jonbenfri/opentrons-colony-counting-picking

Colony counting and picking functionality for OpenTrons

OpenTrons Hack-A-Tron 2019 Hackathon

May 10 6PM EDT - May 11 6PM EDT

Slack workspace: https://hack-a-tron.slack.com

Invite link

Join the #colony_counting channel.

Colony Counting and Picking

Description

Colony counting and picking functionality for OpenTrons.

Pre-prepped agar plates with visible colonies are loaded onto the OpenTrons deck. Colony location data is uploaded to OpenTrons. After calibration, OpenTrons picks individual colonies using pipettor for processing.

Ideas

• Add a camera to head assembly

  • Take photo
  • Add custom instrument
  • Add capture_image function

• Add agar plate labware

  • Calibrate robot position
  • https://docs.opentrons.com/examples.html?highlight=vector#precision-pipetting

• Add count_colonies function

  • ImageJ --headless: https://imagej.nih.gov/ij/plugins/colony-counter.html
  • Run macro: ImageJ --headless -macro custom_macro.ijm
  • Returns count

• Add second camera for depth detection

  • https://docs.opencv.org/3.1.0/dd/d53/tutorial_py_depthmap.html

• Add analyze_colonies function

  • Returns count
  • Returns list of colonies with (x,y,z) coordinates

• Adjust agar plate labware object to include (x,y,z) coordinates of each colony

  • Intuitive interface can be something like: plate.colonies('1')
  • Use center of mass for each colony (e.g. move_to plate.colonies('1') will go to center of mass or centroid)

• Alternative: Take image of agar plate, each pixel is a "well" e.g. 800 x 600 image = 480000 wells

• Image Normalization

  • Perspective correction using homography with OpenCV:
    • https://docs.opencv.org/3.4.1/d9/dab/tutorial_homography.html#tutorial_homography_Demo2
    • https://stackoverflow.com/questions/41995916/opencv-straighten-an-image-with-python

Instructions

This repository provides a way to import ImageJ regions of interest (ROI) data into python.

ImageJ outputs ROI data in a binary format. Dylan Muir provides a Matlab script to read binary ROI data to import into Matlab.

We use Octave, an open-source Matlab clone, to convert ROI data from binary to csv format for import into python.

Requirements

• Octave
• ImageJ
• ImageJ Colony Counter plugin
• OpenTrons

1. Obtain plate image
2. Run ImageJ Colony Counter on plate image (note location of ROI zip file output)
3. Edit variables in ColonyROI2csv.m (create DATA_DIR if it doesn't exist)
4. Import data into python (example):

from colony_roi_import import ColonyROIImportCSV

DATA_DIR = "data"

roi_import = ColonyROIImportCSV(DATA_DIR)

colony_count = roi_import.get_colony_count()
colony_roi_data = roi_import.colony_data()

Colony roi data is a python dictionary:

colony_data = {
	'colony_name': colony_name,
	'roi_data': roi_data,
    }

Where roi_data is a list of (x,y) pixel coordinates for the boundaries drawn around colonies by ImageJ.

Future

• Create another python file to extend labware functionality interface, e.g. plate.colonies('1')
• Improve agar_plate.py to use colony interface
• Use a machine learning algorithm to improve performance of colony detection