HypoQuantyl

Image analysis tool used for tracking localized growth of hypocotyls using high-throughput cloud computing. This program uses a variety of machine learning algorithms to identify a hypocotyl, extract the growth region, form a midline, then quantify the elemental growth rate in which cellular elongation occurs at many locations along that midline.

Note

Used BFG to remove all objects > 500K in size [07-09-2019]

Summary [re-iterate above description and expand on use, novelty, and similar software]

A) What is this program for?
Learning algorithms: predict contour, predict ending of midline
Image analysis: select points along seedling
Quantification: Localized measurements of midline, model growth of growing seedlings
Miscellaneous features: late-decision segmentation for error checking, ease of loading data
Output Data: models full developmental process of a growing, de-etiolating hypocotyl

B) How does this program work?
User stores time-lapse image stacks of a single type in individual folders
A full experiment is a set of individual image stacks in a parent folder
Multiple experiments can be loaded as well

C) Instructions for using HypoQuantyl can be found in HOWTO.md
- Sample data can be found in Sample Data

Getting Started

Dependencies

• Lowest usable MATLAB version
• Run scripts for toolbox dependencies (like from QuantDRaCALA)
• Maybe something about CyVerse if necessary?

Installation

• (to-do) Install program as Matlab executable
  
• (to-do) Program runs on CyVerse
  

Version Information

02/05/2018 - HypoQuantyl Version 0.1 [ date created ] 02/06/2018 - HypoQuantyl Version 0.1 [ name chosen ] 02/22/2018 - HypoQuantyl Version 0.3 [ class structure finalized ] 05/24/2018 - HypoQuantyl Version 0.4 [ midpoint-normalization method ] 08/02/2018 - HypoQuantyl Version 0.5 [ routes method –> curves method ] 12/05/2018 - HypoQuantyl Version 0.6 [ optimized image read/write ] 02/18/2019 - HypoQuantyl Version 0.7 [ learning-based segmentation framework ]

Authors

Main Author

Julian Bustamante, Cellular and Molecular Biology Program (jbustamante@wisc.edu)
University of Wisconsin - Madison
Department of Botany

Contributing Author

Nathan Miller, Senior Scientist (ndmill@gmail.com)
University of Wisconsin - Madison
Department of Botany

License

MIT license found in LICENSE

Acknowledgements

• Guosheng Wu
  
• Nathan Miller
  

To-Do:

• Segmentation with Convolution Neural Net (CNN)
  
  • CNN Architecture
    
  • Feature analysis
    
• Junction Finder with CNN
  
  • Click Method
    
  • Slice Method
    
• Midline Region Splitter and Feature Identifier
  
  • Track patches between frames
    
• Kinematics Analysis
  
  • Elemental Growth Rate methods
    
• Validation Methods
  
  • Manual methodology
    
• Deployment to CyVerse
  
  • Compile to use with MCR
    
  • Synchronization with Open Science Grid
    
• True Application Testing
  
  • CRY1 mutants
    
  • Cell Wall mutants
    

Class Structure and Main Functions

Main Classes

| Class | Description | | — | — | | HypoQuantyl | Initializes program and loads Experiment folders | | Experiment | Loads folders containing multiple image stacks | | Genotype | Loads image stacks containing growing seedlings | | Seedling | Represents a single seedling throughout a time-lapse | | Hypocotyl | Represents the hypocotyl portion of a single seedlings | | CircuitJB | Main contour of a hypocotyl, segmented by anchor points | | Route | Individual segments around a hypocotyl’s contour |

Algorithms

| Algorithm | Description | | — | — | | FindHypocotyl | Identifies region on a seedling containing the hypocotyl | | TrackHypocotyl | Keeps track of hypocotyl during de-etiolation process | | FindGoodFrames | Runs error-checking of each frame for each Seedling | | IntegrationAlgorithm | Integration method for measuring growing Hypocotyl |

Implementing to Cyber Infrastructure

| Infrastructure | Description | | — | — | | Create accounts | University Network, CyVerse, HTCondor | | Set up submit files | Holds MATLAB MCR, HypoQuantyl, Data, I-Commands | | Run Test Data | Run HypoQuantyl on Sample data on cloud computing environment |