The Real-Time and Patient-Specific Prediction for Duration and Recovery Profile of Cisatracurium Based on Deep...

Area of Science:

• Anesthesiology
• Artificial Intelligence
• Biomedical Engineering

Background:

• Monitoring patient muscle relaxation during general anesthesia is critical.
• Current Train-of-four ratio (TOFR) monitoring provides only static data.
• Cisatracurium is a commonly used muscle relaxant requiring careful management.

Purpose of the Study:

• To develop real-time prediction models for cisatracurium's TOFR.
• To enable patient-specific and timely TOFR evaluation for anesthesiologists.
• To explore the application of deep learning in anesthesia monitoring.

Main Methods:

• Recurrent Neural Network (RNN), Gated Recurrent Unit (GRU), and Long Short-Term Memory (LSTM) models were employed.
• Transfer learning utilizing patient similarity (BMI, age) was implemented for personalized prediction.
• Model performance was evaluated for accuracy in TOFR prediction.

Main Results:

• The GRU model demonstrated superior performance in TOFR prediction compared to RNN and LSTM.
• Transfer learning based on patient similarity significantly improved prediction accuracy over random model selection.
• The study confirmed the feasibility of real-time, patient-specific TOFR prediction for cisatracurium.

Conclusions:

• Real-time TOFR prediction for cisatracurium is feasible and holds practical significance in general anesthesia.
• Patient-specific prediction using transfer learning enhances clinical applicability and precision medicine.
• Deep learning models offer a promising approach for dynamic muscle relaxation monitoring.