SpecVis

SpecVis is a repository of R functions to visualize quantitative MRS results from different linear-combination algorithms.

Getting started

Prerequisites

SpecVis was developed in RStudio (Version 1.2.5019) on macOS Catalina (Version 10.15.3 (19D76)) and the needed libraries are downloaded and installed automatically.

For high-resolution pdf files you need to have a functioning cairo_pdf device. You can always change the ggsave output format to any other format.

Installation

Download the latest SpecVis code from its GitHub repository, then include the SpecVis folder as workdir (setwd()). Make sure to regularly check for updates, as we frequently commit new features, bug fixes, and improved functions.

Example markdown and script

An example markdown is included in the repository. You can adapt your own script based on this function.

Features

• Load LCM-native result files from Osprey (.csv), LCModel (.coord), and Tarquin (.csv).
• Load statistics .csv-files which include group variables and correlation measures
• Raincloud plots (https://wellcomeopenresearch.org/articles/4-63) with individual datapoints, boxplots, distributions, and mean +- SD representations.
• Boxplots with individual datapoints
• Correlation plots with collapsed-correlations, group-level correlations, and indicators for sub-groups.
• Correlation plots with group-level facets and correlations for sub-groups.
• Bland-Altman plots with distribution ellipse with collapsed-distribution and group-distributions.
• Statistics script which automatically performs appropriate statistics, including descriptive statistics, tests for normality, variance analysis, and post hoc tests.

Supported file formats

• Osprey .csv-files
• LCModel .coord-files
• Tarquin .csv-files
• arbitrary .csv-files

Example plots

Raincloud plot (https://wellcomeopenresearch.org/articles/4-63)

Correlation plot

Bland Altmann plot

Correlation facet plot

Planned features

• Integration of spectra visualization

Contact, Feedback, Suggestions

For any sort of questions, feedback, suggestions, or critique, please reach out to us via [email protected]. We also welcome your direct contributions to SpecVis here in the GitHub repository.

Developers

• Helge J. Zöllner
• Georg Oeltzschner

Should you publish material that made use of SpecVis, please cite the following publication:

HJ Zöllner, Michal Považan, SCN Hui, S Tapper, RAE Edden, Georg Oeltzschner, Agreement between different linear-combination modelling algorithms for short-TE proton spectra. bioRxiv 2020.

Should you publish material that made use of the Raincloud plot script, please additionally cite:

Allen M, Poggiali D, Whitaker K et al. Raincloud plots: a multi-platform tool for robust data visualization. Wellcome Open Res 2019, 4:63

Acknowledgements

We wish to thank Martin Wilson (University of Birmingham, Birmingham) for shared import code from the 'spant' R-package https://martin3141.github.io/spant/index.html. This code builds on modified version of the raincloud plots by Davide Poggiali https://github.com/RainCloudPlots/RainCloudPlots.

This work has been supported by NIH grants R01EB016089, P41EB15909, R01EB023963, and K99AG062230.