Share：
Share this content in WeChat
X
[Chinese] [PDF] 25 1
• Technical Article •
Application value of deep learning reconstruction algorithm in accelerated lumbar spine MRI: A clinical study on image quality and efficiency optimization
HE Xiaoqun  LIU Meishan  ZHANG Zhiwei  LUO Xi  ZHANG Xingyue  YU Bin  PAN Boyang  GUO Ziheng  GONG Nanjie 

Cite this article as HE X Q, LIU M S, ZHANG Z W, et al. Application value of deep learning reconstruction algorithm in accelerated lumbar spine MRI: A clinical study on image quality and efficiency optimization[J]. Chin J Magn Reson Imaging, 2026, 17(6): 125-132. DOI:10.12015/issn.1674-8034.2026.06.016.


[Abstract] Objective To investigate the clinical value of accelerated scanning (AS) combined with deep learning reconstruction (DLR) in lumbar spine magnetic resonance imaging (MRI).Materials and Methods A retrospective analysis was conducted on 116 patients who underwent lumbar spine MRI. Among them, 54 patients received standard imaging with standard reconstruction (group A), and 62 patients underwent AS (group B). The AS images were further processed using two reconstruction methods: standard reconstruction (group B1, n = 62) and DLR (group B2, n = 62). Image quality was assessed using signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR), and a 5-point Likert scale. Inter-observer agreement was evaluated using the intra-class correlation coefficient (ICC) and Kappa statistics.Results No significant differences were observed in age, sex, or referral reasons between group A and group B (all P > 0.05). For scan Time: Compared with group A, group B achieved a 46% reduction in total acquisition time (185 seconds reduction). For objective image quality: On T1-weighted imaging: SNR of all tissues and CNR of vertebral body were higher in group B2 than in group A and group B1 (all P < 0.05), while no significant difference was found in CNR of the spinal cord among the three groups (P > 0.05). On T2-weighted imaging, no significant differences were observed in SNR or CNR of any tissues among the three groups (all P > 0.05). For subjective image quality: On T1-weighted imaging, group B2 demonstrated higher scores for artifacts, noise, overall image quality, and diagnostic confidence than group A and group B1 (all P < 0.001); the score for anatomical structure display was higher in group B2 than in group A (P < 0.05), and the noise score was higher in group B1 than in group A (P < 0.05). On T2-weighted imaging, group B2 showed higher scores for anatomical structure display, artifacts, noise, overall image quality, and diagnostic confidence than group A and group B1 (all P < 0.001). For consistency analysis: The ICC values for SNR and CNR measurements were 0.980 (95% CI: 0.963 to 0.990) and 0.972 (95% CI: 0.943 to 0.990), respectively. The ICC values for subjective scores on both T1WI and T2WI were all > 0.900. The Kappa values for the grading diagnosis of spinal stenosis and disc abnormalities ranged from 0.876 to 0.944 (all P < 0.001), indicating good inter-observer agreement.Conclusions AS combined with DLR significantly shortens lumbar spine MRI acquisition time while improving image quality and demonstrating good diagnostic consistency with standard imaging for various spinal abnormalities, offering substantial clinical value for enhancing workflow efficiency and patient comfort.
[Keywords] magnetic resonance imaging；deep learning reconstruction；image reconstruction；lumbar degenerative diseases；image quality

HE Xiaoqun1   LIU Meishan1   ZHANG Zhiwei1*   LUO Xi1   ZHANG Xingyue1   YU Bin1   PAN Boyang2   GUO Ziheng3   GONG Nanjie4, 5  

1 Department of Radiology, the First Affiliated Hospital of Chongqing Medical University, Chongqing, 400016, China

2 Tsinghua Cross-Strait Research Institute, Xiamen 361000, China

3 RadioDynamic Medical, Shanghai, 200031, China

4 Institute of Magnetic Resonance and Molecular Imaging in Medicine, East China Normal University, Shanghai, 200062, China

5 School of Physics, East China Normal University, Shanghai, 200062, China

Corresponding author: ZHANG Z W, E-mail: zhangzhiweicqmu@163.com

Conflicts of interest   None.

Received  2025-09-29
Accepted  2026-05-30
DOI: 10.12015/issn.1674-8034.2026.06.016
Cite this article as HE X Q, LIU M S, ZHANG Z W, et al. Application value of deep learning reconstruction algorithm in accelerated lumbar spine MRI: A clinical study on image quality and efficiency optimization[J]. Chin J Magn Reson Imaging, 2026, 17(6): 125-132. DOI:10.12015/issn.1674-8034.2026.06.016.

[1]
BERKARDA Z, WIEDEMANN S, WILPERT C, et al. Deep learning reconstructed T2-weighted Dixon imaging of the spine: Impact on acquisition time and image quality[J/OL]. Eur J Radiol, 2024, 178: 111633 [2025-09-21]. https://pubmed.ncbi.nlm.nih.gov/39067266/. DOI: 10.1016/j.ejrad.2024.111633.
[2]
ESTLER A, HAUSER T K, BRUNNÉE M, et al. Deep learning-accelerated image reconstruction in back pain-MRI imaging: reduction of acquisition time and improvement of image quality[J]. Radiol Med, 2024, 129(3): 478-487. DOI: 10.1007/s11547-024-01787-x.
[3]
KNEZEVIC N N, CANDIDO K D, VLAEYEN J W S, et al. Low back pain[J]. Lancet, 2021, 398(10294): 78-92. DOI: 10.1016/S0140-6736(21)00733-9.
[4]
ESPOSITO D, BROWN B, HANCOCK M J, et al. MRI-based grading systems for assessing lumbar disc degeneration: a scoping review[J/OL]. JOR Spine, 2025, 8(3): e70113 [2025-08-22]. https://pubmed.ncbi.nlm.nih.gov/40959741/. DOI: 10.1002/jsp2.70113.
[5]
LIAWRUNGRUEANG W, PARK J B, CHOLAMJIAK W, et al. Artificial intelligence-assisted MRI diagnosis in lumbar degenerative disc disease: a systematic review[J]. Global Spine J, 2025, 15(2): 1405-1418. DOI: 10.1177/21925682241274372.
[6]
SETSOMPOP K, GAGOSKI B A, POLIMENI J R, et al. Blipped-controlled aliasing in parallel imaging for simultaneous multislice echo planar imaging with reduced g-factor penalty[J]. Magn Reson Med, 2012, 67(5): 1210-1224. DOI: 10.1002/mrm.23097.
[7]
TSAO J, KOZERKE S. MRI temporal acceleration techniques[J]. J Magn Reson Imaging, 2012, 36(3): 543-560. DOI: 10.1002/jmri.23640.
[8]
YANG J, MIAO Y W. Progress in the clinical application of compressed sensing technology in brain MRI[J]. Chin J Magn Reson Imaging, 2022, 13(9): 144-147, 159. DOI: 10.12015/issn.1674-8034.2022.09.034.
[9]
Chinese Association of Radiologists, Chinese Society of Imaging Technology of Chinese Medical Association, Chinese Professional Committee of Medical Technologists, et al. Chinese expert consensus on the clinical application of deep learning image reconstruction algorithms for CT[J]. Chin J Radiol, 2024, 58(7): 697-706. DOI: 10.3760/cma.j.cn112149-20240111-00021.
[10]
WU Y, ZHANG X Q. Artificial intelligence in medical image processing: progress and prospect[J]. J Third Mil Med Univ, 2021, 43(18): 1707-1712. DOI: 10.16016/j.1000-5404.202106194.
[11]
WANG S S, WU R Y, JIA S, et al. Knowledge-driven deep learning for fast MR imaging: Undersampled MR image reconstruction from supervised to un-supervised learning[J]. Magn Reson Med, 2024, 92(2): 496-518. DOI: 10.1002/mrm.30105.
[12]
Chinese Society of Imaging Technology, Chinese Society of Radiology. Expert consensus on MRI examination techniques [J]. Chin J Radiol, 2016, 50(10): 724-739. DOI: 10.3760/cma.j.issn.1005-1201.2016.10.002.
[13]
GOODFELLOW I, POUGET-ABADIE J, MIRZA M, et al. Generative adversarial networks[J]. Commun ACM, 2020, 63(11): 139-144. DOI: 10.1145/3422622.
[14]
LEDIG C, THEIS L, HUSZAR F, et al. Photo-realistic single image super-resolution using a generative adversarial network[C]//2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). July 21-26, 2017. Honolulu, HI. IEEE, 2017: 105-114. DOI: 10.1109/cvpr.2017.19.
[15]
WANG X T, YU K, WU S X, et al. ESRGAN: enhanced super-resolution generative adversarial networks[EB/OL]. 2018: arXiv: 1809.00219 [2025-08-22]. https://arxiv.org/abs/1809.00219. DOI: 10.1007/978-3-030-01264-9_1.
[16]
HUANG G, LIU Z, VAN DER MAATEN L, et al. Densely connected convolutional networks[C]//2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). July 21-26, 2017. Honolulu, HI. IEEE, 2017: 2261-2269. DOI: 10.1109/cvpr.2017.243.
[17]
MIYATO T, KATAOKA T, KOYAMA M, et al. Spectral normalization for generative adversarial networks[EB/OL]. 2018: arXiv: 1802.05957 [2025-08-22]. https://arxiv.org/abs/1802.05957. DOI: 10.48550/arXiv.1802.05957.
[18]
ZHU B, LIU J Z, CAULEY S F, et al. Image reconstruction by domain-transform manifold learning[J]. Nature, 2018, 555(7697): 487-492. DOI: 10.1038/nature25988.
[19]
Professional Committee for Osteoporosis Prevention and Rehabilitation of Chinese Association for Rehabilitation Medicine, Orthopedic Minimally Invasive Branch of China Association for the Elderly Health Care. Expert consensus on diagnosis and treatment for degenerative lumbar spinal stenosis[J]. Chin J Bone Jt Surg, 2023, 16(2): 97-103. DOI: 10.3969/j.issn.2095-9958.2023.02.01.
[20]
QIN X K, SUN K, FENG T X, et al. Methodological quality evaluation and comprehensive analysis of clinical practice guidelines and expert consensus for lumbar disc herniation[J]. Chin Gen Pract, 2024, 27(31): 3857-3864. DOI: 10.12114/j.issn.1007-9572.2024.0051.
[21]
PAN K, HUANG X H, LIU N, et al. Study on application value of proton density weighted imaging accelerated with artificial intelligence-compressed sensing in assessing cartilage injury in osteoarthritis of the knee[J]. Chin J Magn Reson Imaging, 2022, 13(10): 138-143, 156. DOI: 10.12015/issn.1674-8034.2022.10.021.
[22]
FERVERS P, ZAESKE C, RAUEN P, et al. 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[J/OL]. Diagnostics (Basel), 2023, 13(3): 418 [2025-08-22]. https://pubmed.ncbi.nlm.nih.gov/36766523/. DOI: 10.3390/diagnostics13030418.
[23]
HU Y H, ZHAO W J, WANG W J, et al. Clinical application of lumbar spine MRI based on deep learning algorithm[J]. Chin J CT MRI, 2024, 22(12): 160-163. DOI: 10.3969/j.issn.1672-5131.2024.12.050.
[24]
WANG X T, XIE L B, DONG C, et al. Real-ESRGAN: training real-world blind super-resolution with pure synthetic data[EB/OL]. 2021: arXiv: 2107.10833 [2025-08-22]. https://arxiv.org/abs/2107.10833.
[25]
REKA S, PRABA T S, PRASANNA M, et al. Automated high precision PCOS detection through a segment anything model on super resolution ultrasound ovary images[J/OL]. Sci Rep, 2025, 15(1): 16832 [2025-08-22]. https://pubmed.ncbi.nlm.nih.gov/40369044/. DOI: 10.1038/s41598-025-01744-2.
[26]
JOHNSON P M, LIN D J, ZBONTAR J, et al. Deep learning reconstruction enables prospectively accelerated clinical knee MRI[J/OL]. Radiology, 2023, 307(2): e220425 [2025-08-22]. https://pubmed.ncbi.nlm.nih.gov/36648347/. DOI: 10.1148/radiol.220425.
[27]
KANIEWSKA M, DEININGER-CZERMAK E, GETZMANN J M, et al. Application of deep learning-based image reconstruction in MR imaging of the shoulder joint to improve image quality and reduce scan time[J]. Eur Radiol, 2023, 33(3): 1513-1525. DOI: 10.1007/s00330-022-09151-1.
[28]
HALLINAN J T P D. Deep learning for spine MRI: reducing time not quality[J/OL]. Radiology, 2023, 306(3): e222410 [2025-08-22]. https://pubmed.ncbi.nlm.nih.gov/36648347/. DOI: 10.1148/radiol.222410.
[29]
ALMANSOUR H, HERRMANN J, GASSENMAIER S, et al. Deep learning reconstruction for accelerated spine MRI: prospective analysis of interchangeability[J/OL]. Radiology, 2023, 306(3): e212922 [2025-08-22]. https://pubmed.ncbi.nlm.nih.gov/36318032/. DOI: 10.1148/radiol.212922.
[30]
TANG H, HONG M, YU L, et al. Deep learning reconstruction for lumbar spine MRI acceleration: a prospective study[J/OL]. Eur Radiol Exp, 2024, 8(1): 67 [2025-08-22]. https://pubmed.ncbi.nlm.nih.gov/38902467/. DOI: 10.1186/s41747-024-00470-0.
[31]
WU X X, LU X F, LIU C S, et al. Clinical feasibility of 2D FSE sequences of the knee MRI protocol using deep-learning image reconstruction[J]. Chin J Magn Reson Imaging, 2023, 14(5): 53-59. DOI: 10.12015/issn.1674-8034.2023.05.011.
[32]
YAN F H. The clinical application and development prospect of deep learning MRI reconstruction algorithm[J]. Chin J Magn Reson Imaging, 2023, 14(5): 8-10. DOI: 10.12015/issn.1674-8034.2023.05.002.
[33]
DEL GRANDE F, GUGGENBERGER R, FRITZ J. Rapid musculoskeletal MRI in 2021: value and optimized use of widely accessible techniques[J]. Am J Roentgenol, 2021, 216(3): 704-717. DOI: 10.2214/ajr.20.22901.

PREV A study on associated factors of adolescent knee joint sports injuries based on MRI examinations
NEXT Performance evaluation of metamaterials for rat T1/T2 weighted imaging in 5.0 T MRI: A validation study of image quality and imaging efficiency
  


LINK


Tel & Fax: +8610-67113815    E-mail: editor@cjmri.cn