Smartphone-Based Point-of-Care Urinalysis Assessment.

Area of Science:

• Biomedical Engineering
• Medical Diagnostics
• Machine Learning Applications

Background:

• Dipstick urinalysis is a common diagnostic tool.
• Manual color assessment is subjective and prone to errors.
• Existing automated urine analyzers are expensive.

Purpose of the Study:

• To develop a smartphone-based, machine learning approach for automated dipstick urinalysis.
• To provide a cost-effective and accurate alternative to manual and benchtop methods.

Main Methods:

• Utilized a unique calibration chart and multivariate linear regression for color correction.
• Employed least Euclidean distance for reagent pad color matching.
• Tested with five smartphone cameras and three illumination settings.

Main Results:

• Demonstrated high accuracy in reagent pad color assessment using synthetic dipsticks.
• The method effectively corrects for camera distortions and ambient light variations.
• Experimental results show consistent performance across different devices and conditions.

Conclusions:

• The smartphone-based machine learning method is accurate, efficient, and affordable for automated urinalysis.
• This technology can be deployed in clinical, at-home, and point-of-care settings.
• Offers a viable alternative to manual interpretation and expensive urine analyzers.