Share：
Share this content in WeChat
X
[Chinese][PDF]551
• Technical Article •
Clinical applications of deep learning-based methods for generating high-resolution magnetic resonance enhanced images in carotid arteries
CAO Bo  YU Fan  FENG Mengmeng  LU Jie 

Cite this article as: CAO B, YU F, FENG M M, et al. Clinical applications of deep learning-based methods for generating high-resolution magnetic resonance enhanced images in carotid arteries[J]. Chin J Magn Reson Imaging, 2024, 15(10): 141-147. DOI:10.12015/issn.1674-8034.2024.10.024.


[Abstract] Objective Utilizing deep learning methodologies, we study the features of arteries and plaques within high-resolution magnetic resonance imaging (HR-MRI) of the carotid artery. This enables the generation of virtual contrast-enhanced T1WI (vce-T1WI) from plain T1WI. Furthermore, the detection level of the lipid-rich necrotic core within the carotid artery plaque in these generated vce-T1WI images is evaluated.Materials and Methods Incorporating 303 cases of patients with carotid artery stenosis, a total of 486 bilateral carotid arteries were scanned using T1WI images and actual contrast-enhanced T1WI (CE-T1WI) images. These were divided into training, validation, and testing sets at a ratio of 4∶1∶1, and the generative network was trained using five-fold cross-validation. The test set comprised 81 carotid artery images. Two distinct deep learning strategies (pix2pix, Cycle GAN) were employed to generate vce-T1WI images from plain T1WI scans. The quality of the vce-T1WI images generated by the two models was evaluated using the peak signal-to-noise ratio (PSNR), structural similarity (SSIM), and subjective visual quality scores. The diagnostic efficacy of the generated vce-T1WI images for the lipid-rich necrotic core within the plaque was assessed, with the physician's judgment based on the actual CE-T1WI images serving as the gold standard.Results The vce-T1WI images generated by pix2pix and Cycle GAN achieved PSNR scores of 20.206 and 19.717 respectively in the same test set, with SSIM scores of 0.591 and 0.635 respectively. The proportion of images scoring more than 2 points in the subjective visual quality assessment was 95.1% and 97.5% respectively. The detection accuracy for the lipid-rich necrotic core within the plaque was 82.7% and 74.1% respectively.Conclusions Deep learning methodologies can effectively generate high-quality vce-T1WI images from plain T1WI scans of the carotid artery HR-MRI. Furthermore, the virtual contrast-enhanced images generated by Cycle GAN exhibit a high detection accuracy for the lipid-rich necrotic core within the plaque. Deep learning techniques can broaden the clinical application scope of HR-MRI and reduce the risk of adverse reactions to contrast agents.
[Keywords] atherosclerotic plaque；magnetic resonance imaging；deep learning；image generation；computer aided diagnosis

CAO Bo1, 2   YU Fan1, 2   FENG Mengmeng1, 2   LU Jie1, 2*  

1 Department of Radiology and Nuclear Medicine, Xuanwu Hospital Capital Medical University, Beijing 100053, China

2 Beijing Key Lab of MRI and Brain Informatics, Beijing 100053, China

Corresponding author: LU J, E-mail: imaginglu@hotmail.com

Conflicts of interest   None.

Received  2024-06-27
Accepted  2024-10-10
DOI: 10.12015/issn.1674-8034.2024.10.024
Cite this article as: CAO B, YU F, FENG M M, et al. Clinical applications of deep learning-based methods for generating high-resolution magnetic resonance enhanced images in carotid arteries[J]. Chin J Magn Reson Imaging, 2024, 15(10): 141-147. DOI:10.12015/issn.1674-8034.2024.10.024.

[1]
YANG R W, YUAN J, CHEN X E, et al. Vessel wall magnetic resonance imaging of symptomatic middle cerebral artery atherosclerosis: a systematic review and meta-analysis[J/OL]. Clin Imaging, 2022, 90: 90-96 [2024-06-26]. https://pubmed.ncbi.nlm.nih.gov/35952437/. DOI: 10.1016/j.clinimag.2022.08.001.
[2]
MANDELL D M, MOSSA-BASHA M, QIAO Y, et al. Intracranial vessel wall MRI: principles and expert consensus recommendations of the American society of neuroradiology[J]. AJNR Am J Neuroradiol, 2017, 38(2): 218-229. DOI: 10.3174/ajnr.A4893.
[3]
BOS D, ARSHI B, VAN DEN BOUWHUIJSEN Q J A, et al. Atherosclerotic carotid plaque composition and incident stroke and coronary events[J]. J Am Coll Cardiol, 2021, 77(11): 1426-1435. DOI: 10.1016/j.jacc.2021.01.038.
[4]
WANG Y L, WANG T, LUO Y M, et al. Identification markers of carotid vulnerable plaques: an update[J/OL]. Biomolecules, 2022, 12(9): 1192 [2024-06-26]. https://pubmed.ncbi.nlm.nih.gov/36139031/. DOI: 10.3390/biom12091192.
[5]
BENSON J C, SABA L, BATHLA G, et al. MR imaging of carotid artery atherosclerosis: updated evidence on high-risk plaque features and emerging trends[J]. AJNR Am J Neuroradiol, 2023, 44(8): 880-888. DOI: 10.3174/ajnr.A7921.
[6]
GIMNICH O A, ZIL-E-ALI A, BRUNNER G. Imaging approaches to the diagnosis of vascular diseases[J]. Curr Atheroscler Rep, 2022, 24(2): 85-96. DOI: 10.1007/s11883-022-00988-x.
[7]
MA Z L, HUO M J, XIE J J, et al. Wall characteristics of atherosclerotic middle cerebral arteries in patients with single or multiple infarcts: a high-resolution MRI Study[J/OL]. Front Neurol, 2022, 13: 934926 [2024-06-26]. https://pubmed.ncbi.nlm.nih.gov/36408522/. DOI: 10.3389/fneur.2022.934926.
[8]
SABA L C, NARDI V, CAU R, et al. Carotid artery plaque calcifications: lessons from histopathology to diagnostic imaging[J]. Stroke, 2022, 53(1): 290-297. DOI: 10.1161/STROKEAHA.121.035692.
[9]
TENG Z Z, BROWN A J, GILLARD J H. From ultrasonography to high resolution magnetic resonance imaging: towards an optimal management strategy for vulnerable carotid atherosclerotic plaques[J/OL]. EBioMedicine, 2016, 3: 2-3 [2024-06-26]. https://pubmed.ncbi.nlm.nih.gov/26870831/. DOI: 10.1016/j.ebiom.2016.01.001.
[10]
SHIRAKAWA M, YAMADA K, WATASE H, et al. Atherosclerotic carotid plaque characteristics vary with time from ischemic event: a multicenter, prospective magnetic resonance vessel wall imaging registry study[J/OL]. J Neurol Sci, 2023, 446: 120582 [2024-06-26]. https://pubmed.ncbi.nlm.nih.gov/36796273/. DOI: 10.1016/j.jns.2023.120582.
[11]
LIU H N, SUN J, HIPPE D S, et al. Improved carotid lumen delineation on non-contrast MR angiography using SNAP (Simultaneous Non-Contrast Angiography and Intraplaque Hemorrhage) imaging[J/OL]. Magn Reson Imaging, 2019, 62: 87-93 [2024-06-26]. https://pubmed.ncbi.nlm.nih.gov/31247251/. DOI: 10.1016/j.mri.2019.06.012.
[12]
PEREIRA T, BETRIU A, ALVES R. Non-invasive imaging techniques and assessment of carotid Vasa vasorum neovascularization: promises and pitfalls[J]. Trends Cardiovasc Med, 2019, 29(2): 71-80. DOI: 10.1016/j.tcm.2018.06.007.
[13]
SABA L, YUAN C, HATSUKAMI T S, et al. Carotid artery wall imaging: perspective and guidelines from the ASNR vessel wall imaging study group and expert consensus recommendations of the American society of neuroradiology[J/OL]. AJNR Am J Neuroradiol, 2018, 39(2): E9-E31 [2024-06-26]. https://pubmed.ncbi.nlm.nih.gov/29326139/. DOI: 10.3174/ajnr.A5488.
[14]
ZHANG Y, LU J. Research progresses in evaluating the stability of carotid atherosclerotic plaque with positron emission tomography/computed tomography[J]. J Cap Med Univ, 2021, 42(1): 31-36. DOI: 10.3969/j.issn.1006-7795.2021.01.006].
[15]
AHN Y H, KANG D Y, PARK S B, et al. Allergic-like hypersensitivity reactions to gadolinium-based contrast agents: an 8-year cohort study of 154 539 patients[J]. Radiology, 2022, 303(2): 329-336. DOI: 10.1148/radiol.210545.
[16]
BEHZADI A H, ZHAO Y Z, FAROOQ Z, et al. Immediate allergic reactions to gadolinium-based contrast agents: a systematic review and meta-analysis[J]. Radiology, 2018, 286(2): 471-482. DOI: 10.1148/radiol.2017162740.
[17]
WEBERLING L D, KIESLICH P J, KICKINGEREDER P, et al. Increased signal intensity in the dentate nucleus on unenhanced T1-weighted images after gadobenate dimeglumine administration[J]. Invest Radiol, 2015, 50(11): 743-748. DOI: 10.1097/RLI.0000000000000206.
[18]
DAYARATHNA S, ISLAM K T, URIBE S, et al. Deep learning based synthesis of MRI, CT and PET: review and analysis[J/OL]. Med Image Anal, 2024, 92: 103046 [2024-06-26]. https://pubmed.ncbi.nlm.nih.gov/38052145/. DOI: 10.1016/j.media.2023.103046.
[19]
MÜLLER-FRANZES G, NIEHUES J M, KHADER F, et al. A multimodal comparison of latent denoising diffusion probabilistic models and generative adversarial networks for medical image synthesis[J/OL]. Sci Rep, 2023, 13(1): 12098 [2024-06-26]. https://pubmed.ncbi.nlm.nih.gov/37495660/. DOI: 10.1038/s41598-023-39278-0.
[20]
LI W, XIAO H N, LI T, et al. Virtual contrast-enhanced magnetic resonance images synthesis for patients with nasopharyngeal carcinoma using multimodality-guided synergistic neural network[J]. Int J Radiat Oncol Biol Phys, 2022, 112(4): 1033-1044. DOI: 10.1016/j.ijrobp.2021.11.007.
[21]
XIE Q, LIN Y S, WANG M Y, et al. Synthesis of gadolinium-enhanced glioma images on multisequence magnetic resonance images using contrastive learning[J]. Med Phys, 2024, 51(7): 4888-4897. DOI: 10.1002/mp.17004.
[22]
LYU J H, FU Y, YANG M L, et al. Generative adversarial network-based noncontrast CT angiography for aorta and carotid arteries[J/OL]. Radiology, 2023, 309(2): e230681 [2024-06-26]. https://pubmed.ncbi.nlm.nih.gov/37962500/. DOI: 10.1148/radiol.230681.
[23]
ISOLA P, ZHU J Y, ZHOU T H, et al. Image-to-image translation with conditional adversarial networks[C]//2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Honolulu, HI, USA. IEEE, 2017: 5967-5976. DOI: 10.1109/CVPR.2017.632.
[24]
ZHU J Y, PARK T, ISOLA P, et al. Unpaired image-to-image translation using cycle-consistent adversarial networks[C]//2017 IEEE International Conference on Computer Vision (ICCV). Venice, Italy. IEEE, 2017: 2242-2251. DOI: 10.1109/ICCV.2017.244.
[25]
ZHOU Z W, SIDDIQUEE M M R, TAJBAKHSH N, et al. UNet++: redesigning skip connections to exploit multiscale features in image segmentation[J]. IEEE Trans Med Imaging, 2020, 39(6): 1856-1867. DOI: 10.1109/TMI.2019.2959609.
[26]
CAI J M, HATSUKAMI T S, FERGUSON M S, et al. Classification of human carotid atherosclerotic lesions with in vivo multicontrast magnetic resonance imaging[J]. Circulation, 2002, 106(11): 1368-1373. DOI: 10.1161/01.cir.0000028591.44554.f9.
[27]
JI S, YANG D J, LEE J, et al. Synthetic MRI: technologies and applications in neuroradiology[J]. J Magn Reson Imaging, 2022, 55(4): 1013-1025. DOI: 10.1002/jmri.27440.
[28]
WANG J, YU F, ZHANG M Z, et al. A 3D framework for segmentation of carotid artery vessel wall and identification of plaque compositions in multi-sequence MR images[J/OL]. Comput Med Imaging Graph, 2024, 116: 102402 [2024-06-26]. https://pubmed.ncbi.nlm.nih.gov/38810486/. DOI: 10.1016/j.compmedimag.2024.102402.
[29]
LEE J, KIM B, PARK H. MC2-Net: motion correction network for multi-contrast brain MRI[J]. Magn Reson Med, 2021, 86(2): 1077-1092. DOI: 10.1002/mrm.28719.
[30]
LIU J, WANG C Y, WANG J N, et al. Motion detection and correction for carotid MRI using a markerless optical system[J/OL]. Magn Reson Imaging, 2022, 94: 161-167 [2024-06-26]. https://pubmed.ncbi.nlm.nih.gov/36191857/. DOI: 10.1016/j.mri.2022.09.010.
[31]
BAO Q J, CHEN Y L, BAI C X, et al. Retrospective motion correction for preclinical/clinical magnetic resonance imaging based on a conditional generative adversarial network with entropy loss[J/OL]. NMR Biomed, 2022, 35(12): e4809 [2024-06-26]. https://pubmed.ncbi.nlm.nih.gov/35925046/. DOI: 10.1002/nbm.4809.

PREV Study on the correlation between DWI, IVIM, and DCE-MRI parameters and Ki-67 expression in soft tissue tumors
NEXT Application value of intelligent quick magnetic resonance technology in supraspinatus tendon injuries
  


LINK


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