Detection and classification of urine components utilizing quantitative phase imaging and machine learning

Abstract

We explored the capabilities of quantitative phase imaging (QPI) with digital holographic microscopy (DHM) in combination with machine learning (ML) approaches for the characterization and classification of urine sediments. Bright-field images and off-axis holograms of a liquid control for urine analysis were acquired with a modular DHM system. From the retrieved images, particle morphology parameters were extracted by segmentation procedures. In addition, the ability of supervised ML-algorithms to classify and identify urine sediment components based on biophysical parameters was evaluated. The results demonstrate DHM in combination with ML as a prospective tool for urine analysis.