We developed a convolutional neural network approach that can identify subjects with congestive heart failure with extremely high accuracy through analysis of just one raw electrocardiogram (ECG) heartbeat.
Congestive heart failure (CHF) is a chronic progressive condition that affects the pumping power of the heart muscles. Associated with high prevalence, significant mortality rates, and sustained healthcare costs, clinical practitioners and health systems urgently require efficient detection processes.
Sebastiano Massaro — Director of the Organizational Neuroscience Laboratory and Associate Professor at the University of Surrey (UK) — worked with colleagues Ernesto Iadanza (University of Florence, Italy), Mihaela Porumb, and Leandro Pecchia (Warwick University, UK) to tackle these important concerns.
Recently published in Biomedical Signal Processing and Control, this research uses Convolutional Neural Networks (CNN) — hierarchical neural networks highly effective in recognizing patterns and structures in data — to drastically improves existing CHF detection methods. These methods are typically focused on heart rate variability (HRV) that, whilst effective, are time-consuming and prone to errors. Conversely, this new model uses a combination of advanced signal processing and machine learning tools on raw ECG signals, delivering an almost perfect detection accuracy at the subject level.
Specifically, a CNN model was trained and tested on large publicly available ECG datasets featuring subjects with CHF as well as healthy, non-arrhythmic hearts. By checking just one heartbeat it was ultimately possible to detect whether or not a person within this sample had severe heart failure (CHF Class 3 and 4). The model is also one of the first known to be able to identify the ECG' s morphological features specifically associated to the severity of the condition.
With approximately 26 million people worldwide affected by a form of heart failure, our research presents a major advancement on the current methodology. The prospect of enabling clinical practitioners to access an accurate CHF detection tool promises to make a significant societal impact, with patients benefitting from early and more efficient diagnosis and easing pressures on healthcare systems' resources.
Further Readings
Porumb, M., Massaro, S., Iadanza, E., & Pecchia, L. (2020). A Convolutional Neural Network Approach to Detect Congestive Heart Failure. Biomedical Signal Processing and Control, 5. (2020)
Featured In
