mr_series_selection

Small library for tag-based MRI series selection, spun out from the lumbar spine MR project.

Installation

Standard Installation

Do this in order to use the library, without needing to edit it:

pip install git+https://github.com/QTIM-Lab/mr-series-selection.git

It is recommended to install this in a virtual environment.

Editable Installation

This is will clone the code into your current location and then install it in your environment via a link. This means that you can edit the code and re-run without needing to reinstall.

git clone https://github.com/QTIM-Lab/mr-series-selection.git
cd mr_series_selection
pip install -e .

It is recommended to install this in a virtual environment.

Usage

Single Study Mode

To run on a study and print out detailed debug information to the console, just specify an input directory containing the files for a single study.

python -m mr_series_selection /path/to/some/study/directory

Note that you can run this from anywhere as long as the virtualenv in which the package is installed is active.

Batch Mode

To run on an entire batch of studies, each within a sub-directory of an input directory, use the -b option, pass the input directory containing the study directories, and specify a location for the output CSV with -o (default is series_selection_results.csv in the current working directory).

python -m mr_series_selection -b /path/to/some/directory -o /path/to/output.csv

Again note that you can run this from anywhere as long as the virtualenv in which the package is installed is active.

By default, batch mode will be run with 8 threads to increase throughput. The number of threads can be controlled with the -t option. Setting to 0 will use the main thread, which may be useful for debugging purposes.

How far down from the current directory it needs to look to find study directories can be controlled with the -r option. The default is 1, implying that each immediate sub-directory of the input directory are assumed to contain a study.

Additionally, which type of classification it should perform can be controlled with the -c option. Currently, it supports 'lumbar', 'prostate', and 'brain'. The default option is 'lumbar'.