eprintid: 10196662 rev_number: 6 eprint_status: archive userid: 699 dir: disk0/10/19/66/62 datestamp: 2024-09-06 13:09:59 lastmod: 2024-09-06 13:09:59 status_changed: 2024-09-06 13:09:59 type: article metadata_visibility: show sword_depositor: 699 creators_name: Jaubert, Olivier creators_name: Pascale, Michele creators_name: Montalt-Tordera, Javier creators_name: Akesson, Julius creators_name: Virsinskaite, Ruta creators_name: Knight, Daniel creators_name: Arridge, Simon creators_name: Steeden, Jennifer creators_name: Muthurangu, Vivek title: Training deep learning based dynamic MR image reconstruction using open-source natural videos ispublished: pub divisions: UCL divisions: B02 divisions: B04 divisions: D14 divisions: F48 divisions: GA1 keywords: Real-time, Dynamic MRI, Deep learning, Image reconstruction, Machine learning note: © 2024 Springer Nature Limited. This article is licensed under a Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/). abstract: To develop and assess a deep learning (DL) pipeline to learn dynamic MR image reconstruction from publicly available natural videos (Inter4K). Learning was performed for a range of DL architectures (VarNet, 3D UNet, FastDVDNet) and corresponding sampling patterns (Cartesian, radial, spiral) either from true multi-coil cardiac MR data (N = 692) or from synthetic MR data simulated from Inter4K natural videos (N = 588). Real-time undersampled dynamic MR images were reconstructed using DL networks trained with cardiac data and natural videos, and compressed sensing (CS). Differences were assessed in simulations (N = 104 datasets) in terms of MSE, PSNR, and SSIM and prospectively for cardiac cine (short axis, four chambers, N = 20) and speech cine (N = 10) data in terms of subjective image quality ranking, SNR and Edge sharpness. Friedman Chi Square tests with post-hoc Nemenyi analysis were performed to assess statistical significance. In simulated data, DL networks trained with cardiac data outperformed DL networks trained with natural videos, both of which outperformed CS (p < 0.05). However, in prospective experiments DL reconstructions using both training datasets were ranked similarly (and higher than CS) and presented no statistical differences in SNR and Edge Sharpness for most conditions.The developed pipeline enabled learning dynamic MR reconstruction from natural videos preserving DL reconstruction advantages such as high quality fast and ultra-fast reconstructions while overcoming some limitations (data scarcity or sharing). The natural video dataset, code and pre-trained networks are made readily available on github. date: 2024-05-23 date_type: published publisher: NATURE PORTFOLIO official_url: http://dx.doi.org/10.1038/s41598-024-62294-7 oa_status: green full_text_type: pub language: eng primo: open primo_central: open_green verified: verified_manual elements_id: 2279095 doi: 10.1038/s41598-024-62294-7 medium: Electronic pii: 10.1038/s41598-024-62294-7 lyricists_name: Muthurangu, Vivek lyricists_name: Arridge, Simon lyricists_name: Steeden, Jennifer lyricists_id: VMUTH99 lyricists_id: SRARR14 lyricists_id: JAEDG41 actors_name: Muthurangu, Vivek actors_id: VMUTH99 actors_role: owner funding_acknowledgements: MR/S032290/1 [UK Research and Innovation]; RG2661/17/20 [Heart Research UK]; NH/18/1/33511 [British Heart Foundation]; PG/17/6/32797 [British Heart Foundation] full_text_status: public publication: Scientific Reports volume: 14 number: 1 article_number: 11774 pages: 10 event_location: England citation: Jaubert, Olivier; Pascale, Michele; Montalt-Tordera, Javier; Akesson, Julius; Virsinskaite, Ruta; Knight, Daniel; Arridge, Simon; ... Muthurangu, Vivek; + view all <#> Jaubert, Olivier; Pascale, Michele; Montalt-Tordera, Javier; Akesson, Julius; Virsinskaite, Ruta; Knight, Daniel; Arridge, Simon; Steeden, Jennifer; Muthurangu, Vivek; - view fewer <#> (2024) Training deep learning based dynamic MR image reconstruction using open-source natural videos. Scientific Reports , 14 (1) , Article 11774. 10.1038/s41598-024-62294-7 <https://doi.org/10.1038/s41598-024-62294-7>. Green open access document_url: https://discovery.ucl.ac.uk/id/eprint/10196662/1/Training%20deep%20learning%20based%20dynamic%20MR%20image%20reconstruction%20using%20open-source%20natural%20videos.pdf