Systematic evaluation of isolation processes of microorganisms using spatial statistics

The evaluation of inoculation processes of microorganisms by robotic systems as well as by lab-technicians is compensable and can be missing consistency as human judgment will depend on the individual and may therefore be biased and less effective than models and algorithms evaluating spatial patterns. To address this problem, nearest neighbor analysis was used to investigate if it could be utilized as a method to evaluate isolation processes. The nearest neighbor analysis results in a comparable numeric value on the isolation process, which can be used to assess results of different inoculation processes.

In this article, images of Petri dishes and simulated plates are used to investigate the effectiveness of nearest neighbor analysis, which is a method within spatial statistics. This analysis is applied to spatial data created by applying computer vision to localize the colonies on the plates.

When evaluating plates made with the streaking technique method, it was found to be ineffective as the dense parts of the distribution resulted in the computer vision being unable to locate all of the colonies. Therefore, the nearest neighbor analysis is not suitable to evaluate streaking plates and other methods to evaluate such plates should be developed. However, when evaluating Petri dishes where the spread plating technique had been applied, it was found that nearest neighbor analysis can be a useful way to systematically evaluate isolation processes.