Automated CKD Detection via Image Processing Employing Delta-E Metric and CNN

Chronic Kidney Disease (CKD) is a global health concern that often progresses silently until advanced stages, making early detection critical. This paper presents a point-of-care system that automates the color comparison between a test strip and a reference strip to detect biomarkers like albumin associated with kidney dysfunction. The system leverages mobile-captured images, applying image processing techniques such as segmentation and color extraction, and compares the test strip's top color with reference colors on the parent strip using the Delta-E metric in the LAB color space. The closest match quantifies the biomarker concentration, providing a reliable diagnostic measure. To enhance diagnostic precision, this research incorporates Convolutional Neural Networks (CNNs) to refine image preprocessing and anomaly detection. The CNN model complements existing methods by mitigating environmental influences like lighting variability and improving feature extraction accuracy. This hybrid approach ensures consistency while adapting to diverse datasets and image conditions. The result is a cost-effective, fast, and accessible diagnostic tool that reduces manual errors and supports early CKD detection in resource-limited settings. Future work will focus on expanding the application of this system to detect other diseases and further optimize the integration of CNNs for enhanced scalability and robustness.