Development of an automated interpretation system for urine test strips using feature detection and color comparison processing in images taken using a smartphone

Abstract

　Interpretations of urine test strips are typically determined by visually comparing urine test strips to a color chart, leading to different interpretations among individuals. The purpose of this study was to improve the accuracy of the urine qualitative evaluation program by improving the reproducibility of the results and reducing the evaluator′s subjective influence. An automated urine test strip colorimetric program had been previously developed using Python as an objective evaluation method for urine test strips. The current study optimized and standardized the camera shooting environment; color temperature, lighting, contrast, urine test strip use timing, and feature detection. Accelerated KAZE was used for image feature point extraction processing. The imaging environment was adjusted so that elements such as the color temperature, contrast, and gamma value could be compared under the same camera settings by simultaneously imaging the urinalysis test strip and the color chart. Negative judgments, which are normal values, were consistent with the results of the automatic urine analyzer for all items. Outlier positive results were consistent, except for the occult blood item. Urine test strip interpretation accuracy improved, compared to previously published automated urine test strip colorimetric program methods, by using accelerated KAZE and adjusting the photography environment.