Using elastography-based multilayer perceptron model to evaluate renal fibrosis in chronic kidney disease

Area of Science:

• Nephrology
• Medical Imaging
• Machine Learning

Background:

• Chronic kidney disease (CKD) requires noninvasive methods for assessing renal fibrosis severity.
• Accurate risk stratification is crucial for managing progressive kidney disease.

Purpose of the Study:

• To develop and validate a multilayer perceptron (MLP) model for assessing renal fibrosis in CKD patients.
• To integrate real-time two-dimensional shear wave elastography (2D-SWE) with clinical variables for fibrosis assessment.

Main Methods:

• A prospective study included 162 CKD patients undergoing kidney biopsy and 2D-SWE.
• Renal cortex stiffness was measured using 2D-SWE, with patients categorized by fibrosis severity.
• An MLP classifier was trained using elastic values and clinical features, with performance evaluated in training and test cohorts.

Main Results:

• The MLP model showed good calibration and discrimination in both training (AUC=0.93) and test (AUC=0.86) cohorts.
• Decision curve and clinical impact analyses indicated a positive clinical utility with minimal adverse effects.

Conclusions:

• The developed MLP model effectively identifies the risk of moderate-severe renal fibrosis in CKD patients.
• This noninvasive approach can aid in clinical management and treatment decisions for CKD.