Neuron Finding

This can best be described as object-finding or image segmentation, where the goal is to design a model whose output is the coordinates to regions of interest in an image. There’s no discrete label; rather, the model needs to learn segments in a continuous two-dimensional plane; relevant information to learning these segments, however, may be strewn over a third dimension of time. This makes for a very high-dimensional, large-scale problem: the data are height-by-width-by-time, and the model needs to learn a height-by-width mapping of pixels, where each pixel is either part of a neuron, or isn’t. Each folder of training and testing images is a single plane, and the images are numbered according to their temporal ordering. The neurons in the images will “flicker” on and off, as calcium (Ca 2+ ) is added, activating the action potential gates. You’ll have to use this information in order to locate the neurons and segment them out from the surrounding image.

Approach

• NMF to get neuron region coordinates using Thunder-Extraction

NMF Flow

• Use thunder library and import that in your code.
• Load the testing dataset.
• Create the algorithm with various parameters.
• Fit the model in our algorithm.
• Transform and merge the overlapping coordinates.
• Save the output in desired format

Data

The datasets we used to train and test is provided by Dr. Shannon Quinn for the course CSCI 8360: Data Science Practicum.

There are total 9 datasets(test) which are being evaluated.

Prerequisites

The project requires the following technologies to be installed.

• Instructions to download and install Python can be found here. https://www.python.org/
• After the python is installed, the thunder package can be installed using the following commands in the command prompt/terminal.
  • pip install thunder-extraction
  • pip install thunder-python

How to Run

• Download the repository

• Run the script main_script.py using the following command.

$ python3 main_script.py

This script will:

• Download the data in data/download folder
• It will extract the zip file in data/test or data/test depending on the type of file.
• It will run the model
• The result will be saved in the submission file
• finally it will remove the data from the folder

Results

k	percentile	max_iter	overlap	score
10	98	50	0.1	3.00907
10	98	100	0.1	2.98822
10	97	50	0.1	2.96129
20	99	100	0.2	2.95583

Contribution

• See Contributors file for more details

License

This Project is under the MIT License. For more details visit License file.