Share：
Share this content in WeChat
X
[Chinese][PDF]1851
• Special Focu •
A study on preoperative prediction of rectal cancer vascular invasion using MRI-based deep transfer learning radiomics
YUAN Quan  WU Shujian  FAN Lifang  ZHAI Jian 

Cite this article as: YUAN Q, WU S J, FAN L F, et al. A study on preoperative prediction of rectal cancer vascular invasion using MRI-based deep transfer learning radiomics[J]. Chin J Magn Reson Imaging, 2025, 16(1): 54-60. DOI:10.12015/issn.1674-8034.2025.01.009.


[Abstract] Objective To explore the application value of preoperative prediction of lymphovascular invasion (LVI) in rectal cancer patients using axial high-resolution T2WI and deep transfer learning radiomics.Materials and Methods A retrospective analysis was conducted on clinical and imaging data of 384 patients diagnosed with rectal cancer by postoperative pathology at Yijishan Hospital of Wannan Medical College from January 2018 to December 2023. Patients were divided into an LVI-positive group (81 cases) and an LVI-negative group (303 cases) based on pathological LVI status, and randomly assigned to a training group (n = 269) and a validation group (n = 115) in a 7∶3 ratio. The ResNet-34 model was used as the base model for deep transfer learning feature extraction. Deep transfer learning features and traditional radiomics features were extracted from the tumor body, and feature dimension reduction was performed using Spearman rank correlation and least absolute shrinkage and selection operator (LASSO) regression to eliminate redundant features and retain those with the highest predictive value. Six machine learning algorithms [adaptive boosting (AdaBoost), naïve Bayes (NB), elastic net (Enet), gradient boosting machine (GBM), neural networks (NN), and support vector machine (SVM)] were used to construct prediction models based on traditional radiomics features, deep transfer learning features, and combined features. Evaluate the diagnostic performance of each model using receiver operating characteristic (ROC) curves, which demonstrated the models' effectiveness.Results After dimension reduction through Spearman rank correlation and LASSO regression, 23 optimal features were selected, including 6 traditional radiomics features and 17 deep transfer learning features. All constructed models based on combined features model demonstrated a higher area under the curve (AUC) than those based on individual features alone. The AUCs for the training group were 0.956, 0.802, 0.879, 0.966, 0.973, and 0.944, respectively, and for the validation group, 0.924, 0.868, 0.901, 0.892, 0.817, and 0.905, respectively.Conclusions The model based on combined features demonstrates high efficacy in predicting LVI status in rectal cancer, aiding in preoperative individualized prediction and potentially improving patient prognosis.
[Keywords] rectal cancer；lymphovascular invasion；magnetic resonance imaging；deep transfer learning；radiomics

YUAN Quan1   WU Shujian1   FAN Lifang2   ZHAI Jian1*  

1 Department of Radiology, the First Affiliated Hospital of Wannan Medical College (Yijishan Hospital), Wuhu 241001, China

2 Department of Medical Imaging, Wannan Medical College, Wuhu 241002, China

Corresponding author: ZHAI J, E-mail: yjszhaij@126.com

Conflicts of interest   None.

Received  2024-05-06
Accepted  2024-10-10
DOI: 10.12015/issn.1674-8034.2025.01.009
Cite this article as: YUAN Q, WU S J, FAN L F, et al. A study on preoperative prediction of rectal cancer vascular invasion using MRI-based deep transfer learning radiomics[J]. Chin J Magn Reson Imaging, 2025, 16(1): 54-60. DOI:10.12015/issn.1674-8034.2025.01.009.

[1]
YOU J, YIN J D. Performances of whole tumor texture analysis based on MRI: predicting preoperative T stage of rectal carcinomas[J/OL]. Front Oncol, 2021, 11: 678441 [2024-03-20]. https://pubmed.ncbi.nlm.nih.gov/34414105/. DOI: 10.3389/fonc.2021.678441.
[2]
STANZIONE A, VERDE F, ROMEO V, et al. Radiomics and machine learning applications in rectal cancer: current update and future perspectives[J]. World J Gastroenterol, 2021, 27(32): 5306-5321. DOI: 10.3748/wjg.v27.i32.5306.
[3]
LI H, YUAN Y, CHEN X L, et al. Value of intravoxel incoherent motion for assessment of lymph node status and tumor response after chemoradiation therapy in locally advanced rectal cancer[J/OL]. Eur J Radiol, 2022, 146: 110106 [2024-03-20]. https://pubmed.ncbi.nlm.nih.gov/34922118/. DOI: 10.1016/j.ejrad.2021.110106.
[4]
REBECCA L S, Kimberly D M, STACEY A F, et al. Colorectal cancer statistics, 2017[J]. CA Cancer J Clin, 2017, 67(3): 177-193. DOI: 10.3322/caac.21395.
[5]
GE Y X, XU W B, WANG Z, et al. Prognostic value of CT radiomics in evaluating lymphovascular invasion in rectal cancer: diagnostic performance based on different volumes of interest[J]. J Xray Sci Technol, 2021, 29(4): 663-674. DOI: 10.3233/XST-210877.
[6]
XU H S, ZHAO W Y, GUO W B, et al. Prediction model combining clinical and MR data for diagnosis of lymph node metastasis in patients with rectal cancer[J]. J Magn Reson Imaging, 2021, 53(3): 874-883. DOI: 10.1002/jmri.27369.
[7]
YANG Y S, FENG F, QIU Y J, et al. High-resolution MRI-based radiomics analysis to predict lymph node metastasis and tumor deposits respectively in rectal cancer[J]. Abdom Radiol, 2021, 46(3): 873-884. DOI: 10.1007/s00261-020-02733-x.
[8]
GOLLUB M J, LAKHMAN Y, MCGINTY K, et al. Does gadolinium-based contrast material improve diagnostic accuracy of local invasion in rectal cancer MRI? A multireader study[J]. AJR Am J Roentgenol, 2015, 204(2): W160-W167. DOI: 10.2214/AJR.14.12599.
[9]
RYU J, EOM S, KIM H C, et al. Chest X-ray-based opportunistic screening of sarcopenia using deep learning[J]. J Cachexia Sarcopenia Muscle, 2023, 14(1): 418-428. DOI: 10.1002/jcsm.13144.
[10]
MZOUGHI H, NJEH I, SLIMA M BEN, et al. Deep efficient-nets with transfer learning assisted detection of COVID-19 using chest X-ray radiology imaging[J]. Multimed Tools Appl, 2023: 1-23. DOI: 10.1007/s11042-023-15097-3.
[11]
WEBB J M, ADUSEI S A, WANG Y N, et al. Comparing deep learning-based automatic segmentation of breast masses to expert interobserver variability in ultrasound imaging[J/OL]. Comput Biol Med, 2021, 139: 104966 [2024-03-20]. https://pubmed.ncbi.nlm.nih.gov/34715553/. DOI: 10.1016/j.compbiomed.2021.104966.
[12]
WANG X H, YU J Z, ZHU Q, et al. Potential of deep learning in assessing pneumoconiosis depicted on digital chest radiography[J]. Occup Environ Med, 2020, 77(9): 597-602. DOI: 10.1136/oemed-2019-106386.
[13]
LIN X, ZHAO S, JIANG H J, et al. A radiomics-based nomogram for preoperative T staging prediction of rectal cancer[J]. Abdom Radiol, 2021, 46(10): 4525-4535. DOI: 10.1007/s00261-021-03137-1.
[14]
JIANG X F, ZHAO H Y, SALDANHA O L, et al. An MRI deep learning model predicts outcome in rectal cancer[J/OL]. Radiology, 2023, 307(5): e222223 [2024-03-20]. https://pubmed.ncbi.nlm.nih.gov/37278629/. DOI: 10.1148/radiol.222223.
[15]
LIU X Y, ZHANG D F, LIU Z Y, et al. Deep learning radiomics-based prediction of distant metastasis in patients with locally advanced rectal cancer after neoadjuvant chemoradiotherapy: A multicentre study[J/OL]. EBioMedicine, 2021, 69: 103442 [2024-03-20]. https://pubmed.ncbi.nlm.nih.gov/34157487/. DOI: 10.1016/j.ebiom.2021.103442.
[16]
YANG L, DONG D, FANG M J, et al. Can CT-based radiomics signature predict KRAS/NRAS/BRAF mutations in colorectal cancer?[J]. Eur Radiol, 2018, 28(5): 2058-2067. DOI: 10.1007/s00330-017-5146-8.
[17]
MENG X C, XIA W, XIE P Y, et al. Preoperative radiomic signature based on multiparametric magnetic resonance imaging for noninvasive evaluation of biological characteristics in rectal cancer[J]. Eur Radiol, 2019, 29(6): 3200-3209. DOI: 10.1007/s00330-018-5763-x.
[18]
CALABRESE A, SANTUCCI D, LANDI R, et al. Radiomics MRI for lymph node status prediction in breast cancer patients: the state of art[J]. J Cancer Res Clin Oncol, 2021, 147(6): 1587-1597. DOI: 10.1007/s00432-021-03606-6.
[19]
RAUSEO E, IZQUIERDO MORCILLO C, RAISI-ESTABRAGH Z, et al. New imaging signatures of cardiac alterations in ischaemic heart disease and cerebrovascular disease using CMR radiomics[J/OL]. Front Cardiovasc Med, 2021, 8: 716577 [2024-03-20]. https://pubmed.ncbi.nlm.nih.gov/34631820/. DOI: 10.3389/fcvm.2021.716577.
[20]
SHIN J, SEO N, BAEK S E, et al. MRI radiomics model predicts pathologic complete response of rectal cancer following chemoradiotherapy[J]. Radiology, 2022, 303(2): 351-358. DOI: 10.1148/radiol.211986.
[21]
LI H, CHEN X L, LIU H, et al. MRI-based multiregional radiomics for preoperative prediction of tumor deposit and prognosis in resectable rectal cancer: a bicenter study[J]. Eur Radiol, 2023, 33(11): 7561-7572. DOI: 10.1007/s00330-023-09723-9.
[22]
LI M, ZHU Y Z, ZHANG Y C, et al. Radiomics of rectal cancer for predicting distant metastasis and overall survival[J]. World J Gastroenterol, 2020, 26(33): 5008-5021. DOI: 10.3748/wjg.v26.i33.5008.
[23]
LIU H H, YIN H K, LI J N, et al. A deep learning model based on MRI and clinical factors facilitates noninvasive evaluation of KRAS mutation in rectal cancer[J]. J Magn Reson Imaging, 2022, 56(6): 1659-1668. DOI: 10.1002/jmri.28237.
[24]
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.
[25]
LIU S Y, YU X P, YANG S H, et al. Machine learning-based radiomics nomogram for detecting extramural venous invasion in rectal cancer[J/OL]. Front Oncol, 2021, 11: 610338 [2024-03-20]. https://pubmed.ncbi.nlm.nih.gov/33842316/. DOI: 10.3389/fonc.2021.610338.
[26]
ZHANG W, YIN H K, HUANG Z X, et al. Development and validation of MRI-based deep learning models for prediction of microsatellite instability in rectal cancer[J]. Cancer Med, 2021, 10(12): 4164-4173. DOI: 10.1002/cam4.3957.
[27]
AHMED M, AHMED A. Palm tree disease detection and classification using residual network and transfer learning of inception ResNet[J]. PLoS One, 2023, 18(3): e0282250 [2024-03-20]. https://pubmed.ncbi.nlm.nih.gov/36862665/. DOI: 10.1371/journal.pone.0282250.
[28]
YANG X L, YE Q Y, CAI G F, et al. PD-ResNet for classification of Parkinson's disease from gait[J]. IEEE J Transl Eng Health Med, 2022, 10: 2200111 [2024-03-20]. https://pubmed.ncbi.nlm.nih.gov/35795875/. DOI: 10.1109/JTEHM.2022.3180933.
[29]
CEJUDO J E, CHAURASIA A, FELDBERG B, et al. Classification of dental radiographs using deep learning[J/OL]. J Clin Med, 2021, 10(7): 1496 [2024-03-20]. https://pubmed.ncbi.nlm.nih.gov/33916800/. DOI: 10.3390/jcm10071496.
[30]
LI M, JIN Y M, ZHANG Y C, et al. Radiomics for predicting perineural invasion status in rectal cancer[J]. World J Gastroenterol, 2021, 27(33): 5610-5621. DOI: 10.3748/wjg.v27.i33.5610.
[31]
LI J, ZHOU Y, WANG X X, et al. An MRI-based multi-objective radiomics model predicts lymph node status in patients with rectal cancer[J]. Abdom Radiol, 2021, 46(5): 1816-1824. DOI: 10.1007/s00261-020-02863-2.
[32]
JANG J H, KIM T Y, YOON D. Effectiveness of transfer learning for deep learning-based electrocardiogram analysis[J]. Healthc Inform Res, 2021, 27(1): 19-28. DOI: 10.4258/hir.2021.27.1.19.
[33]
LI M, JIN Y M, RUI J, et al. Computed tomography-based radiomics for predicting lymphovascular invasion in rectal cancer[J/OL]. Eur J Radiol, 2022, 146: 110065 [2024-03-20]. https://pubmed.ncbi.nlm.nih.gov/34844171/. DOI: 10.1016/j.ejrad.2021.110065.
[34]
WONG C, LIU T, ZHANG C Y, et al. Preoperative detection of lymphovascular invasion in rectal cancer using intravoxel incoherent motion imaging based on radiomics[J]. Med Phys, 2024, 51(1): 179-191. DOI: 10.1002/mp.16821.
[35]
ZHANG Y Y, HE K, GUO Y, et al. A novel multimodal radiomics model for preoperative prediction of lymphovascular invasion in rectal cancer[J/OL]. Front Oncol, 2020, 10: 457 [2024-03-20]. https://pubmed.ncbi.nlm.nih.gov/32328460/. DOI: 10.3389/fonc.2020.00457.

PREV Prediction of microsatellite instability in colorectal cancer based on MRI-ADC and clinicopathological features
NEXT The role of radiomics in predicting no disease progression after treatment of locally advanced rectal cancer
  


LINK


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