Deep-learning-based reconstruction of undersampled MRI to reduce scan times: a multicentre, retrospective, cohort study.

in The Lancet. Oncology by Aditya Rastogi, Gianluca Brugnara, Martha Foltyn-Dumitru, Mustafa Ahmed Mahmutoglu, Chandrakanth J Preetha, Erich Kobler, Irada Pflüger, Marianne Schell, Katerina Deike-Hofmann, Tobias Kessler, Martin J van den Bent, Ahmed Idbaih, Michael Platten, Alba A Brandes, Burt Nabors, Roger Stupp, Denise Bernhardt, Jürgen Debus, Amir Abdollahi, Thierry Gorlia, Jörg-Christian Tonn, Michael Weller, Klaus H Maier-Hein, Alexander Radbruch, Wolfgang Wick, Martin Bendszus, Hagen Meredig, Felix T Kurz, Philipp Vollmuth

TLDR

  • The study aimed to improve the efficiency of MRI-based clinical workflows by developing a deep learning algorithm that can reconstruct MRI images from undersampled data. The algorithm was trained on a large dataset of MRI images and was able to accurately reconstruct images with a 10-fold reduction in scan time. The study found that the algorithm was able to accurately measure the size and shape of tumors in the brain, which is important for diagnosing and treating brain tumors. The study's findings suggest that this algorithm could be used to improve the speed and accuracy of MRI-based diagnosis and treatment planning in brain tumors, which could lead to better patient outcomes.

Abstract

The extended acquisition times required for MRI limit its availability in resource-constrained settings. Consequently, accelerating MRI by undersampling k-space data, which is necessary to reconstruct an image, has been a long-standing but important challenge. We aimed to develop a deep convolutional neural network (dCNN) optimisation method for MRI reconstruction and to reduce scan times and evaluate its effect on image quality and accuracy of oncological imaging biomarkers. In this multicentre, retrospective, cohort study, MRI data from patients with glioblastoma treated at Heidelberg University Hospital (775 patients and 775 examinations) and from the phase 2 CORE trial (260 patients, 1083 examinations, and 58 institutions) and the phase 3 CENTRIC trial (505 patients, 3147 examinations, and 139 institutions) were used to develop, train, and test dCNN for reconstructing MRI from highly undersampled single-coil k-space data with various acceleration rates (R=2, 4, 6, 8, 10, and 15). Independent testing was performed with MRIs from the phase 2/3 EORTC-26101 trial (528 patients with glioblastoma, 1974 examinations, and 32 institutions). The similarity between undersampled dCNN-reconstructed and original MRIs was quantified with various image quality metrics, including structural similarity index measure (SSIM) and the accuracy of undersampled dCNN-reconstructed MRI on downstream radiological assessment of imaging biomarkers in oncology (automated artificial intelligence-based quantification of tumour burden and treatment response) was performed in the EORTC-26101 test dataset. The public NYU Langone Health fastMRI brain test dataset (558 patients and 558 examinations) was used to validate the generalisability and robustness of the dCNN for reconstructing MRIs from available multi-coil (parallel imaging) k-space data. In the EORTC-26101 test dataset, the median SSIM of undersampled dCNN-reconstructed MRI ranged from 0·88 to 0·99 across different acceleration rates, with 0·92 (95% CI 0·92-0·93) for 10-times acceleration (R=10). The 10-times undersampled dCNN-reconstructed MRI yielded excellent agreement with original MRI when assessing volumes of contrast-enhancing tumour (median DICE for spatial agreement of 0·89 [95% CI 0·88 to 0·89]; median volume difference of 0·01 cm[95% CI 0·00 to 0·03] equalling 0·21%; p=0·0036 for equivalence) or non-enhancing tumour or oedema (median DICE of 0·94 [95% CI 0·94 to 0·95]; median volume difference of -0·79 cm[95% CI -0·87 to -0·72] equalling -1·77%; p=0·023 for equivalence) in the EORTC-26101 test dataset. Automated volumetric tumour response assessment in the EORTC-26101 test dataset yielded an identical median time to progression of 4·27 months (95% CI 4·14 to 4·57) when using 10-times-undersampled dCNN-reconstructed or original MRI (log-rank p=0·80) and agreement in the time to progression in 374 (95·2%) of 393 patients with data. The dCNN generalised well to the fastMRI brain dataset, with significant improvements in the median SSIM when using multi-coil compared with single-coil k-space data (p<0·0001). Deep-learning-based reconstruction of undersampled MRI allows for a substantial reduction of scan times, with a 10-times acceleration demonstrating excellent image quality while preserving the accuracy of derived imaging biomarkers for the assessment of oncological treatment response. Our developments are available as open source software and hold considerable promise for increasing the accessibility to MRI, pending further prospective validation. Deutsche Forschungsgemeinschaft (German Research Foundation) and an Else Kröner Clinician Scientist Endowed Professorship by the Else Kröner Fresenius Foundation.

Overview

  • The study aimed to develop a deep convolutional neural network (dCNN) optimization method for MRI reconstruction and reduce scan times while evaluating its effect on image quality and accuracy of oncological imaging biomarkers. The study used MRI data from patients with glioblastoma treated at Heidelberg University Hospital and from the phase 2 CORE and phase 3 CENTRIC trials, as well as the phase 2/3 EORTC-26101 trial. The dCNN was trained and tested to reconstruct MRI from highly undersampled single-coil k-space data with various acceleration rates. Independent testing was performed with MRIs from the phase 2/3 EORTC-26101 trial, and the public NYU Langone Health fastMRI brain test dataset was used to validate the generalisability and robustness of the dCNN. The study evaluated the similarity between undersampled dCNN-reconstructed and original MRIs using various image quality metrics and assessed the accuracy of undersampled dCNN-reconstructed MRI on downstream radiological assessment of imaging biomarkers in oncology. The dCNN generalised well to the fastMRI brain dataset, with significant improvements in the median SSIM when using multi-coil compared with single-coil k-space data. The study demonstrated that deep-learning-based reconstruction of undersampled MRI allows for a substantial reduction of scan times, with a 10-times acceleration demonstrating excellent image quality while preserving the accuracy of derived imaging biomarkers for the assessment of oncological treatment response. The study's findings hold considerable promise for increasing the accessibility to MRI, pending further prospective validation. The study was funded by the Deutsche Forschungsgemeinschaft (German Research Foundation) and an Else Kröner Clinician Scientist Endowed Professorship by the Else Kröner Fresenius Foundation. The dCNN is available as open source software.

Comparative Analysis & Findings

  • The study compared the outcomes observed under different experimental conditions or interventions, specifically the effect of accelerating MRI by undersampling k-space data on image quality and accuracy of oncological imaging biomarkers. The study found that the dCNN optimized for MRI reconstruction was able to reconstruct MRI from highly undersampled single-coil k-space data with various acceleration rates, with a 10-times acceleration demonstrating excellent image quality while preserving the accuracy of derived imaging biomarkers for the assessment of oncological treatment response. The study also found that the dCNN generalised well to the fastMRI brain dataset, with significant improvements in the median SSIM when using multi-coil compared with single-coil k-space data. The study's findings suggest that deep-learning-based reconstruction of undersampled MRI is a promising approach for increasing the accessibility to MRI and improving the efficiency of MRI-based clinical workflows.

Implications and Future Directions

  • The study's findings have significant implications for the field of research or clinical practice, as they demonstrate the potential of deep-learning-based reconstruction of undersampled MRI to improve the efficiency and accessibility of MRI-based clinical workflows. The study's findings also suggest that the dCNN optimized for MRI reconstruction could be used as a tool for improving the accuracy and speed of MRI-based diagnosis and treatment planning in oncology. Future research directions could include further validation of the dCNN's performance in clinical settings, exploration of the dCNN's potential for improving the accuracy and speed of MRI-based diagnosis and treatment planning in other clinical applications, and development of novel deep-learning-based approaches for MRI reconstruction that can further improve image quality and accuracy while reducing scan times.
Read Full Article