J Shanghai Jiaotong Univ Sci

Journal of Shanghai Jiaotong University(Science) >

2026 , Vol. 31 >Issue 3: 611 - 621

DOI: https://doi.org/10.1007/s12204-026-2927-8

Celebrating 30 Years

Deep Learning-Based Reconstruction of Temporomandibular Joint MRI at 1.5-T and 3-T

Expand
  • 1. Department of Radiology, Shanghai Ninth People's Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai 200011, China; 2. College of Basic Medical Sciences, Shanghai Jiao Tong University School of Medicine, Shanghai 200025, China

Received date: 2026-01-04

  Revised date: 2026-02-28

  Accepted date: 2026-03-20

  Online published: 2026-05-16

Fold

Abstract

Image quality and diagnostic performance between a fast spin-echo sequence with standard parameters (FSE-SD) and a deep learning-reconstructed fast spin echo with low parameters (FSE-DL) are compared for temporomandibular joint (TMJ) magnetic resonance imaging (MRI) at 1.5-T and 3-T scanners. This study enrolled 183 patients (94 scanned at 1.5-T and 89 at 3-T) who underwent both FSE-SD and FSE-DL acquisitions for temporomandibular disorders. Scan time, subjective image quality, detection rates of pathological features, and the inter-protocol and inter-reader agreement were assessed by two readers. The Wilcoxon signed-rank test, McNemar test, and unweighted/linearly weighted Cohen’s κ statistics were used. Using the deep learning algorithm, total scan time at 1.5-T was reduced from 316 s (FSE-SD) to 211 s (FSE-DL), representing a 33.23% reduction. At 3-T, scan time was similarly shortened from 329 s (FSE-SD) to 189 s (FSE-DL), with a 42.55% decrease. Compared with FSE-SD, FSE-DL showed lower noise and fewer artifacts, with overall image quality similar or slightly improved at both 1.5-T and 3-T; most of these differences were statistically significant. Diagnostic confidence was not significantly different between the two protocols. Interreader and interprotocol agreement ranged from moderate to almost perfect (κ values ranging from 0.602 to 1.000). Detection rates of major TMJ pathologies were similar between the two protocols for both readers at 1.5-T and 3-T (p ⩾ 0.375). FSE-DL achieved a significant reduction in scan time for TMJ MRI at both 1.5-T and 3-T while maintaining overall image quality, and demonstrated pathological feature detection performance comparable to that of the FSE-SD.

Cite this article

Zhou Xin, Li Xiaomin, Zhang Zhengjia, Ma Hairong, Cao Ran, Wu Jianwei, Sun Qi, Ai Songtao . Deep Learning-Based Reconstruction of Temporomandibular Joint MRI at 1.5-T and 3-T[J]. Journal of Shanghai Jiaotong University(Science), 2026 , 31(3) : 611 -621 . DOI: 10.1007/s12204-026-2927-8

References

[1] Chagas-Neto F A, de Sá Neto J L, Agnollitto P M. Temporomandibular joint: Review of the anatomy, pathology, and magnetic resonance imaging techniques [M]//Craniofacial pain. Cham: Springer Nature Switzerland, 2024: 91-122.
[2] Wilkie G, Al-Ani Z. Temporomandibular joint anatomy, function and clinical relevance [J]. British Dental Journal, 2022, 233(7): 539-546.
[3] Gharavi S M, Qiao Y J, Faghihimehr A, et al. Imaging of the temporomandibular joint [J]. Diagnostics, 2022, 12(4): 1006.
[4] Mallya S M, Ahmad M, Cohen J R, et al. Recommendations for imaging of the temporomandibular joint. position statement from the American academy of oral and maxillofacial radiology and the American academy of orofacial pain [J]. Journal of Oral and Facial Pain and Headache, 2023, 37(1): 7-15.
[5] Wang G. Making “CASES” for AI in medicine [J]. BME Frontiers, 2024, 5: 36.
[6] Heye T, Knoerl R, Wehrle T, et al. The energy consumption of radiology: Energy- and cost-saving opportunities for CT and MRI operation [J]. Radiology, 2020, 295(3): 593-605.
[7] Lustig M, Donoho D L, Santos J M, et al. Compressed sensing MRI [J]. IEEE Signal Processing Magazine, 2008, 25(2): 72-82.
[8] Schlemper J, Caballero J, Hajnal J V, et al. A deep cascade of convolutional neural networks for dynamic MR image reconstruction [J]. IEEE Transactions on Medical Imaging, 2018, 37(2): 491-503.
[9] Lustig M, Donoho D, Pauly J M. Sparse MRI: The application of compressed sensing for rapid MR imaging [J]. Magnetic Resonance in Medicine, 2007, 58(6): 1182-1195.
[10] Aggarwal H K, Mani M P, Jacob M. MoDL: Model-based deep learning architecture for inverse problems [J]. IEEE Transactions on Medical Imaging, 2019, 38(2): 394-405.
[11] Xie Y X, Tao H Y, Li X W, et al. Prospective comparison of standard and deep learning-reconstructed turbo spin-echo MRI of the shoulder [J]. Radiology, 2024, 310(1): e231405.
[12] Almansour H, Herrmann J, Gassenmaier S, et al. Deep learning reconstruction for accelerated spine MRI: Prospective analysis of interchangeability [J]. Radiology, 2023, 306(3): e212922.
[13] Johnson P M, Lin D J, Zbontar J, et al. Deep learning reconstruction enables prospectively accelerated clinical knee MRI [J]. Radiology, 2023, 307(2): e220425.
[14] Keller G, Estler A, Herrmann J, et al. Prospective intraindividual comparison of a standard 2D TSE MRI protocol for ankle imaging and a deep learning-based 2D TSE MRI protocol with a scan time reduction of 48% [J]. La Radiologia Medica, 2023, 128(3): 347-356.
[15] Herrmann J, Afat S, Gassenmaier S, et al. Image quality and diagnostic performance of accelerated 2D hip MRI with deep learning reconstruction based on a deep iterative hierarchical network [J]. Diagnostics, 2023, 13(20): 3241.
[16] 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]. European Radiology, 2023, 33(3): 1513-1525.
[17] Jo G D, Jeon K J, Choi Y J, et al. Deep learning reconstruction for temporomandibular joint MRI: Diagnostic interchangeability, image quality, and scan time reduction [J]. European Radiology, 2026, 36(3): 1801-1812.
[18] Dong M J, Sun Q, Yu Q, et al. Determining the optimal magnetic resonance imaging sequences for the efficient diagnosis of temporomandibular joint disorders [J]. Quantitative Imaging in Medicine and Surgery, 2021, 11(4): 1343-1353.
[19] Subtle Medical Inc. SubtleMR - FDA 510(k) Summary (K223623) [EB/OL]. (2023-05-11) [2025-09-25]. https://www.accessdata.fda.gov/cdrh_docs/pdf22/K223623.pdf.
[20] Haris M, Shakhnarovich G, Ukita N. Deep back-projection networks for super-resolution [C]//2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Salt Lake City: IEEE, 2018: 1664-1673.
[21] Ahmad M, Hollender L, Anderson Q, et al. Research diagnostic criteria for temporomandibular disorders (RDC/TMD): Development of image analysis criteria and examiner reliability for image analysis [J]. Oral Surgery, Oral Medicine, Oral Pathology, Oral Radiology, and Endodontology, 2009, 107(6): 844-860.
[22] Litjens G, Kooi T, Bejnordi B E, et al. A survey on deep learning in medical image analysis [J]. Medical Image Analysis, 2017, 42: 60-88.
[23] Li Z Q, Karis J P, Pipe J G. A 2D spiral turbo-spin-echo technique [J]. Magnetic Resonance in Medicine, 2018, 80(5): 1989-1996.
[24] He K M, Zhang X Y, Ren S Q, et al. Deep residual learning for image recognition [C]//2016 IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas: IEEE, 2016: 770-778.
[25] Dratsch T, Siedek F, Zäske C, et al. Reconstruction of shoulder MRI using deep learning and compressed sensing: A validation study on healthy volunteers [J]. European Radiology Experimental, 2023, 7(1): 66.
[26] Tang H, Hong M, Yu L, et al. Deep learning reconstruction for lumbar spine MRI acceleration: A prospective study [J]. European Radiology Experimental, 2024, 8(1): 67.
[27] Manoliu A, Spinner G, Wyss M, et al. Quantitative and qualitative comparison of MR imaging of the temporomandibular joint at 1.5 and 3.0 T using an optimized high-resolution protocol [J]. Dento Maxillo Facial Radiology, 2016, 45(1): 20150240.
[28] Strzelecki M, Piórkowski A, Obuchowicz R. Effect of matrix size reduction on textural information in clinical magnetic resonance imaging [J]. Journal of Clinical Medicine, 2022, 11(9): 2526.
[29] Estler A, Zerweck L, Brunnée M, et al. Deep learning-accelerated image reconstruction in MRI of the orbit to shorten acquisition time and enhance image quality [J]. Journal of Neuroimaging, 2024, 34(2): 232-240.
[30] Koo S J, Yoo R E, Choi K S, et al. Deep learning-based reconstruction for accelerated cervical spine MRI: Utility in the evaluation of myelopathy and degenerative diseases [J]. AJNR American Journal of Neuroradiology, 2025, 46(4): 750-757.
[31] Herrmann J, Gassenmaier S, Keller G, et al. Deep learning MRI reconstruction for accelerating turbo spin echo hand and wrist imaging: A comparison of image quality, visualization of anatomy, and detection of common pathologies with standard imaging [J]. Academic Radiology, 2023, 30(11): 2606-2615.
[32] Feinstein A R, Cicchetti D V. High agreement but low Kappa: I. The problems of two paradoxes [J]. Journal of Clinical Epidemiology, 1990, 43(6): 543-549.

Options
Outlines

/

〈 〉