Efficacy of Dynamics-based Features for Machine Learning Classification of Renal Hemodynamics

Area of Science:

• Physiology
• Biomedical Engineering
• Machine Learning

Background:

• Renal hemodynamics analysis is crucial for understanding kidney function.
• Machine learning (ML) offers advanced analytical capabilities for complex physiological data.
• Distinguishing subtle variations in hemodynamic characteristics between animal groups is challenging.

Purpose of the Study:

• To develop and compare machine learning approaches for analyzing renal hemodynamics.
• To evaluate the effectiveness of different feature sets for ML-based classification of rats.
• To demonstrate the potential of ML in classifying Sprague-Dawley rats based on hemodynamic data.

Main Methods:

• Time series data of arterial blood pressure and renal blood flow rate were used.
• Machine learning models including deep neural network (DNN), random forest, support vector machine, and multilayer perceptron were employed.
• Classification was performed using raw physiological measurements and a feature vector derived from a nonlinear dynamic system.

Main Results:

• Deep neural networks utilizing raw data achieved high classification accuracy.
• Models using hemodynamics-based features also demonstrated promising classification performance, albeit slightly reduced.
• The study successfully classified Sprague-Dawley rats from different suppliers based on subtle hemodynamic differences.

Conclusions:

• Machine learning models can effectively classify rats using renal hemodynamic data.
• Hemodynamics-based features alone are sufficient for reasonably accurate ML classification.
• This work highlights the potential of ML applications in physiological research and animal colony management.