A Model to Predict Heartbeat Rate Using Deep Learning Algorithms.

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Healthcare
• Cardiovascular Monitoring

Background:

• The COVID-19 pandemic necessitated contactless health monitoring solutions.
• Traditional electrocardiography (ECG) requires physical contact, posing challenges during pandemics.
• There is a need for non-invasive methods to assess cardiac function.

Purpose of the Study:

• To develop a dependable, touchless technique for estimating heart rate using reflected light from the skin.
• To leverage deep learning models for accurate heart rate detection from video streams.
• To validate the proposed method's performance against established metrics.

Main Methods:

• Utilized deep learning models including AlexNet, Convolutional Neural Networks (CNNs), Long Short-Term Memory Networks (LSTMs), and ResNet50V2.
• Processed video streams by converting them into frames and images for analysis.
• Conducted trials on volunteers to assess accuracy, Mean Absolute Error (MAE), and Mean Squared Error (MSE).

Main Results:

• Achieved an average accuracy of 99.78% when combining LSTMs and ResNet50V2.
• Reported a Mean Absolute Error (MAE) of 0.142 and Mean Squared Error (MSE) of 1.82.
• Demonstrated superior performance compared to existing literature methods in terms of MAE and MSE.

Conclusions:

• The developed touchless heart rate estimation method is viable and accurate.
• The algorithm shows significant potential for application in healthcare facilities, assisting physicians.
• This contactless approach addresses public health concerns while enabling effective cardiac monitoring.