Title: Using knee-trained Deep Neural Networks for Brain MRIs, Understanding Domain Shift in Learned Magnetic Resonance Imaging (MRI) Reconstruction: A Quantitative Analysis on fastMRI Knee and Neuro Sequences
Authors: He, Shizhe, TUMKolleg, München
Contributors: Editor: Physikalisch-Technische Bundesanstalt (PTB), ISNI: 0000 0001 2186 1887
HostingInstitution: Physikalisch-Technische Bundesanstalt (PTB), ISNI: 0000 0001 2186 1887
Pages:18
Language:en
DOI:10.7795/320.202401
Resource Type: Text / Article
Publisher: Physikalisch-Technische Bundesanstalt (PTB)
Rights: Download for personal/private use only, if your national copyright law allows this kind of use.
Dates: Available: 2024-01-29
Accepted: 2023-05-09
Submitted: 2023-02-12
File: Download File (application/pdf) 5.44 MB (5699026 Bytes)
MD5 Checksum: 2eac6c9a6b411e0979b44610115b8849
SHA256 Checksum: 5ef4c143437e110cf7c9fbdddd5e88bf4a0e1c883685b139d92d28c4fdbfb3e9
Keywords Deep Learning ; fastMRI ; Domain Shift ; Physics-based Reconstruction ; Statistical Analysis ; Knee Sequences ; Neuro Sequences
Abstract: We investigate the problem of domain shift in the context of state-of-the-art MRI reconstruction networks with respect to variations in training data. We provide visualization tools and support our findings with statistical analysis for the networks evaluated on the 1.5 T/ 3 T fastMRI knee/neuro data. We observe that the signal-to-noise ratio of the examined sequences plays an essential role, and we statistically prove the hypothesis that both the type and amount of training data are less important for low acceleration factors.
Series Information: Junge Wissenschaft. Paper 01/2024
Other: In der Jungen Wissenschaft werden Forschungsarbeiten von Schüler/innen, die selbstständig, z.B. in einer Schule oder einem Schülerforschungszentrum, durchgeführt wurden, veröffentlicht.
Citation: He, Shizhe. Using knee-trained Deep Neural Networks for Brain MRIs. Understanding Domain Shift in Learned Magnetic Resonance Imaging (MRI) Reconstruction: A Quantitative Analysis on fastMRI Knee and Neuro Sequences. Physikalisch-Technische Bundesanstalt (PTB), 2024. Verfügbar unter: https://doi.org/10.7795/320.202401