Multimodal data integration and label scarcity pose significant challenges for machine learning in medical settings. To address these issues, we conduct an in-depth evaluation of the newly proposed Multimodal Variational Mixture-of-Experts (MMVM) VAE on the challenging MIMIC-CXR dataset. Our analysis demonstrates that the MMVM VAE consistently outperforms other multimodal VAEs and fully supervised approaches, highlighting its strong potential for real-world medical applications.


    Access

    Download


    Export, share and cite



    Title :

    Weakly-Supervised Multimodal Learning on MIMIC-CXR


    Contributors:

    Publication date :

    2024


    Remarks:

    Findings paper presented at Machine Learning for Health (ML4H) symposium 2024, December 15-16, 2024, Vancouver, Canada, 13 pages. arXiv admin note: text overlap with arXiv:2403.05300


    Type of media :

    Preprint


    Type of material :

    Electronic Resource


    Language :

    English




    Weakly supervised learning of actions from transcripts

    Kuehne, Hilde / Richard, Alexander / Gall, Juergen | British Library Online Contents | 2017


    Weakly Supervised Localization and Learning with Generic Knowledge

    Deselaers, T. / Alexe, B. / Ferrari, V. | British Library Online Contents | 2012


    Weakly-Supervised Crack Detection

    Inoue, Yuki / Nagayoshi, Hiroto | IEEE | 2023


    Learning Cascaded Detection Tasks with Weakly-Supervised Domain Adaptation

    Hanselmann, Niklas / Schneider, Nick / Ortelt, Benedikt et al. | IEEE | 2021


    Weakly-Supervised Cross-Domain Dictionary Learning for Visual Recognition

    Zhu, F. / Shao, L. | British Library Online Contents | 2014