PetriPlate AI: Automating Colony Counting and Serial Dilution with Computer Vision

Manual quantification of Colony Forming Units (CFU) on agar plates is one of the most tedious, error-prone tasks in biotechnology, microbiology, and quality control. Studies show inter-operator counting variability of 15% to 32% on plates with clustered colonies, spreading morphologies, or agar meniscus reflections. PetriPlate AI applies advanced browser-based computer vision to automate this workflow with sub-second accuracy.

1. The Algorithmic Pipeline

Automating colony detection requires a multi-stage image processing pipeline that isolates the agar petri dish and suppresses background reflections:

1. Circular ROI Extraction: The petri dish outer rim is identified using a modified circular Hough transform, masking out benchtop clutter, petri edges, and hand shadows.
2. Adaptive Color Thresholding & Contrast CLAHE: Contrast Limited Adaptive Histogram Equalization normalizes uneven illumination caused by room lighting or flash reflections.
3. Euclidean Distance Transform & Watershed Segmentation: Clustered colonies that touch or overlap are separated by computing distance maps from colony edges to local centroids.
4. Size and Circularity Filtering: Artifacts such as air bubbles, dust specks, and agar tears are filtered out by rejecting contours with low circularity (4π × Area / Perimeter² < 0.65).

2. Serial Dilution Mathematics and Standard Errors

Once counts are obtained, calculating microbial concentration requires weighting across dilution series:

CFU / mL = (Σ C) / [ (n₁ + 0.1 × n₂) × d × V_plated ]

Where ΣC is the sum of colonies across countable plates (30–300 colonies), n₁ and n₂ are the number of replicates at consecutive dilution steps, d is the primary dilution factor, and V_plated is the volume inoculated (mL). PetriPlate AI pairs with our online scientific suite to output regulatory-compliant laboratory reports.