| Datensatz noch nicht abgeschlossen | |
| Zugriffsnummer | 56091 |
| Dokumenttyp | Konferenzartikel |
| Sprache | Englisch |
| Titel | From simulations to actual data - generalizability and robustness of learned image reconstruction for portable low-field MRI |
| Autor(in); Institution |
Schote, David; 8.1, Biomedizinische Magnetresonanz, PTB-Berlin
Herthum, Helge; Berlin Center for Advanced Neuroimaging, Charité – Universitätsmedizin Berlin, GERMANY
Calatroni, Luca; MaLGa Center, DIBRIS, Universit‘a di Genova, ITALY
Papafitsoros, Kostas; School of Mathematical Sciences, Queen Mary University of London, London, UK
|
| Quelle/Jahr | Proceedings of the International Society for Magnetic Resonance in Medicine:(2026) |
| Availability | [online only] |
| ISSN | 1065-9889 (online) |
| URL | |
| Verlag | ISMRM |
| Konferenzangaben | Annual Meeting of the ISMRM, Capetown, 9-14, Mai, 2026, South Africa |
| Freie Schlagworte | image reconstruction ; ultra-low-field MRI ; portable MRI ; neural network AI |
| Zusammenfassung | Reconstruction models often rely on large simulated datasets for training, but discrepancies between simulated and target data can lead to distribution shifts and degraded performance. Goals: To assess the generalizability and robustness to data-distribution shifts of three different learning-based methods for low-field MRI. Approach: We pre-trained a model-based deep learning method and two recently proposed sparsity-based approaches with learned spatially adaptive regularization parameter maps on the fastMRI brain data and subsequently applied them to different in-vivo and phantom low-field (50 mT) MR data. Results: All methods showed good generalization properties, with the sparsity-based methods showing improved robustness to data-distribution shifts compared to MoDL |
Zitierung
Schote, D., Herthum, H., Kolbitsch, C., Calatroni, L., Papafitsoros, K., & Kofler, A. (2026). From simulations to actual data - generalizability and robustness of learned image reconstruction for portable low-field MRI. Annual Meeting of the ISMRM, Capetown, 9-14, Mai, 2026, South Africa.