MCount: An automated colony counting tool for high-throughput microbiology

Area of Science:

• Microbiology
• Bioinformatics
• Image Analysis

Background:

• Accurate microbial colony counting is essential for high-throughput workflows.
• Existing automated solutions face challenges with merged colonies, impacting accuracy.
• There is a need for robust automated colony counting methods that handle merged colonies effectively.

Purpose of the Study:

• To introduce MCount, a novel automated colony counting solution.
• To address the limitation of merged colonies in existing automated counting methods.
• To provide a user-friendly and accurate tool for microbial colony enumeration.

Main Methods:

• MCount integrates contour information with regional algorithms for colony counting.
• It optimizes the pairing of contours with regional candidate circles to infer merged colonies.
• Statistical methods are proposed for hyperparameter selection with limited or unlabeled data.

Main Results:

• MCount achieved an average error rate of 3.99% on an Escherichia coli dataset.
• This significantly outperforms existing methods like NICE (16.54%), AutoCellSeg (33.54%), and OpenCFU (50.31%).
• The method demonstrates consistent low error rates even with limited labeled data.

Conclusions:

• MCount offers a significant advancement in automated microbial colony counting, particularly for merged colonies.
• Its user-friendly nature and adaptability to limited data scenarios make it suitable for various applications.
• MCount provides an accurate and efficient solution for high-throughput microbial analysis.