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

Abstract: FR-PO711

Predicting Mortality Risk in Neonatal Patients with AKI with an Artificial Neural Network Algorithm

Session Information

  • Pediatric Nephrology - 1
    October 25, 2024 | Location: Exhibit Hall, Convention Center
    Abstract Time: 10:00 AM - 12:00 PM

Category: Pediatric Nephrology

  • 1900 Pediatric Nephrology

Authors

  • Pandya, Aadi H., Akron Nephrology Associates/ Cleveland Clinic Akron General Medical Center, Akron, Ohio, United States
  • Vyas, Arnav, Akron Nephrology Associates/ Cleveland Clinic Akron General Medical Center, Akron, Ohio, United States
  • Dawson, Maximilian, Akron Nephrology Associates/ Cleveland Clinic Akron General Medical Center, Akron, Ohio, United States
  • Sethi, Sidharth Kumar, Akron Nephrology Associates/ Cleveland Clinic Akron General Medical Center, Akron, Ohio, United States
  • Kashani, Kianoush, Akron Nephrology Associates/ Cleveland Clinic Akron General Medical Center, Akron, Ohio, United States
  • Raina, Rupesh, Akron Nephrology Associates/ Cleveland Clinic Akron General Medical Center, Akron, Ohio, United States
Background

Acute Kidney Injury (AKI) presents a substantial burden in healthcare, especially in the neonatal intensive care unit. Physiological variability and limited biomarkers complicate care for neonatal patients, and artificial intelligence shows promise for clinical implications in neonatal AKI.

Methods

Five deep learning models were tested with two independent neonatal patient datasets sourced from eleven healthcare centers. Data preprocessing techniques were enforced to amalgamate these datasets, followed by SMOTE to augment the imbalanced deceased patient class. The data was subsequently split into training and testing cohorts. Initial modeling was conducted with 36 training features, after which feature reduction techniques, including Recursive Feature Elimination, were utilized to isolate the top ten most influential features, creating parsimonious models.

Results

From both datasets, 1,362 patients with a median age of 14.10 hours were included in this study. The chosen algorithm was a custom-constructed Keras sequential artificial neural network (ANN) architecture with excelling validation-cohort metrics: AU-ROC = 0.9859; AU-PRC = 0.9919; accuracy = 0.9731; sensitivity = 0.9657; specificity = 0.9805. Given ten patient parameters for a neonate patient at risk of AKI, the model calculates a mortality risk level, which can be further utilized to dictate the potential intensification of patient treatment procedures. The robustness and generalizability of the ANN-based model were confirmed through K-fold cross-validation studies.

Conclusion

This study implements an ANN algorithm to predict mortality risk for neonates susceptible to AKI based on ten input parameters. Utilizing this healthcare-based deep learning model provides insight to aid physicians in patient management.