bilgelm/bayesian-ps-adni repository overview

This docker image contains code for running the analyses described in our paper, "Predicting time to dementia using a quantitative template of disease progression".

You will need to download the dataset from adni.loni.usc.edu: Go to DATA & SAMPLES -> ACCESS DATA AND SAMPLES, and login After logging in, DOWNLOAD -> Study Data Search for "QT-PAD Challenge" and download the zip file via the link unzip the file and put the entire "ADNI_QT-PAD" directory into a directory named "data" under $FULLPATH.

Once you have this dataset and have pulled this docker image, use the following command to run it:

docker run -p 8888:8888 \
           -v $FULLPATH/results:/home/jovyan/work/results \
           -v $FULLPATH/data:/home/jovyan/work/data:ro \
           bilgelm/bayesian-ps-adni:v1.1

After running this command, you'll see some text like this:

Copy & paste this URL into your web browser to open up jupyter. Navigate to the directory called "work", then to "code".

The first code you need to run is adni_9cog_psmodel.ipynb. The cells need to be run in order. This code will define and fit a series of Bayesian models. You have the option of actually running the fitting (which takes quite a long time), or simply loading the results of the Monte Carlo sampler that are included in the docker image. By default, the code will load in the results and not actually run the sampling. To run the samplers, set the run_in_load_mode to False for each model you want to carry out the fitting for. This notebook will generate several output files that will be saved to the results/ directory.

You can run the remaining files in any order. boxplots_by_dx.ipynb, fitted_trajectory_with_scatter_static.ipynb, and time_to_dx_plot_v2.ipynb will generate figures. compute_predict_demonset_lm.ipynb will carry out linear models for predicting time to dementia onset.