Share：
Share this content in WeChat
X
[Chinese] [PDF] 179 7
• Clinical Article •
Identification of benign and malignant vertebral compression fractures based on multiparameter MRI radiomics model
ZHONG Yi  YANG Haitao  XU Zhangyan  ZHU Tongxin  ZENG Wei  PU Yongliang  JIANG Lu 

DOI:10.12015/issn.1674-8034.2025.08.017.


[Abstract] Objective To evaluate a combined MRI radiomics and semantic features for distinguishing benign from malignant vertebral compression fractures (VCFs), and compare machine learning algorithms' performance.Materials and Method Retrospectively analyzed 449 VCFs patients (550 vertebrae) from First Affiliated Hospital of Chongqing Medical University (center 1) and Chongqing Yubei District Traditional Chinese Medicine Hospital (center 2). The patients of center 1 (229 patients: 103 benign, 126 malignant) were split 7∶3 into training set and internal validation set; The patients of center 2 (220 patients: 163 benign, 57 malignant) served as external validation set. Radiomics features from MRI sagittal sequences [T1WI, T2WI, T2WI- fat saturation (FS)] and semantic features were integrated. Models (clinical, radiomics, combined) were built using logistic regression (LR), random forest (RF), and support vector machine (SVM). The area under the curve (AUC) values of receiver operating characteristic (ROC) curves were compared across datasets.Result The combined model integrated two semantic features and four radiomics features. The AUC of the combined model constructed using LR algorithm in the training set, internal validation set, and external validation set were 0.957 [95% confidence interval (CI): 0.921 to 0.983], 0.936 (95% CI: 0.871 to 0.975) and 0.921 (95% CI: 0.872 to 0.960), respectively, which were significantly better than the clinical model and single sequence radiomics model (DeLong test P < 0.05). The combined model demonstrated good calibration performance and showed higher clinical net benefits in decision curve analysis.Conclusions The combined radiomics-semantic model with LR significantly enhances MRI diagnostic accuracy for VCFs, offering a reliable clinical tool.
[Keywords] benign vertebral compression fracture；malignant vertebral compression fracture；magnetic resonance imaging；radiomics；semantic features

ZHONG Yi1   YANG Haitao1*   XU Zhangyan1   ZHU Tongxin1   ZENG Wei1   PU Yongliang1   JIANG Lu2  

1 Department of Radiology, First Affiliated Hospital of Chongqing Medical University, Chongqing 400010, China

2 Department of Medical Imaging, Chongqing Yubei District Traditional Chinese Medicine Hospital, Chongqing 401120, China

Corresponding author: YANG H T, E-mail: frankyang119@126.com

Conflicts of interest   None.

Received  2025-01-22
Accepted  2025-08-06
DOI: 10.12015/issn.1674-8034.2025.08.017
DOI:10.12015/issn.1674-8034.2025.08.017.

[1]
CHEE C G, YOON M A, KIM K W, et al. Combined radiomics-clinical model to predict malignancy of vertebral compression fractures on CT[J]. Eur Radiol, 2021, 31(9): 6825-6834. DOI: 10.1007/s00330-021-07832-x.
[2]
SU C, TIAN Y X, HOU X Z, et al. Differential diagnosis based on MRI imaging radiomics meta-analysis of benign and malignant vertebral compression fractures[J]. J Clin Radiol, 2024, 43(10): 1771-1776. DOI: 10.13437/j.cnki.jcr.2024.10.009.
[3]
FENG Q Q, ZHENG X, HAN F G, et al. Value of MRI-based radiomics model in differentiation of benign and malignant vertebral compression fractures[J]. Chin Comput Med Imag, 2023, 29(1): 62-68. DOI: 10.19627/j.cnki.cn31-1700/th.2023.01.022.
[4]
Chinese Rehabilitation Medicine Association Osteoporosis Prevention and Rehabilitation Professional Committee, YANG H L, MAO H Q, et al. Expert consensus on diagnosis and treatment of osteoporotic vertebral compression fracture (2021 edition)[J]. Natl Med J China, 2021, 101(41): 3371-3379. DOI: 10.3760/cma.j.cn112137-20210625-01436.
[5]
COLEMAN R E, CROUCHER P I, PADHANI A R, et al. Bone metastases[J/OL]. Nat Rev Dis Primers, 2020, 6: 83 [2025-01-21]. https://pubmed.ncbi.nlm.nih.gov/33060614/. DOI: 10.1038/s41572-020-00216-3.
[6]
MUSA AGUIAR P, ZARANTONELLO P, APARISI GÓMEZ M P. Differentiation between osteoporotic and neoplastic vertebral fractures: state of the art and future perspectives[J]. Curr Med Imaging, 2022, 18(2): 187-207. DOI: 10.2174/1573405617666210412142758.
[7]
CHEN Y M, DONG J G. Analysis of MRI signal characteristics and clinical application value of thoracolumbar compression fracture[J]. Chin J CT MRI, 2019, 17(2): 130-132. DOI: 10.3969/j.issn.1672-5131.2019.02.040.
[8]
SCHWAIGER B J, GERSING A S, HAMMEL J, et al. Three-material decomposition with dual-layer spectral CT compared to MRI for the detection of bone marrow edema in patients with acute vertebral fractures[J]. Skeletal Radiol, 2018, 47(11): 1533-1540. DOI: 10.1007/s00256-018-2981-x.
[9]
ZHONG Y, LIU X, XIAO Y D, et al. Research progress of medical image texture analysis in musculoskeletal diseases[J]. Chin J Magn Reson Imag, 2020, 11(5): 394-397. DOI: 10.12015/issn.1674-8034.2020.05.018.
[10]
WANG H, XU J Y. Research progress of CT imaging in evaluating the therapeutic effect of non-small cell lung cancer[J]. J Clin Radiol, 2024, 43(1): 149-151. DOI: 10.13437/j.cnki.jcr.2024.01.011.
[11]
LIU K, WANG Q Z, CHEN Y Y, et al. Identification of benign and malignant spinal fractures based on the ResNet50 deep learning model[J]. J Clin Radiol, 2021, 40(12): 2350-2355. DOI: 10.13437/j.cnki.jcr.2021.12.021.
[12]
FOREMAN S C, SCHINZ D, HUSSEINI M EL, et al. Deep learning to differentiate benign and malignant vertebral fractures at multidetector CT[J/OL]. Radiology, 2024, 310(3): e231429 [2025-01-21]. https://pubmed.ncbi.nlm.nih.gov/38530172/. DOI: 10.1148/radiol.231429.
[13]
VAN GRIETHUYSEN J J M, FEDOROV A, PARMAR C, et al. Computational radiomics system to decode the radiographic phenotype[J/OL]. Cancer Res, 2017, 77(21): e104-e107 [2025-01-21]. https://pubmed.ncbi.nlm.nih.gov/38530172/. DOI: 10.1158/0008-5472.CAN-17-0339.
[14]
MA T C, GUO Y, REN S, et al. Application of radiomics in diagnosis and treatment of vertebral fracture[J]. Chin J Magn Reson Imag, 2024, 15(10): 228-233. DOI: 10.12015/issn.1674-8034.2024.10.039.
[15]
ZHENG Y R. Boruta feature selection algorithm and its application[D]. Xi'an: University of Finance and Economics, 2020. DOI: 10.27706/d.cnki.gxacj.2020.000024.
[16]
XIAO M Q, ZHANG M, LIU J F, et al. Application of DWI in differential diagnosis of benign and malignant vertebral compression fractures[J]. Chin Imag J Integr Tradit West Med, 2015, 13(3): 303-305. DOI: 10.3969/j.issn.1672-0512.2015.03.025.
[17]
QU B, SHI X M, FENG C M, et al. The value of CT and MRI in differentiating benign from malignant vertebral compression fractures and the study of their imaging features[J]. Chin J CT MRI, 2023, 21(12): 164-166. DOI: 10.3969/j.issn.1672-5131.2023.12.050.
[18]
JONES A, BRAY T J P, MANDL P, et al. Performance of magnetic resonance imaging in the diagnosis of axial spondyloarthritis: a systematic literature review[J]. Rheumatology (Oxford), 2019, 58(11): 1955-1965. DOI: 10.1093/rheumatology/kez172.
[19]
YAMAMOTO Y, IWATA E, SHIGEMATSU H, et al. Differential diagnosis between metastatic and osteoporotic vertebral fractures using sagittal T1-weighted magnetic resonance imaging[J]. J Orthop Sci, 2020, 25(5): 763-769. DOI: 10.1016/j.jos.2019.10.004.
[20]
ZHANG J. Application value of MRI liquid sign in diagnosis and differential diagnosis of benign and malignant vertebral compression fractures[J]. Chin J Mod Drug Appl, 2020, 14(7): 53-55. DOI: 10.14164/j.cnki.cn11-5581/r.2020.07.023.
[21]
XU C J, KONG L L, ZHAO C Y, et al. The application value of dual-energy CT virtual monoenergetic imaging in differentiating spinal benign and malignant compression fracture[J]. Chin J CT MRI, 2022, 20(12): 150-152. DOI: 10.3969/j.issn.1672-5131.2022.12.057.
[22]
WANG H J, NI S J, XU Y S. The diagnostic value of X-ray, CT and MRI multimodal imaging in the vertebral compression fractures caused by osteolytic metastases and osteoporosis[J]. Chin J Postgrad Med, 2022(11): 997-1003. DOI: 10.3760/cma.j.cn115455-20220111-00023.
[23]
WU L, ZHU J J, SHEN L S, et al. MRI differential diagnosis of vertebral compression fracture caused by osteoporosis and bone metastasis in the elderly[J]. J Mudanjiang Med Univ, 2018, 39(3): 69-71, 59. DOI: 10.13799/j.cnki.mdjyxyxb.2018.03.021.
[24]
QI K J, ZHANG E L, WANG Q Z, et al. Progress of MRI in identifying benign and malignant vertebral compression fractures[J]. Chin J Magn Reson Imag, 2020, 11(11): 1077-1080. DOI: 10.12015/issn.1674-8034.2020.11.029.
[25]
PATIL S S, BHOJARAJ S Y, NENE A M. Safety and efficacy of spinal loop rectangle and sublaminar wires for osteoporotic vertebral compression fracture fixation[J]. Asian J Neurosurg, 2017, 12(3): 436-440. DOI: 10.4103/1793-5482.175648.
[26]
LEE D H, NAM J K, JUNG H S, et al. Does T1- and diffusion-weighted magnetic resonance imaging give value-added than bone scintigraphy in the follow-up of vertebral metastasis of prostate cancer[J]. Investig Clin Urol, 2017, 58(5): 324-330. DOI: 10.4111/icu.2017.58.5.324.
[27]
SHI Z J, HAN J K, QIN J, et al. Clinical application of diffusion-weighted imaging and dynamic contrast-enhanced MRI in assessing the clinical curative effect of early ankylosing spondylitis[J/OL]. Medicine (Baltimore), 2019, 98(20): e15227 [2025-01-21]. https://pubmed.ncbi.nlm.nih.gov/31096431/. DOI: 10.1097/MD.0000000000015227.
[28]
FRIGHETTO-PEREIRA L, RANGAYYAN R M, METZNER G A, et al. Shape, texture and statistical features for classification of benign and malignant vertebral compression fractures in magnetic resonance images[J]. Comput Biol Med, 2016, 73: 147-156. DOI: 10.1016/j.compbiomed.2016.04.006.
[29]
ZHANG H, YUAN G J, WANG C, et al. Differentiation of benign versus malignant indistinguishable vertebral compression fractures by different machine learning with MRI-based radiomic features[J]. Eur Radiol, 2023, 33(7): 5069-5076. DOI: 10.1007/s00330-023-09678-x.
[30]
GENG W, ZHU J F, LI M, et al. Radiomics Based on Multimodal magnetic resonance imaging for the Differential Diagnosis of Benign and Malignant Vertebral Compression Fractures[J]. Orthop Surg, 2024, 16(10): 2464-2474. DOI: 10.1111/os.14148.
[31]
LIU B B, JIN Y C, FENG S X, et al. Benign vs malignant vertebral compression fractures with MRI: a comparison between automatic deep learning network and radiologist's assessment[J]. Eur Radiol, 2023, 33(7): 5060-5068. DOI: 10.1007/s00330-023-09713-x.

PREV Analysis of the diagnostic efficacy of multi-sequence optimized VBQs and QCT for osteoporosis
NEXT Feasibility of reduced dose of <sup>18</sup>F-FDG during chest PET/MR examinations
  


LINK


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