ASN's Mission

ASN leads the fight to prevent, treat, and cure kidney diseases throughout the world by educating health professionals and scientists, advancing research and innovation, communicating new knowledge, and advocating for the highest quality care for patients.

learn more

Contact ASN

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

email@asn-online.org

202-640-4660

The Latest on Twitter

Kidney Week

Abstract: PO2068

Deep Learning Analysis of Derived Cardiac Function Metrics for the Detection of CKD and Subsequent Outcome Prediction in Community-Dwelling Individuals

Session Information

Category: Hypertension and CVD

  • 1401 Hypertension and CVD: Epidemiology, Risk Factors, and Prevention

Authors

  • Cosgrave, Niall, University of Dublin Trinity College, Dublin, Ireland
  • O'Connor, John David, University of Dublin Trinity College, Dublin, Ireland
  • Mellotte, George S., National Renal Office, Health Service Executive of Ireland, Dublin, Ireland
  • Griffin, Brenda, Trinity Health Kidney Center, St James's Hospital & Tallaght University Hospital, Dublin, Ireland
  • O'Seaghdha, Conall M., Beaumont Hospital, Dublin, Ireland
  • Reule, Scott, Minneapolis VA Health Care System, Minneapolis, Minnesota, United States
  • Foley, Robert N., United States Renal Data System Coordinating Centre, Minneapolis, Minnesota, United States
  • Knight, Silvin P., University of Dublin Trinity College, Dublin, Ireland
  • Little, Mark Alan, University of Dublin Trinity College, Dublin, Ireland
  • Sexton, Donal J., University of Dublin Trinity College, Dublin, Ireland
Background

Whether subtle abnormalities in cardiac function exist in community dwelling individuals and can be used to agnostically discern those with reduced kidney function and incident adverse outcomes over follow up is unknown.

Methods

The Irish Longitudinal Study on Ageing (TILDA) is a prospective nationally representative cohort study based on random sampling of the community-dwelling general population aged ≥ 50 years in Ireland. Wave 1 was performed between June 2009-June 2011. Participants underwent a detailed health assessment including blood tests, and an active stand test using the Finometer, which measures continuous blood pressure and heart rate for 10 mins while supine at rest, then throughout the standing test and for 2 mins thereafter. Cardiac function metrics are derived : left ventricular ejection time (LVET), cardiac output, and total peripheral resistance. CKD-EPI equation was used for eGFR. We analysed repeated measures data at 10 second intervals over the entire observation period using sequential neural networks with the categorical outcomes of coincident CKD, and incident mortality. Follow up was approximately 10 yrs. Python v 3.77 and TensorFlow v2.0.0 were used for the analysis.

Results

N=4388 TILDA participants were included, N=2013 were male, mean age was 62 (8) yrs, 647 had CKD. 178 died over follow up. Figure 1. A demonstrates profiles in LVET by eGFR category and B an ROC curve output from the Neural Network model predicting conincident CKD AUC = 0.81(resampled model: which included an input layer, 2 hidden layers and an output layer based on cardiac metrics, age and sex). AUC for mortality was 0.75 (resampled model).

Conclusion

Deep Learning analysis of repeated measures of derived cardiac function metrics discerned community dwelling individuals with CKD and subsequent outcomes. This method has potential for population level risk dsicrimination.