Machine Learning Enabled Portable Electrical Impedance Tomography System for Community Level Chronic Kidney Disease...

Area of Science:

• Biomedical Engineering
• Medical Diagnostics
• Machine Learning in Healthcare

Background:

• Chronic kidney disease (CKD) affects 850 million globally, presenting significant healthcare challenges due to late detection and costly treatments.
• Current diagnostic methods (serum creatinine, urinary albumin) are invasive, expensive, and lab-dependent, limiting remote monitoring.
• There is an unmet need for accessible, non-invasive CKD screening and eGFR estimation, particularly for community and home-based settings.

Purpose of the Study:

• To develop and validate a novel, portable, non-invasive approach for CKD screening and eGFR estimation using electrical impedance tomography (EIT) and machine learning.
• To assess the feasibility of integrating patient-specific anthropometric data with EIT signals for enhanced diagnostic accuracy.
• To demonstrate the potential of this system for point-of-care CKD detection and remote monitoring.

Main Methods:

• A portable EIT system was used to assess 138 subjects (healthy volunteers and CKD patients).
• Renal EIT signals were processed using principal component analysis (PCA) for feature extraction.
• A two-stage machine learning model (SVM for classification, XGBoost for eGFR prediction) was developed, incorporating anthropometric parameters.

Main Results:

• The EIT-based system achieved 93% accuracy in classifying healthy individuals from CKD patients (Stages 2-4), with 100% sensitivity and 94% specificity (AUC-ROC: 0.97).
• The eGFR prediction model achieved an R² value of 0.57 (p < 0.001), demonstrating significant differences across CKD stages.
• The proposed approach outperformed anthropometric-based models in both classification and regression tasks.

Conclusions:

• A low-cost, portable EIT system combined with machine learning is feasible for point-of-care CKD detection and eGFR estimation.
• This non-invasive system has the potential to enable widespread CKD screening in community and telemedicine settings, improving early detection and management.
• The technology addresses the urgent need for accessible, self-administrable CKD monitoring, potentially reducing healthcare costs.