ColiFind: A Digital Image Analysis Application to Identify E. coli colonies in Coliscan Easygel Water Quality Tests

Track 4A: Monitoring Water Quality
Friday, October 13, 2017, 10:00 am – 11:50 am

Ana Humphrey
T.C. Williams High School

I developed ColiFind, a digital image analysis application that identifies and counts E. coli colonies in images of Coliscan Easygel tests to create an inexpensive, simple, and accurate method for monitoring waterborne bacteria. ColiFind employs five digital E. coli signatures in the images’ color intensity, saturation, and pattern correlation. ColiFind was assessed using images taken with and without following an imaging protocol. Gen-6, the current code generation, had a 92% detection rate and a 0.07 false positive ratio with protocol images and a 76% detection rate and a 0.20 false positive ratio with non-protocol images. Handwritten labels on Petri dish lids caused 89% of these false positives and could be eliminated by removing lids when photographing dishes. E. coli detection could increase with an evaluation function that isolates and classifies colonies on the edge of ColiFind’s classification parameters. With these changes, ColiFind is expected to achieve an overall accuracy over 95%. ColiFind is designed to become the core of a monitoring system that enables water quality organizations to compile comprehensive, accurate and low-cost databases of waterborne E. coli levels.