Share：
Share this content in WeChat
X
[Chinese] [PDF] 5 1
• Clinical Articles •
The value of deep learning models based on multiparameter MRI in the preoperative prediction of tumor deposits in rectal cancer
HAN Shi  SUN Yancong  NIU Yongchao  XIE Beichen  WANG Shuaina  CHAI Yaxin  DUAN Jinhui  WANG Hongpo  CHENG Yujie  YAN Ruifang 

Cite this article as: HAN S, SUN Y C, NIU Y C, et al. The value of deep learning models based on multiparameter MRI in the preoperative prediction of tumor deposits in rectal cancer[J]. Chin J Magn Reson Imaging, 2026, 17(3): 46-53. DOI:10.12015/issn.1674-8034.2026.03.007.


[Abstract] Objective To investigate the value of deep learning (DL) models based on multiparametric MRI, including diffusion-weighted imaging (DWI), fat-suppressed T2-weighted imaging (T2-FS), and T1-weighted contrast-enhanced imaging (T1CE), in preoperatively predicting tumor deposit (TD) status in rectal cancer.Materials andMethods A retrospective analysis was conducted on 321 patients from two centers who underwent total mesorectal excision and were pathologically diagnosed with rectal adenocarcinoma. Patients were divided into a TD-positive group (n = 81) and a TD-negative group (n = 240) based on pathology. Patients from center 1 (n = 273) were randomly split 8∶2 into a training set and a test set, while patients from center 2 (n = 48) served as an external validation set. Using the ResNet18 DL network, four models were built: a DWI-DL model, a T2-FS-DL model, a T1CE-DL model, and a combined-DL model. Receiver operating characteristic (ROC) curves were plotted, and the area under the curve (AUC) was calculated to evaluate the predictive performance of each DL model.Results In the single-sequence models, the T1CE-DL model achieved AUCs of 0.842 (95% CI: 0.808 to 0.891), 0.792 (95% CI: 0.752 to 0.828), and 0.747 (95% CI: 0.700 to 0.777) in the training, test, and external validation sets, respectively, demonstrating superior predictive performance compared to the T2-FS-DL and DWI-DL models. The T2-FS-DL model yielded AUCs of 0.805 (95% CI: 0.774 to 0.843), 0.766 (95% CI: 0.724 to 0.801), and 0.725 (95% CI: 0.690 to 0.767) in the three datasets, respectively. For the DWI-DL model, the AUCs were 0.801 (95% CI: 0.753 to 0.832), 0.745 (95% CI: 0.703 to 0.775), and 0.747 (95% CI: 0.702 to 0.779), respectively. The combined-DL model achieved the highest AUCs, reaching 0.909 (95% CI: 0.877 to 0.956), 0.875 (95% CI: 0.834 to 0.919), and 0.816 (95% CI: 0.767 to 0.852) in the training test, and external validation sets, respectively. Its diagnostic performance was significantly superior to that of the three single-sequence models (DeLong test, all P < 0.05).Conclusions The combined-DL model shows good predictive value and generalization ability for preoperatively assessing TD status in rectal cancer patients.
[Keywords] rectal cancer；deep learning；multiparametric magnetic resonance imaging；tumor deposit；prediction model；external validation；interpretability

HAN Shi1   SUN Yancong1   NIU Yongchao2   XIE Beichen1   WANG Shuaina1   CHAI Yaxin2   DUAN Jinhui1   WANG Hongpo1   CHENG Yujie1   YAN Ruifang1*  

1 Department of Magnetic Resonance, the First Affiliated Hospital of Henan Medical University, Xinxiang 453100, China

2 Department of Magnetic Resonance, Xinxiang Central Hospital, Xinxiang 453000, China

Corresponding author: YAN R F, E-mail: yrf718@163.com

Conflicts of interest   None.

Received  2025-10-29
Accepted  2026-03-05
DOI: 10.12015/issn.1674-8034.2026.03.007
Cite this article as: HAN S, SUN Y C, NIU Y C, et al. The value of deep learning models based on multiparameter MRI in the preoperative prediction of tumor deposits in rectal cancer[J]. Chin J Magn Reson Imaging, 2026, 17(3): 46-53. DOI:10.12015/issn.1674-8034.2026.03.007.

[1]
GLYNNE-JONES R, WYRWICZ L, TIRET E, et al. Rectal cancer: ESMO Clinical Practice Guidelines for diagnosis, treatment and follow-up[J/OL]. Ann Oncol, 2017, 28(suppl_4): iv22-iv40 [2025-11-04] https://pubmed.ncbi.nlm.nih.gov/29741565/. DOI: 10.1093/annonc/mdx224.
[2]
MORGAN E, ARNOLD M, GINI A, et al. Global burden of colorectal cancer in 2020 and 2040: incidence and mortality estimates from GLOBOCAN[J]. Gut, 2023, 72(2): 338-344. DOI: 10.1136/gutjnl-2022-327736.
[3]
WEISER M R. AJCC 8th edition: colorectal cancer[J]. Ann Surg Oncol, 2018, 25(6): 1454-1455. DOI: 10.1245/s10434-018-6462-1.
[4]
MARJASUO S T, LEHTIMÄKI T E, KOSKENVUO L E, et al. Impact of mesorectal extranodal tumor deposits in magnetic resonance imaging on outcome of rectal cancer patients[J/OL]. Eur J Surg Oncol, 2024, 50(6): 108337 [2025-11-04] https://pubmed.ncbi.nlm.nih.gov/38657373/. DOI: 10.1016/j.ejso.2024.108337.
[5]
XIAN Z Y, SONG Y W, ZHANG Z J, et al. Combining pathological risk factors and T, N staging to optimize the assessment for risk stratification and prognostication in low-risk stage Ⅲ colon cancer[J/OL]. World J Surg Oncol, 2024, 22(1): 10 [2025-11-04] https://pubmed.ncbi.nlm.nih.gov/38178080/. DOI: 10.1186/s12957-023-03299-w.
[6]
DELATTRE J F, COHEN R, HENRIQUES J, et al. Prognostic value of tumor deposits for disease-free survival in patients with stage III colon cancer: a post hoc analysis of the IDEA France phase III trial (PRODIGE-GERCOR)[J]. J Clin Oncol, 2020, 38(15): 1702-1710. DOI: 10.1200/JCO.19.01960.
[7]
BASNET S, LOU Q F, LIU N, et al. Tumor deposit is an independent prognostic indicator in patients who underwent radical resection for colorectal cancer[J]. J Cancer, 2018, 9(21): 3979-3985. DOI: 10.7150/jca.27475.
[8]
ZHENG H, ZHANG J X, LIU Y C, et al. Prognostic value of tumor deposits in locally advanced rectal cancer: a retrospective study with propensity score matching[J]. Int J Clin Oncol, 2021, 26(6): 1109-1119. DOI: 10.1007/s10147-021-01885-0.
[9]
UENO H, MOCHIZUKI H, SHIROUZU K, et al. Multicenter study for optimal categorization of extramural tumor deposits for colorectal cancer staging[J]. Ann Surg, 2012, 255(4): 739-746. DOI: 10.1097/SLA.0b013e31824b4839.
[10]
NAGTEGAAL I D, KNIJN N, HUGEN N, et al. Tumor deposits in colorectal cancer: improving the value of modern staging-a systematic review and meta-analysis[J]. J Clin Oncol, 2017, 35(10): 1119-1127. DOI: 10.1200/JCO.2016.68.9091.
[11]
DIAGNOSIS AND TREATMENT GUIDELINES FOR COLORECTAL CANCER WORKING GROUP C S O C O C. Chinese Society of Clinical Oncology (CSCO) diagnosis and treatment guidelines for colorectal cancer 2018 (English version)[J]. Chin J Cancer Res, 2019, 31(1): 117-134. DOI: 10.21147/j.issn.1000-9604.2019.01.07.
[12]
BAI R, TAN Y N, LI D, et al. Development and validation of a novel prognostic nomogram including tumor deposits could better predict survival for colorectal cancer: a population-based study[J/OL]. Ann Transl Med, 2021, 9(8): 620 [2025-11-04] https://pubmed.ncbi.nlm.nih.gov/33987318/. DOI: 10.21037/atm-20-4728.
[13]
COHEN R, SHI Q, MEYERS J, et al. Combining tumor deposits with the number of lymph node metastases to improve the prognostic accuracy in stage Ⅲ colon cancer: a post hoc analysis of the CALGB/SWOG 80702 phase III study (Alliance) [J]. Ann Oncol, 2021, 32(10): 1267-1275. DOI: 10.1016/j.annonc.2021.07.009.
[14]
KARBHARI A, BAHETI A D, ANKATHI S K, et al. MRI in rectal cancer patients on 'watch and wait': patterns of response and their evolution[J]. Abdom Radiol (NY), 2023, 48(11): 3287-3296. DOI: 10.1007/s00261-023-04003-y.
[15]
KE J, JIN C, TANG J H, et al. A longitudinal MRI-based artificial intelligence system to predict pathological complete response after neoadjuvant therapy in rectal cancer: A multicenter validation study[J/OL]. Dis Colon Rectum, 2023, 66(12): e1195-e1206 [2025-11-04] https://pubmed.ncbi.nlm.nih.gov/37682775/. DOI: 10.1097/DCR.0000000000002931.
[16]
KIKANO E G, MATALON S A, ESKIAN M, et al. Concordance of MRI with pathology for primary staging of rectal cancer in routine clinical practice: a single institution experience[J]. Curr Probl Diagn Radiol, 2024, 53(1): 68-72. DOI: 10.1067/j.cpradiol.2023.08.016.
[17]
FERNANDES M C, GOLLUB M J, BROWN G. The importance of MRI for rectal cancer evaluation[J/OL]. Surg Oncol, 2022, 43: 101739 [2025-11-04] https://pubmed.ncbi.nlm.nih.gov/35339339/. DOI: 10.1016/j.suronc.2022.101739.
[18]
LORD A C, D'SOUZA N, SHAW A, et al. MRI-diagnosed tumor deposits and EMVI status have superior prognostic accuracy to current clinical TNM staging in rectal cancer[J]. Ann Surg, 2022, 276(2): 334-344. DOI: 10.1097/SLA.0000000000004499.
[19]
LORD A C, MORAN B, ABULAFI M, et al. Can extranodal tumour deposits be diagnosed on MRI Protocol for a multicentre clinical trial (the COMET trial)[J/OL]. BMJ Open, 2020, 10(10): e033395 [2025-11-04] https://pubmed.ncbi.nlm.nih.gov/33033006/. DOI: 10.1136/bmjopen-2019-033395.
[20]
ZHAO W, XU H, ZHAO R, et al. MRI-based radiomics model for preoperative prediction of lateral pelvic lymph node metastasis in locally advanced rectal cancer[J]. Acad Radiol, 2024, 31(7): 2753-2772. DOI: 10.1016/j.acra.2023.07.016.
[21]
FU C L, SHAO T T, HOU M, et al. Preoperative prediction of tumor deposits in rectal cancer with clinical-magnetic resonance deep learning-based radiomic models[J/OL]. Front Oncol, 2023, 13: 1078863 [2025-11-04] https://pubmed.ncbi.nlm.nih.gov/36890815/. DOI: 10.3389/fonc.2023.1078863.
[22]
YANG R, ZHAO H X, WANG X X, et al. Magnetic resonance imaging radiomics modeling predicts tumor deposits and prognosis in stage T3 lymph node positive rectal cancer[J]. Abdom Radiol (NY), 2023, 48(4): 1268-1279. DOI: 10.1007/s00261-023-03825-0.
[23]
FENG F W, LIU Y Q, BAO J Y, et al. Multiregional-based magnetic resonance imaging radiomics model for predicting tumor deposits in resectable rectal cancer[J]. Abdom Radiol (NY), 2023, 48(11): 3310-3321. DOI: 10.1007/s00261-023-04013-w.
[24]
SUN Z, XIA F, LV W Z, et al. Radiomics based on T2-weighted and diffusion-weighted MR imaging for preoperative prediction of tumor deposits in rectal cancer[J]. Am J Surg, 2024, 232: 59-67. DOI: 10.1016/j.amjsurg.2024.01.002.
[25]
WU S K, WANG N, AO W Q, et al. Deep learning-based multi-parametric magnetic resonance imaging (mp-MRI) nomogram for predicting Ki-67 expression in rectal cancer[J]. Abdom Radiol (NY), 2024, 49(9): 3003-3014. DOI: 10.1007/s00261-024-04232-9.
[26]
WAN L J, HU J S, CHEN S, et al. Prediction of lymph node metastasis in stage T1-2 rectal cancers with MRI-based deep learning[J]. Eur Radiol, 2023, 33(5): 3638-3646. DOI: 10.1007/s00330-023-09450-1.
[27]
WEI Y R, WANG H J, CHEN Z W, et al. Deep learning-based multiparametric MRI model for preoperative T-stage in rectal cancer[J]. J Magn Reson Imaging, 2024, 59(3): 1083-1092. DOI: 10.1002/jmri.28856.
[28]
HU H B, ZHAO S, JIANG H, et al. Deep learning based on multiparametric magnetic resonance imaging features to predict BRAF gene mutation status in rectal cancer patients[J]. Chin J Magn Reson Imaging, 2025, 16(1): 22-28. DOI: 10.12015/issn.1674-8034.2025.01.004.
[29]
HOSNY A, PARMAR C, QUACKENBUSH J, et al. Artificial intelligence in radiology[J]. Nat Rev Cancer, 2018, 18(8): 500-510. DOI: 10.1038/s41568-018-0016-5.
[30]
MOON J Y, LEE M R, HA G W. Prognostic value of tumor deposits for long-term oncologic outcomes in patients with stage III colorectal cancer: a systematic review and meta-analysis[J]. Int J Colorectal Dis, 2022, 37(1): 141-151. DOI: 10.1007/s00384-021-04036-z.
[31]
JIN Y M, YIN H K, ZHANG H L, et al. Predicting tumor deposits in rectal cancer: a combined deep learning model using T2-MR imaging and clinical features[J/OL]. Insights Imaging, 2023, 14(1): 221 [2025-11-04] https://pubmed.ncbi.nlm.nih.gov/38117396/. DOI: 10.1186/s13244-023-01564-w.
[32]
AO W Q, WANG N, CHEN X, et al. Multiparametric MRI-based deep learning models for preoperative prediction of tumor deposits in rectal cancer and prognostic outcome[J]. Acad Radiol, 2025, 32(3): 1451-1464. DOI: 10.1016/j.acra.2024.10.004.
[33]
DIJKHOFF R A P, BEETS-TAN R G H, LAMBREGTS D M J, et al. Value of DCE-MRI for staging and response evaluation in rectal cancer: a systematic review[J]. Eur J Radiol, 2017, 95: 155-168. DOI: 10.1016/j.ejrad.2017.08.009.
[34]
LIU Y Y, XU W J, HAO D P. Research progress of quantitative parameters of DCE-MRI in evaluating biological behavior of soft tissue tumors[J]. Radiol Pract, 2021, 36(10): 1326-1329. DOI: 10.13609/j.cnki.1000-0313.2021.10.025.
[35]
XU Q Y, XU Y Y, WANG J, et al. Distinguishing mesorectal tumor deposits from metastatic lymph nodes by using diffusion-weighted and dynamic contrast-enhanced magnetic resonance imaging in rectal cancer[J]. Eur Radiol, 2023, 33(6): 4127-4137. DOI: 10.1007/s00330-022-09328-8.

PREV Intravoxel incoherent motion histogram parameters combined with clinical features to predict the response of neoadjuvant therapy in patients with rectal cancer
NEXT Diagnostic value of magnetic resonance diffusion-weighted imaging for T-staging between T1 and T2 of bladder cancer: A Meta-analysis
  


LINK


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