Doctors increasingly use software for automatic assessment of ECG Holter monitoring signals in adult patients, but no software has been developed for children so far. Cardiomatics and the Medical University of Warsaw are on the way to a breakthrough in paediatric cardiology. They are developing an internationally innovative tool for automatic assessment, analysis and interpretation of electrocardiographic (ECG) signals from paediatric patients.
ECG diagnostics in children today
Holter monitoring is a standard test performed in the clinical practice of paediatric cardiology. Currently, we see an increasing number of children with indications for Holter monitoring. The most common indications in this group of patients include heart defects before and after cardiac surgery, palpitations, syncope, suspected cardiomyopathy, chest pain, evaluation of antiarrhythmic treatment and complete atrioventricular block. To perform a Holter test in children, doctors use the same ECG signal recorders and the same standardised analysis software that is used for adult patients. When standard algorithms are used, manual analysis of children’s long-term ECG recordings can be very time-consuming and difficult.
Compared to adults, there are a number of challenges associated with automated analysis of ECG signal in paediatric patients. This is due to distinct, age-dependent cardiovascular characteristics, which means that the signal morphology and amplitude are strongly dependent on patient age, sinus arrhythmia, or high heart rate. With these factors present, it becomes impossible to perform high-quality automated analysis based on standard software used in adults.
“If we take tools suited for adult ECG analysis and use it in a child, we may end up with an incorrect interpretation of automatic recordings. If automatic signal analysis is incorrect, the entire analysis must be performed manually by humans, which is a tedious and time-consuming process. For this reason, the risk of error, and in particular the risk of overlooking important electrocardiographic occurrences, is greater than in the case of automatic analysis, and may lead to failure to diagnose significant arrhythmias and cardiac conduction abnormalities, which means that the required pharmacological action or pharmacological intervention might not be undertaken in due time,” emphasises Professor Bożena Werner, M.D., Head of the Paediatric Cardiology and General Paediatrics Clinic at the Children’s Clinical Hospital operated by the Medical University of Warsaw (MUW).
There are occasional reports on algorithms which interpret the ECG signal from paediatric patients, but only in the case of resting 12-lead ECG. To date, however, no AI-based solutions have been developed to analyse long-term ECG recordings for children. Until today, paediatric cardiology has remained dormant in the context of the expanding AI systems, and yet it is among the medical specialisations that can benefit more than others from new technological solutions.
Cardiomatics algorithms harnessed to help children with heart diseases
As an R&D company, Cardiomatics is one of the leaders in the use of artificial intelligence in medicine. For more than three years now, Cardiomatics has been facilitating the daily work of physicians by providing a cloud-based tool for automatic ECG signal analysis. The tool ensures secure data processing and is CE-certified as a Class IIa medical device. Every day, Cardiomatics minimises the time needed for cardiologists to evaluate data thanks to innovative AI-based algorithms. The tool is currently used in more than 600 cardiology centres across Europe, helping to analyse long-term ECG recordings in adult patients. As a natural next step for the system, a decision was made to develop a solution to support ECG diagnostics for children.
“The needs of cardiologists have been neglected for years. As we talk to physicians every day, we know all too well that they often have to perform difficult diagnosis of children’s heart diseases while relying on inadequate tools. The problems arising from this situation are related not only to the very poor workflow, but also an enormous risk of errors that doctors run when analysing ECG recordings for children. Based on our experience, which expands day after day thanks to AI and machine learning, we are ready to take up the challenge of changing the future of paediatric cardiology,” emphasises Rafał Samborski, the CEO of Cardiomatics.
A forthcoming breakthrough in paediatric cardiology
Cardiomatics, the developer of the well-known digital ECG analysis technology, joined forces with a team from the Medical University of Warsaw (MUW), recently commencing a project entitled: Developing an innovative tool for automatic assessment of arrhythmia and cardiac conduction abnormalities in paediatric patients – Cardiomatics Junior.
This innovative project will enable more effective assessment of cardiac arrhythmia in children and reduce the time needed by cardiologists to evaluate Holter monitor data by using AI-based Cardiomatics algorithms.
A key challenge will be to build algorithms that enable high-quality analysis of ECG signals in children. The algorithms will be based on deep neural network architectures, such as ResNet, and will rely on filtered ECG signals. At the same time, the algorithm will be primarily trained using a database of signals from paediatric patients, which will be set up in collaboration with MUW researchers.
On the part of Cardiomatics, project implementation will be overseen by Mariusz Mąsior, CTO, and Katarzyna Barczewska, PhD, Head of the Cardiomatics R&D team. On behalf of the MUW, the project will be led by a team of cardiologists headed by Prof. Bożena Werner and Radosław Pietrzak, PhD.
The research team faces numerous technological challenges that must be addressed to ensure excellent performance of Cardiomatics Junior. These include:
- creating a representative database for each age category as a basis to train the algorithms, establishing benchmarks from centile grids,
- eliminating the risk of false positives when detecting atrial fibrillation in children and identifying real arrhythmias,
- using and interpreting information on heart rate activity and morphology,
- considering the variability of the sinus rhythm in children and providing its correct interpretation,
- considering the mode of interpretation for sports medicine and performance assessment,
- considering the singularities of the ECG signal and heart rhythm in children, e.g. irregular or very high rhythm misinterpreted as arrhythmia (P wave hidden in T wave), different volatility (very low in young children, high in older ones), very high T waves in infants, very low QRS complexes,
- eliminating motion artifacts, which are much more common in children due to their higher motor activity compared to adults.
“The development of a reliable Cardiomatics system for automated analysis of Holter ECG recordings in children will not only improve the work of clinicians, but will also enhance accessibility and versatility of this test, which plays a very important role in detecting the heart rhythm and conduction abnormalities in paediatrics. Cardiomatics technology will also improve the diagnosis of these diseases, paving the way for appropriate therapy at an earlier stage. In some situations, it may even help to avoid life-threatening arrhythmia events such as ventricular tachycardia or sudden cardiac arrest in conditions such as channelopathies,” says Radosław Pietrzak, PhD.
In everyday practice, Cardiomatics Junior will potentially minimise the analysis time to 5 minutes, enabling full report generation and delivery within up to 15 minutes. This will enable doctors to change their current patient handling routines: the doctor will be able to inform the patient about the results already on the second visit, when removing the Holter monitor, without having to wait until the next appointment. This solution will also expand the testing opportunities. Automated analysis will improve the availability of Holter diagnostics and its potential implementation in a greater number of medical facilities. In the era of the COVID-19 pandemic, it is recommended to limit the number of face-to-face appointments, especially for children after cardiac surgery, to reduce exposure to pathogens (including viruses and bacteria).
The new product will be an effective tool for rapid diagnosis of cardiac disorders, benefiting many users. Cardiomatics Junior will be primarily targeted at:
- physicians, including paediatric cardiologists, sports physicians, and paediatricians interested in performing arrhythmia screening,
- Holter monitor manufacturers who would like to enable paediatric Holter analysis with their own device,
- clinical trial implementers.
However, the ultimate beneficiaries of successful project completion will be young patients and their parents.
The R&D work will continue for another three years, consuming a budget of PLN 5,363,925. Project implementation is being supported by the National Centre for Research and Development.