Autoregulatory Efficiency Assessment in Kidneys Using Deep Learning

Area of Science:

• Physiology
• Biomedical Engineering
• Machine Learning

Background:

• Renal autoregulation is crucial for maintaining stable kidney function.
• Assessing autoregulation impairment is vital for understanding kidney disease.
• Current methods for assessing renal autoregulation can be complex and invasive.

Purpose of the Study:

• To develop and evaluate a convolutional deep neural network for assessing renal autoregulation.
• To determine the network's ability to classify different types of autoregulatory impairment.
• To investigate the network's performance with various pharmacological and surgical models of impairment.

Main Methods:

• Utilized time series data of arterial blood pressure and blood flow rate in conscious rats.
• Trained a convolutional deep neural network on data from rats with intact and impaired autoregulation.
• Tested the network using data from different impairment models, including calcium channel blockers and renal mass reduction.

Main Results:

• The network achieved effective classification of autoregulation impairment induced by calcium channel blockers.
• Assessment of autoregulation impairment due to renal mass reduction was less clear, indicating distinct hemodynamic signatures.
• Combined impairment (renal mass reduction plus calcium channel blockers) was effectively classified.

Conclusions:

• Convolutional deep neural networks show promise for non-invasive assessment of renal autoregulation.
• The network's performance varies depending on the specific model of autoregulatory impairment.
• Further refinement may be needed to accurately assess impairment from structural changes like renal mass reduction.