A Comparison of Bacterial Colonies Count from Petri Dishes Utilizing Hough Transform and Traditional Manual Counting

TL;DR

This study evaluates the Hough Transform method for automated bacterial colony counting on petri dishes, showing high accuracy and efficiency improvements over manual counting, with potential for high-throughput microbiological analysis.

Contribution

It introduces a novel application of the Hough Circle Transform for bacterial colony detection and compares its performance to manual counting, highlighting its advantages and limitations.

Findings

Automated counting achieved up to 95% accuracy for E. coli.

Automation reduced counting time from 70 to 30 seconds per plate.

Error rates varied among bacteria, with higher errors in S. aureus and P. aeruginosa.

Abstract

Bacterial colony enumeration is an essential stage in microbiological research, allowing susceptibility to antibiotics assessment, monitoring of the environment, and clinical diagnostics. Traditional manual counting methods are costly and susceptible to human mistakes, prompting the creation of automated detection systems. This research compares the efficacy of the Hough Circle Transform method for automated colony detection to hand counting of E. coli, S. aureus, and P. aeruginosa colonies on 200 petri plates. These bacteria are among the most clinically relevant pathogens, with E. coli frequently causing urinary tract infections, S. aureus connected with skin and bloodstream infections, and P. aeruginosa a significant issue in hospital-acquired infections. When colonies were counted automatically without visual correction, the mean difference from manual counts was 59.7%, with overestimation and underestimation occurring in 29% and 45% of cases, whereas S. aureus and P. aeruginosa had higher error rates. The proposed methodology achieved an overall accuracy of 95% for E. coli, 90% for S. aureus, and 84% for P. aeruginosa, with associated recall values of 95%, 91%, and 86%. The F-measure remained continuously high, ranging between 0.85 and 0.95. Regarding efficiency, manual counting required an average of 70 seconds per plate, while automated counting without and with visual correction took 30 seconds. Despite issues with segmentation in high-density plates, automated approaches offer a potential approach to high-throughput bacterial enumeration by decreasing labor-intensive operations while retaining adequate accuracy. Future research should enhance colony algorithmic segmentation and picture preprocessing approaches to improve detection performance, especially on crowded petri plates.