TOP-Net Prediction Model Using Bidirectional Long Short-term Memory and Medical-Grade Wearable Multisensor System for...

Area of Science:

• Artificial Intelligence in Medicine
• Cardiology
• Medical Informatics

Background:

• Tachycardia (tachyarrhythmia) can lead to severe health issues like heart failure and cardiac arrest if not diagnosed and treated promptly.
• Current diagnostic methods for tachycardia require enhancement for early risk detection by physicians.

Purpose of the Study:

• To develop a deep learning model named TOP-Net for early prediction of tachycardia onset.
• To utilize easily accessible data, including vital signs from wearable systems and electronic health records, for the model.

Main Methods:

• The TOP-Net model employs a bidirectional long short-term memory deep learning architecture.
• Data integration included vital signs (heart rate, respiratory rate, SpO2) from wearables and patient information (age, gender, medical history) from EHRs.
• Model training occurred in an ICU setting, with subsequent transfer to a general ward for real-world validation, evaluating performance using six key metrics.

Main Results:

• TOP-Net demonstrated superior predictive performance compared to baseline models in an ICU setting, predicting tachycardia onset up to 6 hours in advance.
• In a general ward setting, the transferred TOP-Net model achieved high performance metrics, predicting onset 2 hours in advance.
• Optimal prediction accuracy was achieved by utilizing comprehensive statistical information from vital signs like heart rate, respiratory rate, and SpO2.

Conclusions:

• TOP-Net effectively predicts tachycardia onset using readily available data from wearable sensors and electronic health records.
• The model's validated performance surpasses baseline models in predicting tachycardia onset in both intensive care and general ward environments.
• TOP-Net's accessibility and data integration capabilities can significantly aid clinicians in the early identification of patients at risk for tachycardia, both in hospitals and at home.