Detection of subjects with ischemic heart disease by using machine learning technique based on heart rate total variability parameters
Related Experiment Video
Updated: Dec 5, 2025
Calculating Heart Rate Variability from ECG Data from Youth with Cerebral Palsy During Active Video Game Sessions
Published on: June 5, 2019
Agostino Accardo1, Giulia Silveri1, Marco Merlo2
1Department of Engineering and Architecture, University of Trieste, Trieste, Italy.
Insights
This study identifies key heart rate total variability (HRTV) parameters and left ventricular ejection fraction (LVEF) for non-invasively detecting ischemic heart disease (IHD) using artificial neural networks (ANNs). The model achieved 79.8% accuracy with LVEF, offering a promising diagnostic tool.
Area of Science:
- Cardiology and Cardiovascular Research
- Biomedical Engineering and Signal Processing
- Machine Learning in Healthcare
Background:
- Ischemic heart disease (IHD) often presents subtly, progressing to severe events like myocardial infarction without clear early symptoms.
- Current IHD diagnosis relies on invasive coronary angiography, highlighting the need for non-invasive detection methods.
- Previous research explored heart rate variability (HRV) and artificial neural networks (ANNs) for cardiovascular conditions, but with limited scope for IHD classification.
Purpose of the Study:
- To identify optimal ANN structures and heart rate total variability (HRTV) parameters for non-invasive IHD patient identification.
- To validate the performance of these parameters and models on a large, matched cohort of IHD and healthy subjects.
- To assess the impact of the non-invasive clinical parameter, left ventricular ejection fraction (LVEF), on classification accuracy.
Main Methods:
- Extracted linear and non-linear HRTV parameters from 24-hour RR signals of 251 healthy and 245 IHD subjects.
- Trained and tested various ANNs incorporating HRTV parameters, age, and gender, varying hidden neuron count and dataset splits.
- Evaluated ANN performance with and without the inclusion of left ventricular ejection fraction (LVEF) as an input feature.
Main Results:
- HRTV parameters exhibited significantly higher variability in IHD patients compared to healthy individuals.
- An ANN model using specific HRTV parameters (mean RR, LF, LF/HF, beta exponent, SD2), age, and gender achieved 71.8% accuracy.
- Incorporating LVEF into the ANN model significantly improved classification accuracy to 79.8%.
Conclusions:
- A combination of specific HRTV parameters, demographic data, and LVEF can effectively identify patients with IHD non-invasively.
- This approach offers a potential advancement in early and reliable detection of ischemic heart disease.
- The findings underscore the utility of machine learning and advanced signal processing in cardiovascular diagnostics.
Abstract:
Objective: Ischemic heart disease (IHD), in its chronic stable form, is a subtle pathology due to its silent behavior before developing in unstable angina, myocardial infarction or sudden cardiac death. The clinical assessment is based on typical symptoms and finally confirmed, invasively, by coronary angiography. Recently, heart rate variability (HRV) analysis as well as some machine learning algorithms like artificial neural networks (ANNs) were used to identify cardiovascular arrhythmias and, only in few cases, to classify IHD segments in a limited number of subjects. The goal of this study was the identification of the ANN structure and the HRV parameters producing the best performance to identify IHD patients in a non-invasive way, validating the results on a large sample of subjects. Moreover, we examined the influence of a clinical non-invasive parameter, the left ventricular ejection fraction (LVEF), on the classification performance.Approach: To this aim, we extracted several linear and non-linear parameters from 24 h RR signal, considering both normal and ectopic beats (heart rate total variability), of 251 normal and 245 IHD subjects, matched by age and gender. ANNs using several different combinations of these parameters together with age and gender were tested. For each ANN, we varied the number of hidden neurons from 2 to 7 and simulated 100 times, changing randomly the training and test dataset.Main results: The HRTV parameters showed significant greater variability in IHD than in normal subjects. The ANN applied to mean RR, LF, LF/HF, beta exponent, SD2 together with age and gender reached a maximum accuracy of 71.8% and, by adding as input LVEF, an accuracy of 79.8%.Significance: The study provides a deep insight into how a combination of some HRTV parameters and LVEF could be exploited to reliably detect the presence of subjects affected by IHD.
Related Concept Videos
01:30Pulse rhythm
Conversely, an irregular pulse pattern is termed dysrhythmia, stemming from disruptions in cardiac...
01:17Ischemic Heart Disease: Overview
Atherosclerosis, the primary malefactor, orchestrates this dangerous condition. It manifests as the accumulation of fatty deposits, akin to insidious plaques, within arterial walls. As time elapses, these plaques metamorphose, hardening and...
01:30Factors Influencing Heart Rate
Let us explore the significant factors affecting heart rate, including age, body temperature, posture, acute pain, chemical influences,...