Colorimetry-based and smartphone-assisted machine-learning model for quantification of urinary albumin

Abstract

Abstract The presence of albumin in the urine is indicative of kidney damage and can occur due to several underlying conditions, such as diabetes. The concentration of albumin in urine is used for the diagnosis and staging of chronic kidney disease (CKD). In clinical samples, the detection of albumin at lower concentrations is crucial for the early diagnosis and monitoring of CKD. Current urine analyzers precisely quantify albumin but are expensive and difficult to use in point-of-care (PoC) settings. Here, we demonstrate the quantification of albumin concentration in a urine sample using colorimetry. This model presents an accessory-free urine analyzer that uses a smartphone and customized machine-learning algorithms. Here, a urine sample is introduced onto a chemically impregnated dipstick that exhibits a change in color with the amount of albumin. Images of the urine dipsticks are captured using a smartphone camera under different illumination/experimental conditions and are processed to extract changes in the color values arising due to changes in the concentration of urinary albumin. Albumin concentrations are estimated from changes in color values. We used customized machine-learning algorithms to classify albumin concentrations and mitigate the effect of ambient light conditions. The k-nearest neighbor algorithm yielded an average classification accuracy of 96% with a detection limit of 4 mg l−1. The proposed scheme can be extensively used to monitor albumin concentration in PoC settings.