deep learning algorithm in some ways is easy. However, developing a truly robust and accurate algorithm requires a great deal of expertise and a large amount of high-quality, well-curated training and validation data. For example, deep learning models rely on simple computational units with trainable parameters that are stacked and connected in a way that allows for modeling of complex relationships. The result is a network with millions of trainable parameters, and with so many trainable parameters deep learning algorithms are extremely sensitive to overfitting. They must be carefully designed, trained and validated in order to ensure that the algorithm performs well when presented with never-before-seen data.

case study

At Preventice, our deep learning algorithms leverage our combined expertise in algorithm development and ECG interpretation to provide the highest quality reporting on factors that influence clinician decision making. For example, the detection of ectopic heart beats that originate from an abnormal focus within the ventricles of the heart (Ventricular Ectopic Beats or VEBs). In the general population, VEBs may be of little clinical consequence, but for some patients these beats signify a potentially life-threatening and treatable cardiac irregularity.

At Preventice, state-of-the-art VEB algorithm performance was achieved by leveraging big-data in combination with a novel deep learning architecture. Training was performed using more than 650,000 beats from nearly 3,500 unique patients with a diverse array of beat morphologies and noise conditions (Figure 2, Figure 3).

case study

The deep learning network incorporated a small two-layer network, which was fed heart rates, and a deep convolutional network, which was fed 3-beat trains (Figure 3), enabling it to classify beats based on heart rate, waveform morphology and context from surrounding beats.

Within the research literature, VEB algorithm validation is typically performed using an 11-patient sub-set of the publicly available MIT-BIH database. Currently, the best performing algorithms within the literature require patient-specific training to achieve > 90% sensitivity.

case study

While patient specific training may be useful in a research environment, it is not practical for real-world implementation where speed is required for critical notifications and thousands of patients may be monitored in a single day.

Table 1 shows how the Preventice VEB algorithm outperforms all previously published work, achieving 99.2% sensitivity while requiring no patient specific training.



Many of the previously published deep learning beat classification algorithms have relied on the publicly available MIT-BIH database, which consists of 30-minute ECG records from 47 patients for both training and testing. The small number of patients within this database prevents researchers from demonstrating that their algorithms perform well for new patients, particularly in cases where patients are not separated for training and validation. It also prevents researchers from identifying algorithm “blind spots”, situations where the algorithm performs poorly during particular arrhythmias that were never seen during training. Blind-spots are likely to occur during critical arrhythmias as they may not be represented in small datasets.

Validation of the Preventice VEB algorithm was performed on ECG captured using the BodyGuardian® Heart device. The algorithm achieved state-of-the-art performance with 95% sensitivity and 98% specificity on the real-world validation dataset. Real-world validation using a large diverse dataset is a necessity for evaluating deep learning algorithms and standard practice for all algorithms deployed under the BeatLogic™ platform.

  1. Michael Mcroberts and Benjamin A. Teplitzky, Fully-automated ventricular ectopic beat classification for use with mobile cardiac telemetry, 2018 IEEE 15th International Conference on Wearable and Implantable Body Sensor Networks, BSN 2018, vol. 2018, pp. 58–61, 2018
  2. Benjamin A. Teplitzky, Michael McRoberts and Suneet Mittal, Real-World Validation of a Deep Learning Algorithm for Fully-Automated Premature Ventricular Beat Classification During Ambulatory External ECG Monitoring, Heart Rhythm Society's 39th Annual Scientific Sessions in Boston, May 9-12, 2018, Abstract-PO03-044