[Working Temperature Predication of Artificial Heart Based on Neural Network]

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Medicine
• Computational Physiology

Background:

• Accurate temperature monitoring is crucial for artificial heart function and patient safety.
• Sensor limitations in implanted devices necessitate indirect temperature prediction methods.
• Artificial neural networks offer potential for complex physiological system modeling.

Purpose of the Study:

• To develop a sensorless method for predicting artificial heart working temperature.
• To evaluate the efficacy of a genetically optimized backpropagation neural network (BPNN) for this task.
• To compare the performance of a standard BPNN against the optimized version.

Main Methods:

• Utilizing detectable external parameters to infer internal pump temperature.
• Implementing a standard backpropagation neural network (BPNN) model.
• Applying genetic algorithms to optimize the BPNN parameters for enhanced prediction.
• Conducting comparative analysis between the standard and optimized BPNN models.

Main Results:

• The genetically optimized BPNN demonstrated superior performance in temperature prediction.
• The optimized model achieved a prediction accuracy with a margin of error less than 1% in 1.84% of cases.
• The study successfully validated the feasibility of sensorless temperature prediction for artificial hearts.

Conclusions:

• Genetic algorithm optimization significantly enhances the accuracy of BPNN for artificial heart temperature prediction.
• Sensorless temperature monitoring is a viable approach for implantable devices, overcoming environmental limitations.
• This predictive model holds promise for improving the safety and reliability of artificial heart technology.