Detecting significant bacteriuria through urine smear image analysis for urinary tract infection

Urinary tract infection is one of the most common bacterial infection in humans and a major cause for outpatient consults. Bacteria is found to be the cause of infection in more than 95% of cases. Detecting significant bacteriuria contributes immensely to the diagnosis, prognosis and treatment of the disease. Prevalent detection methods are either tedious, time-consuming or expensive. In this paper, a new computer-aided method is proposed for significant bacteriuria classification by analysing urine smears. A new protocol was devised for smear preparation on slides. A minimum of 10 distinct non-overlapping bright field microscopy images were captured from each slide. The bacteriuria classification was done based on the average number of bacteria present in these images. Our approach exploits color-intensity-based luminance thresholding for image segmentation. Object identification was performed by a new color-matching technique using a specifically designed color database. The proposed method was implemented on 37 urine samples. The results indicate that our method is a promising approach towards fully automating significant bacteriuria detection.