A neural network approach to glomerular filtration rate estimation: a single-centre retrospective audit

Abstract

Objectives: The 2009 Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation without race correction factor is frequently used for an estimate of glomerular filtration rate (eGFR) and to support a single-sample GFR regime. This study examines whether neural networks offer a potential means to improve the accuracy of GFR estimates using the same initial inputs as eGFR.

Methods: An audit of 865 adult GFR examinations and serum creatinine measurements between January 2010 and 2024 was undertaken. Patient sex, age, creatinine, and measured GFR were used to train a neural network (NN) model with an 80 : 20 train-test split, with test set root mean square error (RMSE), accuracy, median bias, and sensitivity calculated and compared against the 2009 CKD-EPI equation eGFR.

Results: NN GFR showed an improved performance against the 2009 CKD-EPI equation in RMSE: 12.0 vs. 16.6 mL/min/1.73 m 2 ( P < 0.001), median bias: -2.50 vs. 7.86 mL/min/1.73 m 2 ( P < 0.001) and accuracy: 94.2 vs. 83.2% ( P < 0.001). Both NN GFR and the eGFR equation had poor sensitivity across the British Nuclear Medicine Society single-sample ranges of 25-50, 50-70, 70-100, and >100 mL/min/1.73 m 2 : 57.9 vs. 57.9%, 50.0 vs. 26.9%, 84.4 vs. 54.2%, 10.0 vs. 70.0%.

Conclusion: This study has suggested that locally trained NNs can offer a potential avenue to improve GFR predictions, even on small and diverse datasets.

Advances in knowledge: Although the model is not sufficiently sensitive to predict the optimum time-sample point for a single-sample regime, this work can serve as a proof of concept for UK-specific NN GFR models.

Keywords: GFR; eGFR; glomerular filtration rate; machine learning; neural network.

Fig. 1

Plots of residuals and NN GFR, along with Q–Q plots (a) and histograms (b) for the training and test set. The R² scores for the training and test set were 0.692 and 0.702, respectively. GFR, glomerular filtration rate; NN, neural network; Q-Q, quantile-quantile.

Fig. 2

NN GFR vs. mGFR for the test set (n = 173). NN GFR had a smaller RMSE of 12.0 mL/min/1.73 m² and a higher accuracy of 94.2%. GFR, glomerular filtration rate; mGFR, measured GFR; NN, neural network; RMSE, root mean square error.

Fig. 3

eGFR vs. mGFR for the test set (n = 173). The eGFR had a RMSE of 16.6 mL/min/1.73 m² and an accuracy of 83.2%. CKD-EPI, Chronic Kidney Disease Epidemiology Collaboration; eGFR, estimated glomerular filtration rate; mGFR, measured glomerular filtration rate; NN, neural network; RMSE, root mean square error.

Fig. 4

Bland–Altman NN GFR/eGFR vs. measured GFR. The NN GFR and eGFR had mean biases of –1.50 and 8.49 mL/min/1.73 m², respectively. CKD-EPI, Chronic Kidney Disease Epidemiology Collaboration; eGFR, estimated glomerular filtration rate; GFR, glomerular filtration rate; NN, neural network.

Fig. 5

(a–d) Plots of NN GFR vs. mGFR for the BNMS ranges 25–50, 50–70, 70–100, and more than 100 mL/min/1.73 m², with respective sensitivities of 57.9, 50.0, 84.4, and 10.0%. BNMS, British Nuclear Medicine Society; GFR, glomerular filtration rate; mGFR, measured glomerular filtration rate; NN, neural network.

Fig. 6

(a–d) Plots of eGFR vs. mGFR for the BNMS ranges 25–50, 50–70, 70–100, and more than100 mL/min/1.73 m², with respective sensitivities of 57.9, 26.9, 54.2, and 70.0%. BNMS, British Nuclear Medicine Society; CKD-EPI, Chronic Kidney Disease Epidemiology Collaboration; eGFR, estimated glomerular filtration rate; GFR, glomerular filtration rate; mGFR, measured glomerular filtration rate.