How can gender be identified from heart rate data? Evaluation using ALLSTAR heart rate variability big data analysis

Area of Science:

• Cardiology
• Biomedical Engineering
• Data Science

Background:

• Electrocardiography (ECG) and Holter monitoring provide long-term heart rate data.
• Previous research has explored various physiological markers for demographic identification.

Purpose of the Study:

• To investigate the feasibility of gender verification using electrocardiogram (ECG) data.
• To assess the accuracy of different machine learning algorithms in gender identification from heart rate variability.

Main Methods:

• Extraction of ten-dimensional statistics from heart rate data of over 420,000 individuals.
• Comparison of multiple machine learning models including Lasso, linear regression, Support Vector Machines (SVM), random forest, logistic regression, k-means, and Elastic Net.
• Analysis stratified by age groups (under 50 and 50 and over).

Main Results:

• The highest accuracy achieved was 0.681927 using the Random Forest algorithm for individuals under 50.
• No consistent significant differences in accuracy were observed between the two age groups.
• While statistically significant, the discrimination results were not accurate enough for reliable individual gender determination.

Conclusions:

• Gender verification from long-term ECG data using current machine learning techniques is not sufficiently accurate.
• Further research is needed to improve algorithms and feature extraction for more reliable demographic identification from physiological signals.