A Correlation-Based Anomaly Detection Model for Wireless Body Area Networks Using Convolutional Long Short-Term...

Area of Science:

• Biomedical Engineering
• Data Science
• Internet of Healthcare Things (IoHT)

Background:

• Wireless Body Area Networks (WBANs) are crucial for remote patient monitoring within the Internet of Healthcare Things (IoHT).
• Reliable physiological data collection is essential for timely clinical decisions.
• Existing anomaly detection methods in WBANs struggle with big data and novel anomalous patterns.

Purpose of the Study:

• To develop an effective anomaly detection model for WBAN data streams.
• To address the limitations of traditional statistical and machine learning approaches in handling WBAN data.
• To improve the accuracy and reliability of physiological data for healthcare applications.

Main Methods:

• Proposed a hybrid model combining Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks (ConvLSTM).
• Leveraged correlations between different physiological data attributes.
• Applied the model to detect both point anomalies and contextual anomalies in WBAN big data streams.

Main Results:

• The proposed ConvLSTM model achieved an average F1-measure of 98% and 99% accuracy.
• Demonstrated superior performance compared to standalone CNN and LSTM models, which achieved 64%.
• Effectively detected anomalies across different subjects and datasets.

Conclusions:

• The hybrid ConvLSTM model significantly enhances anomaly detection in WBAN data streams.
• This approach improves the reliability of healthcare data, supporting better clinical decision-making.
• The model shows promise for robust anomaly detection in the evolving Internet of Healthcare Things.