Code – Predictive Clinical Neuroscience Lab

Code for different projects is hosted on GitHub, here: http://github.com/amarquand or here: http://github.com/predictive-clinical-neuroscience. All code is released under a GNU public license unless otherwise specified and comes without any warranty whatsoever. Some of our main contributions are listed below. If you use this software please cite some of the main references below.

PCNtoolkit

A comprehensive set of python tools for normative modelling, including warped Bayesian linear regression, hierarchical Bayesian regression, tools for modelling non-Gaussian distributions and federated learning. The latest version can can be installed via:

pip install pcntoolkit

The code repository can be found at: https://github.com/amarquand/PCNtoolkit

With extensive online documentation and tutorials.

PCNportal

A code free portal providing access to many pre-estimated normative models. This can be accessed via:

https://pcnportal.dccn.nl

The underlying code is available at: https://github.com/predictive-clinical-neuroscience/PCNportal

SACCADE (multi-view CCA)

Sparse Asymmetric Canonical Correlation Analysis (SACCADE). An implementation of multi-view sparse CCA based loosely on Daniela Witten’s penalised matrix decomposition. This has several enhancements useful for neuroimaging, as described here. Specifically, this enables uninteresting correlations to be ignored (e.g. between imaging modalities) and also accommodates block-wise missing data. This is available at:

https://github.com/predictive-clinical-neuroscience/saccade

Analysis methods for digital phenotyping

A set of tools for the analysis of smartphone-based digital phenotyping. For example, feature construction approaches based on hidden Markov models are available at:

https://github.com/predictive-clinical-neuroscience/HMM_Digital_Phenotyping

Another key challenge in digital phenotyping is missing data, which can be imputed using Posson process models. Code for doing this is available at:

https://github.com/predictive-clinical-neuroscience/Imputation_Digital_Phenotyping