COMPUTER IDENTIFICATION OF BACTERIA BY COLONY MORPHOLOGY

Abstract Species of bacteria important in medical diagnosis, public health surveys, industrial applications, and biological research can often be identified by the appearance of colonies they form when growing on solid agar. Colonies are located and counted automatically by flying spot scanner examination of photographs. Measurements of colony diameter and optical density profiles on colony diameters permit identification that is better than 90% correct for eight strains already characterized. Extending the library to 50 strains with accuracies exceeding 90% seems quite feasible.