Conventional and Deep-Learning-Based Image Reconstructions of Undersampled K-Space Data of the Lumbar Spine Using Compressed Sensing in MRI: A Comparative Study on 20 Subjects

被引:16
|
作者
Fervers, Philipp [1 ,2 ]
Zaeske, Charlotte [1 ,2 ]
Rauen, Philip [1 ,2 ]
Iuga, Andra-Iza [1 ,2 ]
Kottlors, Jonathan [1 ,2 ]
Persigehl, Thorsten [1 ,2 ]
Sonnabend, Kristina [3 ]
Weiss, Kilian [3 ]
Bratke, Grischa [1 ,2 ]
机构
[1] Univ Hosp Cologne, Inst Diagnost & Intervent Radiol, D-50937 Cologne, Germany
[2] Inst Diagnost & Intervent Radiol, Fac Med, D-50937 Cologne, Germany
[3] Philips GmbH Market DACH, D-22335 Hamburg, Germany
关键词
magnetic resonance imaging; artificial intelligence; image processing; computer-assisted; BACK-PAIN INDICATIONS; ECHO SEQUENCE;
D O I
10.3390/diagnostics13030418
中图分类号
R5 [内科学];
学科分类号
1002 ; 100201 ;
摘要
Compressed sensing accelerates magnetic resonance imaging (MRI) acquisition by undersampling of the k-space. Yet, excessive undersampling impairs image quality when using conventional reconstruction techniques. Deep-learning-based reconstruction methods might allow for stronger undersampling and thus faster MRI scans without loss of crucial image quality. We compared imaging approaches using parallel imaging (SENSE), a combination of parallel imaging and compressed sensing (COMPRESSED SENSE, CS), and a combination of CS and a deep-learning-based reconstruction (CS AI) on raw k-space data acquired at different undersampling factors. 3D T2-weighted images of the lumbar spine were obtained from 20 volunteers, including a 3D sequence (standard SENSE), as provided by the manufacturer, as well as accelerated 3D sequences (undersampling factors 4.5, 8, and 11) reconstructed with CS and CS AI. Subjective rating was performed using a 5-point Likert scale to evaluate anatomical structures and overall image impression. Objective rating was performed using apparent signal-to-noise and contrast-to-noise ratio (aSNR and aCNR) as well as root mean square error (RMSE) and structural-similarity index (SSIM). The CS AI 4.5 sequence was subjectively rated better than the standard in several categories and deep-learning-based reconstructions were subjectively rated better than conventional reconstructions in several categories for acceleration factors 8 and 11. In the objective rating, only aSNR of the bone showed a significant tendency towards better results of the deep-learning-based reconstructions. We conclude that CS in combination with deep-learning-based image reconstruction allows for stronger undersampling of k-space data without loss of image quality, and thus has potential for further scan time reduction.
引用
收藏
页数:13
相关论文
共 25 条
  • [1] Compressed sensing using a deep adaptive perceptual generative adversarial network for MRI reconstruction from undersampled K-space data
    Wu, Kun
    Xia, Yan
    Ravikumar, Nishant
    Frangi, Alejandro F.
    BIOMEDICAL SIGNAL PROCESSING AND CONTROL, 2024, 96
  • [2] Compressed Sensing MRI Reconstruction from Highly Undersampled k-Space Data Using Nonsubsampled Shearlet Transform Sparsity Prior
    Yuan, Min
    Yang, Bingxin
    Ma, Yide
    Zhang, Jiuwen
    Zhang, Runpu
    Zhang, Caiyuan
    MATHEMATICAL PROBLEMS IN ENGINEERING, 2015, 2015
  • [3] Emerging Trends in Fast MRI Using Deep-Learning Reconstruction on Undersampled k-Space Data: A Systematic Review
    Singh, Dilbag
    Monga, Anmol
    de Moura, Hector L.
    Zhang, Xiaoxia
    Zibetti, Marcelo V. W.
    Regatte, Ravinder R.
    BIOENGINEERING-BASEL, 2023, 10 (09):
  • [4] Paired Dictionary Learning Based MR Image Reconstruction from Undersampled k-Space Data
    Liu, Jiaodi
    Sheng, Yuxia
    Yang, Jun
    Xiong, Dan
    2021 PROCEEDINGS OF THE 40TH CHINESE CONTROL CONFERENCE (CCC), 2021, : 2981 - 2986
  • [5] Model-based federated learning for accurate MR image reconstruction from undersampled k-space data
    Wu, Ruoyou
    Li, Cheng
    Zou, Juan
    Liang, Yong
    Wang, Shanshan
    Computers in Biology and Medicine, 2024, 180
  • [6] Radial Undersampled MRI Reconstruction Using Deep Learning With Mutual Constraints Between Real and Imaginary Components of K-Space
    Li, Zhaotong
    Li, Sha
    Zhang, Zeru
    Wang, Fei
    Wu, Fengliang
    Gao, Song
    IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS, 2024, 28 (06) : 3583 - 3596
  • [7] SISMIK for Brain MRI: Deep-Learning-Based Motion Estimation and Model-Based Motion Correction in k-Space
    Dabrowski, Oscar
    Falcone, Jean-Luc
    Klauser, Antoine
    Songeon, Julien
    Kocher, Michel
    Chopard, Bastien
    Lazeyras, Francois
    Courvoisier, Sebastien
    IEEE TRANSACTIONS ON MEDICAL IMAGING, 2025, 44 (01) : 396 - 408
  • [8] Deep learning-based image reconstruction and motion estimation from undersampled radial k-space for real-time MRI-guided radiotherapy
    Terpstra, Maarten L.
    Maspero, Matteo
    d'Agata, Federico
    Stemkens, Bjorn
    Intven, Martijn P. W.
    Lagendijk, Jan J. W.
    van den Berg, Cornelis A. T.
    Tijssen, Rob H. N.
    PHYSICS IN MEDICINE AND BIOLOGY, 2020, 65 (15):
  • [9] Hybrid deep-learning-based denoising method for compressed sensing in pituitary MRI: comparison with the conventional wavelet-based denoising method
    Hiroyuki Uetani
    Takeshi Nakaura
    Mika Kitajima
    Kosuke Morita
    Kentaro Haraoka
    Naoki Shinojima
    Machiko Tateishi
    Taihei Inoue
    Akira Sasao
    Akitake Mukasa
    Minako Azuma
    Osamu Ikeda
    Yasuyuki Yamashita
    Toshinori Hirai
    European Radiology, 2022, 32 : 4527 - 4536
  • [10] Hybrid deep-learning-based denoising method for compressed sensing in pituitary MRI: comparison with the conventional wavelet-based denoising method
    Uetani, Hiroyuki
    Nakaura, Takeshi
    Kitajima, Mika
    Morita, Kosuke
    Haraoka, Kentaro
    Shinojima, Naoki
    Tateishi, Machiko
    Inoue, Taihei
    Sasao, Akira
    Mukasa, Akitake
    Azuma, Minako
    Ikeda, Osamu
    Yamashita, Yasuyuki
    Hirai, Toshinori
    EUROPEAN RADIOLOGY, 2022, 32 (07) : 4527 - 4536