Warning
Problem: The current name of your GitHub Pages repository ("Solution: Please consider renaming the repository to "
http://".
However, if the current repository name is intended, you can ignore this message by removing "{% include widgets/debug_repo_name.html %}" in index.html.
Action required
Problem: The current root path of this site is "baseurl ("_config.yml.
Solution: Please set the
baseurl in _config.yml to "Education
-
Utrecht UniversityMaster of Science, Methodology and Statistics2012 - 2016 -
Utrecht UniversityBachelor of Science, Psychology2016 - 2017
Experience
-
TNOScientist Innovator2022 - Present -
SanquinPHD Candidate2017 - 2022
Publications
Evaluating Cluster-Based Synthetic Data Generation for Blood-Transfusion Analysis
Shannon K. S. Kroes, Matthijs van Leeuwen, Rolf H. H. Groenwold, Mart P. Janssen
Journal of Cybersecurity and Privacy 2023
Cluster-based synthetic data generation (CBSDG) offers an explainable, privacy-preserving way to share sensitive data. We applied CBSDG to 250,729 real blood-transfusion records and trained SVMs to predict donor hemoglobin levels, matching original precision (male 0.997, female 0.987) and showing comparable recall and feature‐impact patterns. Most attributes became harder to infer, with only deferral status and sex remaining detectable—demonstrating CBSDG’s promise for practical use.
Evaluating Cluster-Based Synthetic Data Generation for Blood-Transfusion Analysis
Shannon K. S. Kroes, Matthijs van Leeuwen, Rolf H. H. Groenwold, Mart P. Janssen
Journal of Cybersecurity and Privacy 2023
Cluster-based synthetic data generation (CBSDG) offers an explainable, privacy-preserving way to share sensitive data. We applied CBSDG to 250,729 real blood-transfusion records and trained SVMs to predict donor hemoglobin levels, matching original precision (male 0.997, female 0.987) and showing comparable recall and feature‐impact patterns. Most attributes became harder to infer, with only deferral status and sex remaining detectable—demonstrating CBSDG’s promise for practical use.
Generating synthetic mixed discrete-continuous health records with mixed sum-product networks
Shannon K S Kroes, Matthijs van Leeuwen, Rolf H H Groenwold, Mart P Janssen
Journal of the American Medical Informatics Association (JAMIA) 2022
Mixed sum-product networks (MSPNs) instantiate private data representations from which synthetic patient records are drawn, enabling secure exchange for downstream statistical analyses. Rigorous evaluation against privacy and information-loss metrics demonstrates the approach’s capacity to uphold confidentiality while preserving analytical utility.
Generating synthetic mixed discrete-continuous health records with mixed sum-product networks
Shannon K S Kroes, Matthijs van Leeuwen, Rolf H H Groenwold, Mart P Janssen
Journal of the American Medical Informatics Association (JAMIA) 2022
Mixed sum-product networks (MSPNs) instantiate private data representations from which synthetic patient records are drawn, enabling secure exchange for downstream statistical analyses. Rigorous evaluation against privacy and information-loss metrics demonstrates the approach’s capacity to uphold confidentiality while preserving analytical utility.
Evaluating privacy of individuals in medical data
Shannon K S Kroes, Mart P. Janssen, Rolf H.H. Groenwold, Matthijs van Leeuwen
Health Informatics Journal 2021
A novel, variable-centric paradigm merges rigorous information-theoretic metrics with dynamic visualizations to elevate the understanding of individual privacy risk within medical datasets. By contextualizing each feature’s real-world exploitability, the methodology fulfills regulatory mandates and informs the creation of nuanced anonymization schemes, thereby illuminating the delicate balance between data utility and confidentiality.
Evaluating privacy of individuals in medical data
Shannon K S Kroes, Mart P. Janssen, Rolf H.H. Groenwold, Matthijs van Leeuwen
Health Informatics Journal 2021
A novel, variable-centric paradigm merges rigorous information-theoretic metrics with dynamic visualizations to elevate the understanding of individual privacy risk within medical datasets. By contextualizing each feature’s real-world exploitability, the methodology fulfills regulatory mandates and informs the creation of nuanced anonymization schemes, thereby illuminating the delicate balance between data utility and confidentiality.