Toward organism identification through analysis of bacterial colonies.

The clinical problem of bacterial identification has been approached by applying pattern-recognition techniques to multi-wavelength surface-scattering and reflectance data derived from real-time scans of isolated colonies. Preliminary results, obtained from blood-agar plates inoculated with a mixture of staphylococci, streptococci and escherichieae, indicate that these organisms can be differentiated with better than 90% certainty, provided the colonies are physically separated and their growth conditions closely controlled. Data collection and classification characteristics of the experimental system are briefly described; it is felt that the technique, possibly expanded to include boundary characteristics of the colonies, may offer a viable means of identifying clinically important bacteria.