Readme

Research context

This repository is the result of my (Jef Masereel) master's thesis under guidance of the STADIUS research team (Faculty of Engineering Sciences, Department of Biomedical Engineering, KU Leuven, Belgium). The data under analysis was provided by the MOX research team at the Department of Rehabilitation Sciences (also KU Leuven).

The goal of this study was to explore a broad range of social stress features in noninvasive recordings of blood volume pulse, electrodermal activity and respiration, and to assess their role in social stress regulation under influence of oxytocin treatment.

The dataset remains intellectual property of the university and is not shared in this repository. For more information and news, see the webpage of the MOX study, which is part of broader research on autism at LAuRes. A more in-depth discussion of prior works can be found in the introductory chapter of report.pdf

Overview of contents

• code
  • run init_process.m to build the relevant header files from source data
  • for each signal modality, follow the readme.txt to use the scripts
    1. BVP: blood volume pulse
    2. EDA: electrodermal activity
    3. RSP: respiration (from thoracic breathing belt)
  • combi
    • run merge_logdata.m to collect the extracted features
    • stattests_mat contains all statistical testing scripts
    • visuals_py contains jupyter notebooks with visualization tools
    • stattests_R contains explorative multivariate tests
    • export_*.m scripts can be used to export data to .xlsx
• data
  • ORIGINAL
    • rawdata_signals: formatted .txt files with raw measurements
    • rawdata_placebo.txt: placebo labels (columnwise)
    • rawdata_attachment.txt: SAAM & placebo labels (columnwise)
  • signal modalities
    • BVP, EDA, RSP: intermediate results from corresponding scripts
  • combi
    • intermediate and final results from corresponding scripts
  • header files
    • generated by init_process.m (if dataset was formatted correctly)
    • logdata_all.mat is generated by merge_logdata.m
• report
  • manuscript in .pdf as submitted for my master's thesis
  • contains the relevant citations, discussions and results

Technical remarks and recommendations

Scripts used for the collection, processing and analysis of the multimodal dataset were primarily implemented in MATLAB. A small number of operations is complemented with Python (v3.7) and R scripts to leverage existing tools.

A "devalued" dataset is provided in the repository to respect the IP rights of the MOX team, and to limit the filesize of the repository. The majority of functions remain operational, but the generated results are non-informational. The scripts can be adapted to fit a new dataset, or used directly if the data matches the format shown in ~/data/ORIGINAL/rawdata_signals/*. Additionally, the ~/data/ORIGINAL/rawdata_attachment.txt and/or ~/data/ORIGINAL/rawdata_placebo.txt should be added to include the relevant label information for statistical tests.

When applying (parts of) this repository to a new dataset, it is recommended to use Data Version Control to manage the results without causing excessive overhead for the basic Git version control.