Share：
Share this content in WeChat
X
[Chinese] [PDF] 11 1
• Clinical Article •
Feasibility study on the classification of liver multi-parameter MRI sequences based on deep learning
ZHANG Duoduo  WU Pengsheng  WANG Xiangpeng  ZHANG Xiaodong  WANG Xiaoying 

Cite this article as: ZHANG D D, WU P S, WANG X P, et al. Feasibility study on the classification of liver multi-parameter MRI sequences based on deep learning[J]. Chin J Magn Reson Imaging, 2025, 16(10): 48-54, 82. DOI:10.12015/issn.1674-8034.2025.10.008.


[Abstract] Objective To investigate the feasibility of using a deep learning-based image classification model for distinguishing liver multi-parameter magnetic resonance imaging (mpMRI) sequences.Materials and Methods A retrospective dataset of 1744 liver mpMRI examinations from 1676 patients (November 16, 2022 to June 29, 2023) was collected as model development set, yielding 25 365 independent sequences. These were randomly divided into training [number of series (ns) = 20 207], validation (ns = 2664), and test sets (ns = 2494) at an 8∶1∶1 ratio. A 3D-ResNet model was trained to classify liver mpMRI sequences, with input as image and output categories including: T1-weighted in-phase (T1WI_In), T1-weighted opposed-phase (T1WI_Opp), T2-weighted imaging with fat-suppression (T2WI_Fs), high b-value DWI, ADC maps, and dynamic contrast-enhanced MRI (pre-contrast, arterial, portal venous, delayed). The Cancer Genome Atlas Liver Hepatocellular Carcinoma (TCGA_LIHC) dataset was used as the external validation set for the model, comprising a total of 59 mpMRI examinations involving 38 patients. Radiologists' classifications served as the gold standard. Model performance was evaluated using confusion matrices.Results At the overall classification level, the training, validation and test sets achieved average accuracy, macro-F1, and micro-F1 scores of 97.2% to 99.0%, 0.949 to 0.982 and 0.960 to 0.985, respectively. For individual sequences, the training, validation and test sets demonstrated per-class accuracy (89.6% to 100.0%), sensitivity (81.0% to 100.0%), specificity (98.2% to 100.0%), and F1 scores (0.797 to 1.000). On the external validation set, the model achieved macro-accuracy, macro-F1, and micro-F1 scores of 91.6%, 0.819, and 0.816, respectively. Per-sequence metrics included accuracy (74.1% to 99.4%), sensitivity (55.4% to 100.0%), specificity (92.8% to 100.0%), and F1 score (0.579 to 0.968).Conclusions The deep learning-based model demonstrated high accuracy in classifying liver mpMRI sequences, supporting its potential for automated sequence classification in clinical practice.
[Keywords] liver；magnetic resonance imaging；automated classification of images；artificial intelligence；deep learning

ZHANG Duoduo1   WU Pengsheng2   WANG Xiangpeng2   ZHANG Xiaodong1   WANG Xiaoying1*  

1 Department of Radiology, Peking University First Hospital, Beijing 100034, China

2 Beijing Smart Tree Medical Technology Co., Ltd, Beijing 100011, China

Corresponding author: WANG X Y, E-mail: wangxiaoying@bjmu.edu.cn

Conflicts of interest   None.

Received  2025-06-05
Accepted  2025-09-24
DOI: 10.12015/issn.1674-8034.2025.10.008
Cite this article as: ZHANG D D, WU P S, WANG X P, et al. Feasibility study on the classification of liver multi-parameter MRI sequences based on deep learning[J]. Chin J Magn Reson Imaging, 2025, 16(10): 48-54, 82. DOI:10.12015/issn.1674-8034.2025.10.008.

[1]
BRAY F, LAVERSANNE M, SUNG H, et al. Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries[J]. CA Cancer J Clin, 2024, 74(3): 229-263. DOI: 10.3322/caac.21834.
[2]
GANESAN P, KULIK L M. Hepatocellular carcinoma: new developments[J]. Clin Liver Dis, 2023, 27(1): 85-102. DOI: 10.1016/j.cld.2022.08.004.
[3]
VITALE A, CABIBBO G, IAVARONE M, et al. Personalised management of patients with hepatocellular carcinoma: a multiparametric therapeutic hierarchy concept[J/OL]. Lancet Oncol, 2023, 24(7): e312-e322 [2025-06-04]. https://pubmed.ncbi.nlm.nih.gov/37414020/. DOI: 10.1016/S1470-2045(23)00186-9.
[4]
XIA C F, BASU P, KRAMER B S, et al. Cancer screening in China: a steep road from evidence to implementation[J/OL]. Lancet Public Health, 2023, 8(12): e996-e1005 [2025-06-04]. https://pubmed.ncbi.nlm.nih.gov/38000379/. DOI: 10.1016/S2468-2667(23)00186-X.
[5]
Department of Medical Administration of the National Health Commission of the People's Republic of China. Guideline for diagnosis and treatment of primary liver cancer (2024 edition)[J]. Chin J Magn Reson Imaging, 2024, 15(6): 1-18. DOI: 10.12015/issn.1674-8034.2024.06.001.
[6]
HUANG M Q, SONG C Y, ZHOU X Q, et al. Tissue-matched analysis of MRI evaluating the tumor infiltrating lymphocytes in hepatocellular carcinoma[J]. Int J Cancer, 2025, 156(8): 1634-1643. DOI: 10.1002/ijc.35281.
[7]
SONG Q, GUO Y X, YAO X Z, et al. Comparative study of evaluating the microcirculatory function status of primary small HCC between the CE (DCE-MRI) and Non-CE (IVIM-DWI) MR Perfusion Imaging[J]. Abdom Radiol (NY), 2021, 46(6): 2575-2583. DOI: 10.1007/s00261-020-02945-1.
[8]
DRZAŁ A, JASIŃSKI K, GONET M, et al. MRI and US imaging reveal evolution of spatial heterogeneity of murine tumor vasculature[J/OL]. Magn Reson Imag, 2022, 92: 33-44 [2025-06-04]. https://www.sciencedirect.com/science/article/pii/S0730725X22000972. DOI: 10.1016/j.mri.2022.06.003.
[9]
MAINO C, VERNUCCIO F, CANNELLA R, et al. Liver metastases: The role of magnetic resonance imaging[J]. World J Gastroenterol, 2023, 29(36): 5180-5197. DOI: 10.3748/wjg.v29.i36.5180.
[10]
ICHIKAWA S, GOSHIMA S. Gadoxetic acid-enhanced liver MRI: everything you need to know[J]. Invest Radiol, 2024, 59(1): 53-68. DOI: 10.1097/RLI.0000000000000990.
[11]
MARRERO J A, KULIK L M, SIRLIN C B, et al. Diagnosis, staging, and management of hepatocellular carcinoma: 2018 practice guidance by the American association for the study of liver diseases[J]. Hepatology, 2018, 68(2): 723-750. DOI: 10.1002/hep.29913.
[12]
CHERNYAK V, FOWLER K J, KAMAYA A, et al. Liver imaging reporting and data system (LI-RADS) version 2018: imaging of hepatocellular carcinoma in at-risk patients[J]. Radiology, 2018, 289(3): 816-830. DOI: 10.1148/radiol.2018181494.
[13]
FALLAHPOOR M, NGUYEN D, MONTAHAEI E, et al. Segmentation of liver and liver lesions using deep learning[J]. Phys Eng Sci Med, 2024, 47(2): 611-619. DOI: 10.1007/s13246-024-01390-4.
[14]
WANG J, PENG Y Y, JING S, et al. A deep-learning approach for segmentation of liver tumors in magnetic resonance imaging using UNet+[J/OL]. BMC Cancer, 2023, 23(1): 1060 [2025-06-04]. https://pubmed.ncbi.nlm.nih.gov/37923988/. DOI: 10.1186/s12885-023-11432-x.
[15]
ZHEN S H, CHENG M, TAO Y B, et al. Deep learning for accurate diagnosis of liver tumor based on magnetic resonance imaging and clinical data[J/OL]. Front Oncol, 2020, 10: 680 [2025-06-04]. https://pubmed.ncbi.nlm.nih.gov/32547939/. DOI: 10.3389/fonc.2020.00680.
[16]
OESTMANN P M, WANG C J, SAVIC L J, et al. Deep learning-assisted differentiation of pathologically proven atypical and typical hepatocellular carcinoma (HCC) versus non-HCC on contrast-enhanced MRI of the liver[J]. Eur Radiol, 2021, 31(7): 4981-4990. DOI: 10.1007/s00330-020-07559-1.
[17]
HE X J, XU Y, ZHOU C Y, et al. Prediction of microvascular invasion and pathological differentiation of hepatocellular carcinoma based on a deep learning model[J/OL]. Eur J Radiol, 2024, 172: 111348 [2025-06-04]. https://pubmed.ncbi.nlm.nih.gov/38325190/. DOI: 10.1016/j.ejrad.2024.111348.
[18]
WANG F, CHEN Q Q, CHEN Y N, et al. A novel multimodal deep learning model for preoperative prediction of microvascular invasion and outcome in hepatocellular carcinoma[J]. Eur J Surg Oncol, 2023, 49(1): 156-164. DOI: 10.1016/j.ejso.2022.08.036.
[19]
ZHANG K, HE K, ZHANG L, et al. Gadoxetic acid-enhanced MRI scoring model to predict pathologic features of hepatocellular carcinoma[J/OL]. Radiology, 2025, 314(2): e233229 [2025-06-04]. https://pubmed.ncbi.nlm.nih.gov/39932410/. DOI: 10.1148/radiol.233229.
[20]
BAGHERI M H, ROTH H, KOVACS W, et al. Technical and clinical factors affecting success rate of a deep learning method for pancreas segmentation on CT[J]. Acad Radiol, 2020, 27(5): 689-695. DOI: 10.1016/j.acra.2019.08.014.
[21]
BAUMGÄRTNER G L, HAMM C A, SCHULZE-WEDDIGE S, et al. Metadata-independent classification of MRI sequences using convolutional neural networks: Successful application to prostate MRI[J/OL]. Eur J Radiol, 2023, 166: 110964 [2025-06-04]. https://pubmed.ncbi.nlm.nih.gov/37453274/. DOI: 10.1016/j.ejrad.2023.110964.
[22]
GAURIAU R, BRIDGE C, CHEN L N, et al. Using DICOM metadata for radiological image series categorization: a feasibility study on large clinical brain MRI datasets[J]. J Digit Imaging, 2020, 33(3): 747-762. DOI: 10.1007/s10278-019-00308-x.
[23]
ANAND A, LIU J F, SHEN T C, et al. Automated classification of intravenous contrast enhancement phase of CT scans using residual networks[J/OL]. Proc SPIE Int Soc Opt Eng, 2023, 12465: 124650O [2025-06-04]. https://pubmed.ncbi.nlm.nih.gov/40248190/. DOI: 10.1117/12.2655263.
[24]
ZHU Z, MITTENDORF A, SHROPSHIRE E, et al. 3D Pyramid Pooling Network for Abdominal MRI Series Classification[J]. IEEE Trans Pattern Anal Mach Intell, 2022, 44(4): 1688-1698. DOI: 10.1109/TPAMI.2020.3033990.
[25]
KIM B, MATHAI T S, HELM K, et al. Classification of multi-parametric body MRI series using deep learning[J]. IEEE J Biomed Health Inform, 2024, 28(11): 6791-6802. DOI: 10.1109/JBHI.2024.3448373.
[26]
MILLER C M, ZHU Z, MAZUROWSKI M A, et al. Automated selection of abdominal MRI series using a DICOM metadata classifier and selective use of a pixel-based classifier[J]. Abdom Radiol, 2024, 49(10): 3735-3746. DOI: 10.1007/s00261-024-04379-5.
[27]
CLARK K, VENDT B, SMITH K, et al. The cancer imaging archive (TCIA): maintaining and operating a public information repository[J]. J Digit Imaging, 2013, 26(6): 1045-1057. DOI: 10.1007/s10278-013-9622-7.
[28]
Cancer Genome Atlas Research Network Electronic address: wheeler@bcm edu, Cancer Genome Atlas Research Network. Comprehensive and integrative genomic characterization of hepatocellular carcinoma[J/OL]. Cell, 2017, 169(7): 1327-1341.e23 [2025-06-04]. https://pubmed.ncbi.nlm.nih.gov/28622513/. DOI: 10.1016/j.cell.2017.05.046.
[29]
NOGUCHI T, HIGA D, ASADA T, et al. Artificial intelligence using neural network architecture for radiology (AINNAR): classification of MR imaging sequences[J]. Jpn J Radiol, 2018, 36(12): 691-697. DOI: 10.1007/s11604-018-0779-3.
[30]
MA M M, QIN N S, JIANG Y, et al. Research on automatic classification of breast MRI images based on deep learning[J]. Chin J Magn Reson Imag, 2024, 15(1): 55-60. DOI: 10.12015/issn.1674-8034.2024.01.009.
[31]
YIN X M, WANG K X, WANG L, et al. Algorithms for classification of sequences and segmentation of prostate gland: an external validation study[J]. Abdom Radiol, 2024, 49(4): 1275-1287. DOI: 10.1007/s00261-024-04241-8.
[32]
SUN Z N, CUI Y P, LIU X, et al. Feasibility of prostate mpMRI sequence classification based on deep learning[J]. J Clin Radiol, 2022, 41(8): 1559-1564. DOI: 10.13437/j.cnki.jcr.2022.08.028.
[33]
KAMBADAKONE A R, FUNG A, GUPTA R T, et al. LI-RADS technical requirements for CT, MRI, and contrast-enhanced ultrasound[J]. Abdom Radiol (NY), 2018, 43(1): 56-74. DOI: 10.1007/s00261-017-1325-y.
[34]
HUANG B Y, CASTILLO M. Neurovascular imaging at 1.5 tesla versus 3.0 tesla[J]. Magn Reson Imaging Clin N Am, 2009, 17(1): 29-46. DOI: 10.1016/j.mric.2008.12.005.

PREV Non-invasive prediction of HER-2 overexpression and low expression in NME-type breast cancer using multiparametric MRI radiomics combined with MRI features
NEXT A study on the correlation between quantitative parameters of DCE-MRI and the pathological grading and angiogenesis of extrahepatic cholangiocarcinoma
  


LINK


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