Autoregulatory Efficiency Assessment in Kidneys Using Deep Learning

Abstract

A convolutional deep neural network is employed to assess renal autoregulation using time series of arterial blood pressure and blood flow rate measurements in conscious rats. The network is trained using representative data samples from rats with intact autoregulation and rats whose autoregulation is impaired by the calcium channel blocker amlodipine. Network performance is evaluated using test data of the types used for training, but also with data from other models for autoregulatory impairment, including different calcium channel blockers and also renal mass reduction. The network is shown to provide effective classification for impairments from calcium channel blockers. However, the assessment of autoregulation when impaired by renal mass reduction was not as clear, evidencing a different signature in the hemodynamic data for that impairment model. When calcium channel blockers were given to those animals, however, the classification again was effective.

Keywords: biomedical signal processing; machine learning; nephrology; neural networks; physiology.

Fig. 1.

Deep Neural Network architecture used for classifying intact vs impaired BP/RBF data snippets. The input (12000 × 2) data comes from one minute of BP and RBF data sampled at 200 Hz. The output of the network is from sigmoid activation units taking values between zero and one. Score zero indicates intact AR, and score one indicates AR impairment.

Fig. 2.

Score pdf for test data from rats with intact AR and rats receiving amlodipine, concentration 100 mg/l. The network was trained to distinguish betweeen training data from these two groups. Intact rats (blue) number 78, with 71 scoring below 0.5. Amlodipine 100 rats (red) number nine, with all nine scoring above 0.5.

Fig. 3.

Score pdf for test data from rats receiving amlodipine, concentration 200 mg/l. Number of rats is five, with all five scoring above 0.5.

Fig. 4.

Score pdf for test data from rats receiving mibefradil at two fractions(0.065% and 0.1%). Number of rats is eight, with all eight scoring above 0.5.

Fig. 5.

Score pdf for test data from the RK-NX group. Number of rats is 76, with 21 scoring above 0.5 and 55 scoring below 0.5.

Fig. 6.

Score pdf for test data from RK-NX rats also receiving amlodipine. Number of rats is 21, with 19 scoring above 0.5 and two scoring below 0.5.

Fig. 7.

ROC for test data from rats with intact AR and rats receiving amlodipine, concentration 100 mg/l. The area under the ROC is 0.9858.