PURRlab @ IT University of Copenhagen

Pattern Recognition Revisited

About

PURRLab research interests lie within the broad area of trustworthy machine learning and its applications to medical imaging with a focus on datasets. We are particularly interested in understanding the similarity and diversity of datasets, methods for learning with limited labeled data such as transfer learning, and meta-research on machine learning in medical imaging.

The best way to get a sense of what’s currently going on in the lab is to read about our people and projects.

PURRLab is a part of DASYA research group in the department of Computer Science at the IT University of Copenhagen and is led by Veronika Cheplygina.

PURRLab research is being supported by the Dutch Research Council, Novo Nordisk Foundation and the Independent Research Council of Denmark.

News

May 02, 2024 Théo’s paper on citation practices was accepted at MIDL!
Mar 14, 2024 We have another new preprint to share, this time on model robustness in transfer learning.
Feb 15, 2024 Two new preprints are out, on citation practices and sharing datasets on community platforms, both about medical imaging datasets.
Jan 31, 2024 Several of us will be at the D3A conference, presenting posters about shortcuts, robustness in transfer learning, and citations of medical imaging datasets. Please stop by if you are interested in our work!
Dec 01, 2023 Yucheng Lu joins us as a postdoctoral researcher, welcome Yucheng!
Oct 09, 2023 Théo Sourget now joins us as a research assistant, welcome Théo!
Sep 29, 2023 Dovile’s paper was accepted at TMLR! Amelia will be presenting her work with MSc students Trine and Cathrine at the ICCV DataComp workshop in Paris next week. Come say hi!

Selected publications

  1. citation.png
    [Citation needed] Data usage and citation practices in medical imaging conferences
    Théo Sourget, Ahmet Akkoç , Stinna Winther , Christine Lyngbye Galsgaard , Amelia Jiménez-Sánchez, and 3 more authors
    arXiv preprint arXiv:2402.03003, 2024
  2. actionability.png
    Towards actionability for open medical imaging datasets: lessons from community-contributed platforms for data management and stewardship
    Amelia Jiménez-Sánchez, Natalia-Rozalia Avlona , Dovile JuodelyteThéo Sourget, Caroline Vang-Larsen , and 2 more authors
    arXiv preprint arXiv:2402.06353, 2024
  3. revisiting.png
    Revisiting Hidden Representations in Transfer Learning for Medical Imaging
    Transactions on Machine Learning Research, 2023
  4. augmenting.png
    Augmenting Chest X-ray Datasets with Non-Expert Annotations
    Cathrine Damgaard , Trine Naja Eriksen , Dovile JuodelyteVeronika Cheplygina, and Amelia Jiménez-Sánchez
    arXiv preprint arXiv:2309.02244, 2023
  5. kaggle.png
    Machine learning for medical imaging: methodological failures and recommendations for the future
    Gaël Varoquaux , and Veronika Cheplygina
    NPJ Digital Medicine, 2022
  6. enhance.png
    ENHANCE (ENriching Health data by ANnotations of Crowd and Experts): A case study for skin lesion classification
    Ralf Raumanns, Gerard Schouten , Max Joosten , Josien PW Pluim , and Veronika Cheplygina
    arXiv preprint arXiv:2107.12734, 2021