ASN's Mission

To create a world without kidney diseases, the ASN Alliance for Kidney Health elevates care by educating and informing, driving breakthroughs and innovation, and advocating for policies that create transformative changes in kidney medicine throughout the world.

learn more

Contact ASN

1401 H St, NW, Ste 900, Washington, DC 20005


The Latest on Twitter

Kidney Week

Abstract: SA-PO014

Renal Autoregulation Assessment in Conscious Rats Using Deep Learning Networks

Session Information

  • Bioengineering
    November 05, 2022 | Location: Exhibit Hall, Orange County Convention Center‚ West Building
    Abstract Time: 10:00 AM - 12:00 PM

Category: Bioengineering

  • 300 Bioengineering


  • Williamson, Geoffrey A., Illinois Institute of Technology, Chicago, Illinois, United States
  • Chopde, Purva R., Illinois Institute of Technology, Chicago, Illinois, United States
  • Alphonse, Sebastian Anand A., Livanova, Houston, Texas, United States
  • Polichnowski, Aaron J., East Tennessee State University, Johnson City, Tennessee, United States
  • Griffin, Karen A., Loyola University Chicago Stritch School of Medicine, Maywood, Illinois, United States

Renal autoregulation (AR) maintains constant GFR and prevents glomerular transmission of systemic pressure. AR assessment informs about susceptibility to hypertensive renal injury and thus may influence management of patients with chronic kidney disease (CKD). In rats, AR assessment has typically involved measuring the recovery of renal blood flow (RBF) after acute step blood pressure (BP) changes under anesthesia, via AR indices quantifying fractional changes in RBF relative to BP. We developed a Short Segment AR Index (SSARI) methodology for AR assessment in conscious animals (Bidani et al., JASN 31(2):324-336, 2020). This method locates adjacent short segments (0.5 to 20 s) in the BP/RBF record where mean BP changes by >5 mmHg and averages the AR index for those segment pairs over all pairs. Here, we develop and train a Deep Neural Network (DNN) to effect a similar AR assessment and compare its performance with that of SSARI. Our DNN design leverages our earlier work applying DNNs to differentiate intact and impaired AR (Alphonse et al., 2020 28th EUSIPCO, 1165-1169, 2021).


The DNN input employs a one min. long BP/RBF record sampled at 200 Hz. The DNN architecture includes four convolutional layers followed by two dense layers, each with leaky Relu activation, and max pooling after convolutional layers. The output is a scalar produced with linear activation. Using 2960 30-min. BP/RBF recordings, we trained the DNN to match the average SSARI score (using 2.5 s. segments) for the full 30 min. data set. The DNN was tested using a separate set of 3413 30-min. BP/RBF recordings.


With the test data, we achieved a mean-square error of 0.0555 between the average DNN score using all 30 1-min. segments and the average SSARI value for the full 30 min. Furthermore, averaged DNN scores had 1/10 the variability of the averaged SSARI values.


The DNN can distinguish between intact AR and impaired AR (e.g. via calcium channel blockade or renal mass reduction or both), just as does SSARI as we have reported. The DNN, however, provides equivalent accuracy in doing this with only 1/10 the data length. Further refinement of this methodology will allow us to leverage its use in the clinical setting as DNN will inform AR impairment and susceptibility to hypertensive renal injury in patients with CKD.


  • NIDDK Support