Caleb Kumar

Designing and Validating a Java Program to Diagnose Bladder Cancer

Caleb Kumar

Blaine, MN
Grade: 12

I developed JAVA software to diagnose bladder cancer from urine cytology images. First, I wrote algorithms to determine size, color and curvature of cancerous and benign cells in bladder-cell images. Second, I developed software that defines and then compares characteristics of undiagnosed cell images to pre-classified images. Validation of my software showed accuracy of 99.8% in identifying suspicious bladder cells and accuracy of 87.1% and 92.2% for differentiating normal from abnormal bladder cells, respectively. Holistically, accuracy of my software was deemed as "good" according to a kappa method defined by Altman (1991). A diagnosis for a single bladder-cell image takes less than ten seconds.