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

email@asn-online.org

202-640-4660

The Latest on X

Kidney Week

Please note that you are viewing an archived section from 2020 and some content may be unavailable. To unlock all content for 2020, please visit the archives.

Abstract: SU-OR37

Development of a Deep Learning Model to Predict ESKD in Patients with Immunoglobulin A Nephropathy (IgAN) at Kidney Biopsy Time

Session Information

Category: Glomerular Diseases

  • 1203 Glomerular Diseases: Clinical, Outcomes, and Trials

Authors

  • Schena, Francesco Paolo, Schena Faoudation, Bari, Italy
  • Anelli, Vito walter, Polytecnic of Bari, Bari, Italy
  • Trotta, Joseph, Polytecnic of Bari, Bari, Italy
  • Di noia, Tommaso, Polytecnic of Bari, Bari, Italy
  • Manno, Carlo, Universita degli Studi di Bari Aldo Moro, Bari, Puglia, Italy
  • Tripepi, Giovanni, CNR -IFC, Reggio Calabria, Italy
  • D'Arrigo, Graziella, CNR -IFC, Reggio Calabria, Italy
  • Russo, Maria Luisa, Fondazione Ricerca Molinette, Torino, Piemonte, Italy
  • Stangou, Maria J., Aristotle University of Thessaloniki School of Medicine, Nephrology, Thessaloniki, Greece
  • Papagianni, Aikaterini A., Aristotle University of Thessaloniki School of Medicine, Nephrology, Thessaloniki, Greece
  • Zoccali, Carmine, CNR -IFC, Reggio Calabria, Italy
  • Tesar, Vladimir, University of Prague, Prague, Czechia
  • Coppo, Rosanna, Fondazione Ricerca Molinette, Torino, Piemonte, Italy
Background

Many prediction models to support clinical decision making have been developed for decades but they are based on traditional statistic linear methods. Another approach is the application of artificial intelligence (AI) that is based on machine learning or deep learning algorithms. We developed an artificial neural network (ANN) tool to predict ESKD in IgAN patients at kidney biopsy time.

Methods

The classifier model to predict ESKD was composed of 4 hidden layers with 100 neurons in each layer. The regression model to predict the time-to-event endpoint consisted of 3 layers containing 125 neurons each.

Results

Our tool, based on these two models, was developed in a cohort of 948 IgAN patients of the VALIGA and Greek cohort. Then, the tool was validated in an independent cohort of 167 IgAN patients from 6 nephrology units. After Cox's regression analysis 7 variables (age, sex, blood pressure values, serum creatinine, daily proteinuria, MESTC classification for the kidney biopsy and therapy at baseline) were chosen to develop the ANN model. The AUC of the ANN model in the study cohort was 0.80. The performance was 0.82 (precision 0.83, accuracy 0.80) for ESKD prediction at 5 years of follow-up and 0.89 (precision 0.81; accuracy 0.83) for patients with 10 years of follow-up. Stable renal function and ESKD were correctly predicted in 91% of IgAN patients in the test cohort.

Conclusion

(i) Our ANN is a promising alternative to the mathematical models in solving non-linear and multidimensional problems. (ii) We have developed a new clinical decision support system that provides additional information to identify IgAN patients at high risk of ESKD. (iii) This tool may stratify patients in the context of a personalized therapy. (iv) This tool will be validated in a clinical prospective study.

Funding

  • Government Support - Non-U.S.