Share：
Share this content in WeChat
X
[Chinese][PDF]6801
• Special Focus •
The MRI-based 3D-ResNet101 deep learning model for predicting preoperative grading of gliomas: A multicenter study
LI Darui  HU Wanjun  LIU Guangyao  GAN Tiejun  MA Laiyang  ZHANG Jing 

Cite this article as: LI D R, HU W J, LIU G Y, et al. The MRI-based 3D-ResNet101 deep learning model for predicting preoperative grading of gliomas: A multicenter study[J]. Chin J Magn Reson Imaging, 2023, 14(5): 25-30. DOI:10.12015/issn.1674-8034.2023.05.006.


[Abstract] Objective The preoperative accurate and non-invasive prediction of glioma grading remains challenging. Developing a robust Residual Networks (ResNet) deep learning model based on conventional T2WI images to predict preoperative pathological grading of gliomas.Materials and Methods A retrospective analysis of preoperative T2WI images of 919 patients with pathologically confirmed glioma, of which 708 were data from patients enrolled at the Second Hospital of Lanzhou University from June 2014 to April 2021 and 211 were derived from The Cancer Imaging Archive (TCIA) database. The TCIA dataset was subdivided into a development set (n=135) and an independent test set (n=76). The data from the Second Hospital of Lanzhou University dataset and the TCIA development set were randomly split 7∶3 into a training set (n=590) and a test set (n=253) to construct a 3D-ResNet101 deep learning model based on T2WI images. After the training, the models were validated on the test set and independent test set, where the model efficacy was assessed by macro F1 scores, accuracy (ACC), and receiver operating characteristic (ROC) curves.Results The 3D-ResNet101 deep learning model constructed based on T2WI had ACCs of 99% and 95% in the training and test sets, respectively; The F1 scores were 99% and 95%, respectively; the area under the ROC curve (AUC) were 0.98 and 0.97, respectively; the ACC of the independent test set was 83%, the F1 score was 83%, and the AUC was 0.89.Conclusions The 3D-ResNet101 deep learning model based on T2WI images predicts high- and low-grade gliomas with high accuracy and robustness. The method can be used for the non-invasive prediction of preoperative glioma grading as well as helping to improve the effectiveness of clinical management of patients.
[Keywords] glioma；3D-Residual Networks；deep learning；magnetic resonance imaging；T2 weighted imaging

LI Darui1, 2, 3   HU Wanjun1, 2   LIU Guangyao1, 2   GAN Tiejun1, 2   MA Laiyang1, 2, 3   ZHANG Jing1, 2*  

1 Department of Magnetic Resonance, Lanzhou University Second Hospital, Lanzhou 730030, China

2 Gansu Province Clinical Research Center for Functional and Molecular Imaging, Lanzhou 730030, China

3 Second Clinical School, Lanzhou University, Lanzhou 730030, China

Corresponding author: Zhang J, E-mail: lztong2001@163.com

Conflicts of interest   None.

ACKNOWLEDGMENTS Gansu Health Industry Scientific Research Project (No. GSWSKY2020-68); Lanzhou University Second Hospital "Cuiying Science and Technology Innovation" Program (No. CY2021-BJ-A05).
Received  2023-01-04
Accepted  2023-05-06
DOI: 10.12015/issn.1674-8034.2023.05.006
Cite this article as: LI D R, HU W J, LIU G Y, et al. The MRI-based 3D-ResNet101 deep learning model for predicting preoperative grading of gliomas: A multicenter study[J]. Chin J Magn Reson Imaging, 2023, 14(5): 25-30. DOI:10.12015/issn.1674-8034.2023.05.006.

[1]
JIANG T, NAM D H, RAM Z, et al. Clinical practice guidelines for the management of adult diffuse gliomas[J]. Cancer Lett, 2021, 499: 60-72. DOI: 10.1016/j.canlet.2020.10.050">10.1016/j.canlet.2020.10.050">10.1016/j.canlet.2020.10.050.
[2]
LOUIS D N, PERRY A, REIFENBERGER G, et al. The 2016 World Health Organization Classification of Tumors of the Central Nervous System: a summary[J]. Acta Neuropathol, 2016, 131(6): 803-820. DOI: 10.1007/s00401-016-1545-1">10.1007/s00401-016-1545-1">10.1007/s00401-016-1545-1.
[3]
SU C Q, LU S S, HAN Q Y, et al. Intergrating conventional MRI, texture analysis of dynamic contrast-enhanced MRI, and susceptibility weighted imaging for glioma grading[J]. Acta Radiol, 2019, 60(6): 777-787. DOI: 10.1177/0284185118801127">10.1177/0284185118801127">10.1177/0284185118801127.
[4]
ZHANG B, XUE C Q, LIN X Q, et al. Advances in deep learning for brain glioma imaging[J]. Chinese Journal of Medical Physics, 2021, 38(8): 1048-1052. DOI: 10.3969/j.issn.1005-202X.2021.08.025">10.3969/j.issn.1005-202X.2021.08.025">10.3969/j.issn.1005-202X.2021.08.025.
[5]
SHIN I, KIM H, AHN S S, et al. Development and Validation of a Deep Learning-Based Model to Distinguish Glioblastoma from Solitary Brain Metastasis Using Conventional MR Images[J]. AJNR Am J Neuroradiol, 2021, 42(5): 838-844. DOI: 10.3174/ajnr.A7003">10.3174/ajnr.A7003">10.3174/ajnr.A7003.
[6]
AHAMMED MUNEER K V, RAJENDRAN V R, K P J. Glioma Tumor Grade Identification Using Artificial Intelligent Techniques[J]. J Med Syst, 2019, 43(5): 113. DOI: 10.1007/s10916-019-1228-2">10.1007/s10916-019-1228-2">10.1007/s10916-019-1228-2.
[7]
HE K, ZHANG X, REN S, et al. Deep residual learning for image recognition[C]. Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, 2016: 770-778.
[8]
KORFIATIS P, KLINE T L, LACHANCE D H, et al. Residual Deep Convolutional Neural Network Predicts MGMT Methylation Status[J]. J Digit Imaging, 2017, 30(5): 622-628. DOI: 10.1007/s10278-017-0009-z">10.1007/s10278-017-0009-z">10.1007/s10278-017-0009-z.
[9]
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">10.1007/s10278-013-9622-7">10.1007/s10278-013-9622-7.
[10]
CHOI Y S, BAE S, CHANG J H, et al. Fully automated hybrid approach to predict the IDH mutation status of gliomas via deep learning and radiomics[J]. Neuro Oncol, 2021, 23(2): 304-313. DOI: 10.1093/neuonc/noaa177">10.1093/neuonc/noaa177">10.1093/neuonc/noaa177.
[11]
ZHA C, MENG X, LI L, et al. Neutrophil extracellular traps mediate the crosstalk between glioma progression and the tumor microenvironment via the HMGB1/RAGE/IL-8 axis[J]. Cancer Biol Med, 2020, 17(1): 154-168. DOI: 10.20892/j.issn.2095-3941.2019.0353">10.20892/j.issn.2095-3941.2019.0353">10.20892/j.issn.2095-3941.2019.0353.
[12]
MZOUGHI H, NJEH I, WALI A, et al. Deep Multi-Scale 3D Convolutional Neural Network (CNN) for MRI Gliomas Brain Tumor Classification[J]. J Digit Imaging, 2020, 33(4): 903-915. DOI: 10.1007/s10278-020-00347-9">10.1007/s10278-020-00347-9">10.1007/s10278-020-00347-9.
[13]
SINGH G, MANJILA S, SAKLA N, et al. Radiomics and radiogenomics in gliomas: a contemporary update[J]. Br J Cancer, 2021, 125(5): 641-657. DOI: 10.1038/s41416-021-01387-w">10.1038/s41416-021-01387-w">10.1038/s41416-021-01387-w.
[14]
ZHUGE Y, NING H, MATHEN P, et al. Automated glioma grading on conventional MRI images using deep convolutional neural networks[J]. Med Phys, 2020, 47(7): 3044-3053. DOI: 10.1002/mp.14168">10.1002/mp.14168">10.1002/mp.14168.
[15]
YANG Y, YAN L F, ZHANG X, et al. Glioma Grading on Conventional MR Images: A Deep Learning Study With Transfer Learning[J]. Front Neurosci, 2018, 12: 804. DOI: 10.3389/fnins.2018.00804">10.3389/fnins.2018.00804">10.3389/fnins.2018.00804.
[16]
LI Y, WEI D, LIU X, et al. Molecular subtyping of diffuse gliomas using magnetic resonance imaging: comparison and correlation between radiomics and deep learning[J]. Eur Radiol, 2022, 32(2): 747-758. DOI: 10.1007/s00330-021-08237-6">10.1007/s00330-021-08237-6">10.1007/s00330-021-08237-6.
[17]
KHAWALDEH S, PERVAIZ U, RAFIQ A, et al. Noninvasive grading of glioma tumor using magnetic resonance imaging with convolutional neural networks[J]. Applied Sciences, 2017, 8(1): 27. DOI: 10.3390/app8010027">10.3390/app8010027">10.3390/app8010027.
[18]
ARTZI M, GERSHOV S, BEN-SIRA L, et al. Automatic segmentation, classification, and follow-up of optic pathway gliomas using deep learning and fuzzy c-means clustering based on MRI[J]. Med Phys, 2020, 47(11): 5693-5701. DOI: 10.1002/mp.14489">10.1002/mp.14489">10.1002/mp.14489.
[19]
BANGALORE YOGANANDA C G, SHAH B R, VEJDANI-JAHROMI M, et al. A novel fully automated MRI-based deep-learning method for classification of IDH mutation status in brain gliomas[J]. Neuro Oncol, 2020, 22(3): 402-411. DOI: 10.1093/neuonc/noz199">10.1093/neuonc/noz199">10.1093/neuonc/noz199.
[20]
YU J, YANG B, WANG J, et al. 2D CNN versus 3D CNN for false-positive reduction in lung cancer screening[J]. J Med Imaging (Bellingham), 2020, 7(5): 051202. DOI: 10.1117/1.JMI.7.5.051202">10.1117/1.JMI.7.5.051202">10.1117/1.JMI.7.5.051202.
[21]
FAN F X, HU W J, JIANG Y L, et al. The value of 3D convolution neural network based on multimodal MRI images in the classification of liver fibrosis[J]. Chin J Magn Reson Imaging, 2022, 13(9): 30-34. DOI: 10.12015/issn.1674-8034.2022.09.006">10.12015/issn.1674-8034.2022.09.006">10.12015/issn.1674-8034.2022.09.006.
[22]
ANDRONESI O C, BHATTACHARYYA P K, BOGNER W, et al. Motion correction methods for MRS: experts' consensus recommendations[J/OL]. NMR Biomed, 2021, 34(5): e4364 [2023-01-03]. https://pubmed.ncbi.nlm.nih.gov/33089547/. DOI: 10.1002/nbm.4364">10.1002/nbm.4364">10.1002/nbm.4364.
[23]
SHAKIR T M, FENGLI L, CHENGUANG G, et al. 1H-MR spectroscopy in grading of cerebral glioma: A new view point, MRS image quality assessment[J/OL]. Acta Radiol Open, 2022, 11(2): 20584601221077068 [2023-01-03]. https://pubmed.ncbi.nlm.nih.gov/35237448/. DOI: 10.1177/20584601221077068">10.1177/20584601221077068">10.1177/20584601221077068.
[24]
MEI Z, BI J Y. Comparison of value of three MRI perfusion techniques in the preoperative assessment of brain glioma grading[J]. Chin J Magn Reson Imaging, 2022, 13(2): 83-86, DOI: 10.12015/issn.1674-8034.2022.02.017">10.12015/issn.1674-8034.2022.02.017">10.12015/issn.1674-8034.2022.02.017.
[25]
ZHANG J, LIU H, TONG H, et al. Clinical Applications of. Contrast-Enhanced Perfusion MRI Techniques in Gliomas: Recent Advances and Current Challenges[J]. Contrast Media Mol Imaging, 2017, 2017: 7064120. DOI: 10.1155/2017/7064120">10.1155/2017/7064120">10.1155/2017/7064120.
[26]
RICHTER H, BUCKER P, DUNKER C, et al. Gadolinium deposition in the brain of dogs after multiple intravenous administrations of linear gadolinium based contrast agents[J/OL]. PLoS One, 2020, 15(2): e0227649 [2023-01-03]. https://pubmed.ncbi.nlm.nih.gov/32012163/. DOI: 10.1371/journal.pone.0227649">10.1371/journal.pone.0227649">10.1371/journal.pone.0227649.
[27]
MATHUR M, JONES J R, WEINREB J C. Gadolinium Deposition and Nephrogenic Systemic Fibrosis: A Radiologist's Primer[J]. Radiographics, 2020, 40(1): 153-162. DOI: 10.1148/rg.2020190110">10.1148/rg.2020190110">10.1148/rg.2020190110.
[28]
TALO M, YILDIRIM O, BALOGLU U B, et al. Convolutional neural networks for multi-class brain disease detection using MRI images[J]. Comput Med Imaging Graph, 2019, 78: 101673. DOI: 10.1016/j.compmedimag.2019.101673">10.1016/j.compmedimag.2019.101673">10.1016/j.compmedimag.2019.101673.
[29]
ALBADAWY E A, SAHA A, MAZUROWSKI M A. Deep learning for segmentation of brain tumors: Impact of cross-institutional training and testing[J]. Med Phys, 2018, 45(3): 1150-1158. DOI: 10.1002/mp.12752">10.1002/mp.12752">10.1002/mp.12752.
[30]
BALACHANDAR N, CHANG K, KALPATHY-CRAMER J, et al. Accounting for data variability in multi-institutional distributed deep learning for medical imaging[J]. J Am Med Inform Assoc, 2020, 27(5): 700-708. DOI: 10.1093/jamia/ocaa017">10.1093/jamia/ocaa017">10.1093/jamia/ocaa017.

PREV A comparative study on the enhancement of high resolution coronal image quality by deep learning reconstruction
NEXT Clinical feasibility of breath-hold fat-suppressed T2-weighted sequence with deep learning reconstruction for liver imaging
  


LINK


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