Share：
Share this content in WeChat
X
[Chinese][PDF]1671
• Clinical Article •
Development and external validation of an XGBoost model for differentiating the benign and malignant nature of non-mass breast lesions
YANG Wen  YANG Wei  ZHOU Xiaoping  YANG Yan  ZHANG Ningmei  YIN Qingyun  ZHANG Chaolin  LIU Zhaodi 

Cite this article as: YANG W, YANG W, ZHOU X P, et al. Development and external validation of an XGBoost model for differentiating the benign and malignant nature of non-mass breast lesions[J]. Chin J Magn Reson Imaging, 2025, 16(1): 118-126, 145. DOI:10.12015/issn.1674-8034.2025.01.018.


[Abstract] Objective To develop an extreme gradient boosting (XGBoost) model based on clinical and imaging features to differentiate between benign and malignant non-mass breast lesions.Materials and Methods Data were collected from January 2018 to July 2024 from two institutions, focusing on 480 non-mass breast lesions with pathological results obtained from two types of mammography equipment. Patients were categorized into a modeling group [n = 310, digital mammography (DM) examination], an internal validation group (n = 108, DM examination), and an external validation group [n = 62, digital breast tomosynthesis (DBT) examination]. Preoperative breast X-ray (DM or DBT), MRI, and clinical characteristics were recorded. The XGBoost algorithm and multivariate logistic regression (LR) analysis were employed to develop the XGBoost and LR models, respectively. Diagnostic performance was assessed using receiver operating characteristic (ROC) curves.Results In the modeling group, patients were randomly split in a 7∶3 ratio into a training set (n = 217) and a test set (n = 93). The proportion of malignant non-mass lesions in the training set, test set, internal validation group of the training set, and external validation group of the training set, were 159 (73%), 58 (62%), 73 (68%) and 43 (69%), respectively. The XGBoost model outperformed the LR model in diagnostic accuracy, demonstrating superior performance across the independent training, test, and internal, external validation sets of the training set, with area under the curve (AUC) ranging from 0.884 to 0.913. Additionally, the XGBoost model exhibited good calibration and clinical net benefit in all four cohorts.Conclusions The XGBoost model accurately differentiates between benign and malignant non-mass breast lesions, indicating its potential for widespread clinical application.
[Keywords] non-mass enhancement；breast cancer；extreme gradient boosting；machine learning；magnetic resonance imaging；mammography

YANG Wen1   YANG Wei2*   ZHOU Xiaoping1   YANG Yan3   ZHANG Ningmei4   YIN Qingyun5   ZHANG Chaolin6   LIU Zhaodi7  

1 The First School of Clinical Medicine, Ningxia Medical University, Yinchuan 750004, China

2 Department of Radiology, General Hospital of Ningxia Medical University, Yinchuan 750004, China

3 Information Technology Center, 32752 Troop, Xiangyang 441000, China

4 Department of Pathology, General Hospital of Ningxia Medical University, Yinchuan 750004, China

5 Department of Medical Oncology, General Hospital of Ningxia Medical University, Yinchuan 750004, China

6 Department of Oncology Surgery, General Hospital of Ningxia Medical University, Yinchuan 750004, China

7 Department of Radiology, First People's Hospital of Shizuishan City, Shizuishan 753200, China

Corresponding author: YANG W, E-mail: yangwei_0521@163.com

Conflicts of interest   None.

Received  2024-09-13
Accepted  2025-01-10
DOI: 10.12015/issn.1674-8034.2025.01.018
Cite this article as: YANG W, YANG W, ZHOU X P, et al. Development and external validation of an XGBoost model for differentiating the benign and malignant nature of non-mass breast lesions[J]. Chin J Magn Reson Imaging, 2025, 16(1): 118-126, 145. DOI:10.12015/issn.1674-8034.2025.01.018.

[1]
WANG L J, CHANG L F, LUO R, et al. An artificial intelligence system using maximum intensity projection MR images facilitates classification of non-mass enhancement breast lesions[J]. Eur Radiol, 2022, 32(7): 4857-4867. DOI: 10.1007/s00330-022-08553-5.
[2]
LIU G, LI Y, CHEN S L, et al. Non-mass enhancement breast lesions: MRI findings and associations with malignancy[J/OL]. Ann Transl Med, 2022, 10(6): 357 [2024-09-12]. https://pubmed.ncbi.nlm.nih.gov/35433999/. DOI: 10.21037/atm-22-503.
[3]
WANG S Q, WANG H, LI Y, et al. The value of DCE- MRI of the breast as a diagnostic tool in assessing amorphous calcifications in screening mammography[J/OL]. Front Oncol, 2023, 13: 1151500 [2024-09-12]. https://pubmed.ncbi.nlm.nih.gov/37182168/. DOI: 10.3389/fonc.2023.1151500.
[4]
LUNKIEWICZ M, FORTE S, FREIWALD B, et al. Interobserver variability and likelihood of malignancy for fifth edition BI-RADS MRI descriptors in non-mass breast lesions[J]. Eur Radiol, 2020, 30(1): 77-86. DOI: 10.1007/s00330-019-06312-7.
[5]
MACHIDA Y, TOZAKI M, SHIMAUCHI A, et al. Two distinct types of linear distribution in nonmass enhancement at breast MR imaging: difference in positive predictive value between linear and branching patterns[J]. Radiology, 2015, 276(3): 686-694. DOI: 10.1148/radiol.2015141775.
[6]
CEN D Z, XU L, LI N N, et al. BI-RADS 3-5 microcalcifications can preoperatively predict breast cancer HER2 and Luminal a molecular subtype[J]. Oncotarget, 2017, 8(8): 13855-13862. DOI: 10.18632/oncotarget.14655.
[7]
HOWELL A, ANDERSON A S, CLARKE R B, et al. Risk determination and prevention of breast cancer[J/OL]. Breast Cancer Res, 2014, 16(5): 446 [2024-09-12]. https://pubmed.ncbi.nlm.nih.gov/25467785/. DOI: 10.1186/s13058-014-0446-2.
[8]
BENZEKRY S, MASTRI M, NICOLÒ C, et al. Machine-learning and mechanistic modeling of metastatic breast cancer after neoadjuvant treatment[J/OL]. PLoS Comput Biol, 2024, 20(5): e1012088 [2024-09-12]. https://pubmed.ncbi.nlm.nih.gov/38701089/. DOI: 10.1371/journal.pcbi.1012088.
[9]
MAO J J, LIU L L, SHEN Q, et al. Integrating single-cell transcriptomics and machine learning to predict breast cancer prognosis: a study based on natural killer cell-related genes[J/OL]. J Cell Mol Med, 2024, 28(15): e18549 [2024-09-12]. https://pubmed.ncbi.nlm.nih.gov/39098994/. DOI: 10.1111/jcmm.18549.
[10]
CLIFT A K, DODWELL D, LORD S, et al. Development and internal-external validation of statistical and machine learning models for breast cancer prognostication: cohort study[J/OL]. BMJ, 2023, 381: e073800 [2024-09-12]. https://pubmed.ncbi.nlm.nih.gov/37164379/. DOI: 10.1136/bmj-2022-073800.
[11]
ZHENG G Y, PENG J X, SHU Z Y, et al. Predicting pathological complete response to neoadjuvant chemotherapy in breast cancer patients: use of MRI radiomics data from three regions with multiple machine learning algorithms[J/OL]. J Cancer Res Clin Oncol, 2024, 150(3): 147 [2024-09-12]. https://pubmed.ncbi.nlm.nih.gov/38512406/. DOI: 10.1007/s00432-024-05680-y.
[12]
ZHOU X P, YANG W, YIN Q Y, et al. Combining the X-ray and MRI characteristics with the clinical pathology to predict ductal carcinoma in situ with microinvasion of breast[J]. Chin J Magn Reson Imag, 2024, 15(5): 102-110. DOI: 10.12015/issn.1674-8034.2024.05.017.
[13]
YIN L, WEI X, ZHANG Q, et al. Multimodal ultrasound assessment of mass and non-mass enhancements by MRI: Diagnostic accuracy in idiopathic granulomatous mastitis and breast cancer[J/OL]. Breast, 2024, 78: 103797 [2024-09-12]. https://pubmed.ncbi.nlm.nih.gov/39418768/. DOI: 10.1016/j.breast.2024.103797.
[14]
CHE S N, LI J, XUE M, et al. The value of synthetic MRI in differential diagnosis of benign and malignant breast lesions[J]. Chin J Oncol, 2021, 43(8): 872-877. DOI: 10.3760/cma.j.cn112152-20210322-00254.
[15]
SPAK D A, PLAXCO J S, SANTIAGO L, et al. BI-RADS® fifth edition: a summary of changes[J]. Diagn Interv Imaging, 2017, 98(3): 179-190. DOI: 10.1016/j.diii.2017.01.001.
[16]
WECSLER J, JEONG Y J, RAGHAVENDRA A S, et al. Factors associated with MRI detection of occult lesions in newly diagnosed breast cancers[J]. J Surg Oncol, 2020, 121(4): 589-598. DOI: 10.1002/jso.25855.
[17]
WATT G P, THAKRAN S, SUNG J S, et al. Association of breast cancer odds with background parenchymal enhancement quantified using a fully automated method at MRI: the IMAGINE study[J/OL]. Radiology, 2023, 308(3): e230367 [2024-09-12]. https://pubmed.ncbi.nlm.nih.gov/37750771/. DOI: 10.1148/radiol.230367.
[18]
YANG X P, DONG M S, LI S, et al. Diffusion-weighted imaging or dynamic contrast-enhanced curve: a retrospective analysis of contrast-enhanced magnetic resonance imaging-based differential diagnoses of benign and malignant breast lesions[J]. Eur Radiol, 2020, 30(9): 4795-4805. DOI: 10.1007/s00330-020-06883-w.
[19]
WANG Q, LIN Y Y, DING C, et al. Multi-modality radiomics model predicts axillary lymph node metastasis of breast cancer using MRI and mammography[J]. Eur Radiol, 2024, 34(9): 6121-6131. DOI: 10.1007/s00330-024-10638-2.
[20]
LIU K H, YANG W, TIAN H P, et al. Association between programmed cell death ligand-1 expression in patients with cervical cancer and apparent diffusion coefficient values: a promising tool for patient's immunotherapy selection[J]. Eur Radiol, 2024, 34(10): 6726-6737. DOI: 10.1007/s00330-024-10759-8.
[21]
KAYADIBI Y, SARACOGLU M S, KURT S A, et al. Differentiation of malignancy and idiopathic granulomatous mastitis presenting as non-mass lesions on MRI: radiological, clinical, radiomics, and clinical-radiomics models[J]. Acad Radiol, 2024, 31(9): 3511-3523. DOI: 10.1016/j.acra.2024.03.025.
[22]
KIM Y, JUNG H K, PARK A Y, et al. Diagnostic value of mammography for accompanying non-mass enhancement on preoperative breast MRI[J]. Acta Radiol, 2022, 63(8): 1032-1042. DOI: 10.1177/02841851211030771.
[23]
HADI Q, MASROOR I, HUSSAIN Z. Mammographic criteria for determining the diagnostic accuracy of microcalcifications in the detection of malignant breast lesions[J/OL]. Cureus, 2019, 11(10): e5919 [2024-09-12]. https://pubmed.ncbi.nlm.nih.gov/31788377/. DOI: 10.7759/cureus.5919.
[24]
DEN DEKKER B M, BAKKER M F, DE LANGE S V, et al. Reducing false-positive screening MRI rate in women with extremely dense breasts using prediction models based on data from the DENSE trial[J]. Radiology, 2021, 301(2): 283-292. DOI: 10.1148/radiol.2021210325.
[25]
BAHL M, BAKER J A, KINSEY E N, et al. MRI predictors of tumor-positive margins after breast-conserving surgery[J/OL]. Clin Imaging, 2019, 57: 45-49 [2024-09-12]. https://pubmed.ncbi.nlm.nih.gov/31128385/. DOI: 10.1016/j.clinimag.2019.05.006.
[26]
ZHOU J, LI M, LIU D Q, et al. Differential diagnosis of benign and malignant breast papillary neoplasms on MRI with non-mass enhancement[J/OL]. Acad Radiol, 2023, 30(Suppl 2): S127-S132 [2024-09-12]. https://pubmed.ncbi.nlm.nih.gov/36906443/. DOI: 10.1016/j.acra.2023.02.010.
[27]
CHADASHVILI T, GHOSH E, FEIN-ZACHARY V, et al. Nonmass enhancement on breast MRI: review of patterns with radiologic-pathologic correlation and discussion of management[J]. AJR Am J Roentgenol, 2015, 204(1): 219-227. DOI: 10.2214/AJR.14.12656.
[28]
KUBOTA K, MORI M, FUJIOKA T, et al. Magnetic resonance imaging diagnosis of non-mass enhancement of the breast[J]. J Med Ultrason, 2023, 50(3): 361-366. DOI: 10.1007/s10396-023-01290-2.
[29]
WANG X, JING L X, YAN L X, et al. A conditional inference tree model for predicting cancer risk of non-mass lesions detected on breast ultrasound[J]. Eur Radiol, 2024, 34(7): 4776-4788. DOI: 10.1007/s00330-023-10504-7.
[30]
CHEN Y H, WANG J D, ZHANG X X, et al. Correlation between apparent diffusion coefficient and pathological characteristics of patients with invasive breast cancer[J/OL]. Ann Transl Med, 2021, 9(2): 143 [2024-09-12]. https://pubmed.ncbi.nlm.nih.gov/33569445/. DOI: 10.21037/atm-20-7746.
[31]
LIU K H, YANG W, TIAN H P, et al. Nomogram based on clinical, pathological, and DWI quantitative parameters for predicting the programmed death-ligand 1 positive expression in cervical cancer: Comparison of different ROI options[J]. Chin J Magn Reson Imag, 2023, 14(10): 98-104, 115. DOI: 10.12015/issn.1674-8034.2023.10.017.
[32]
AVENDANO D, MARINO M A, LEITHNER D, et al. Limited role of DWI with apparent diffusion coefficient mapping in breast lesions presenting as non-mass enhancement on dynamic contrast-enhanced MRI[J/OL]. Breast Cancer Res, 2019, 21(1): 136 [2024-09-12]. https://pubmed.ncbi.nlm.nih.gov/31801635/. DOI: 10.1186/s13058-019-1208-y.
[33]
JANSEN S A, FAN X B, KARCZMAR G S, et al. DCEMRI of breast lesions: is kinetic analysis equally effective for both mass and nonmass-like enhancement?[J]. Med Phys, 2008, 35(7): 3102-3109. DOI: 10.1118/1.2936220.
[34]
WU Q, WANG Z, NING N, et al. Differential diagnostic value of IVIM combining with dynamic enhanced MRI in non-mass enhancement adenosis and breast cancer[J]. Chin J Magn Reson Imag, 2023, 14(2): 37-43, 49. DOI: 10.12015/issn.1674-8034.2023.02.007.
[35]
NIE T T, FENG M W, YANG K, et al. Correlation between dynamic contrast-enhanced MRI characteristics and apparent diffusion coefficient with Ki-67-positive expression in non-mass enhancement of breast cancer[J/OL]. Sci Rep, 2023, 13(1): 21451 [2024-09-12]. https://pubmed.ncbi.nlm.nih.gov/38052920/. DOI: 10.1038/s41598-023-48445-2.
[36]
XIAO J, RAHBAR H, HIPPE D S, et al. Dynamic contrast-enhanced breast MRI features correlate with invasive breast cancer angiogenesis[J/OL]. NPJ Breast Cancer, 2021, 7(1): 42 [2024-09-12]. https://pubmed.ncbi.nlm.nih.gov/33863924/. DOI: 10.1038/s41523-021-00247-3.
[37]
YANG W, YANG Y, ZHANG C L, et al. A clinicopathological-imaging nomogram for the prediction of pathological complete response in breast cancer cases administered neoadjuvant therapy[J/OL]. Magn Reson Imaging, 2024, 111: 120-130 [2024-09-12]. https://pubmed.ncbi.nlm.nih.gov/38703971/. DOI: 10.1016/j.mri.2024.05.002.
[38]
FITZPATRICK D, PIRIE K, REEVES G, et al. Combined and progestagen-only hormonal contraceptives and breast cancer risk: a UK nested case-control study and meta-analysis[J/OL]. PLoS Med, 2023, 20(3): e1004188 [2024-09-12]. https://pubmed.ncbi.nlm.nih.gov/36943819/. DOI: 10.1371/journal.pmed.1004188.
[39]
ZHANG X Y, ZHAO L G, NGO L H, et al. Prediagnostic plasma proteomics profile for hepatocellular carcinoma[J]. J Natl Cancer Inst, 2024, 116(8): 1343-1355. DOI: 10.1093/jnci/djae079.
[40]
ZHANG W, DANG R Y, LIU H Y, et al. Machine learning-based investigation of regulated cell death for predicting prognosis and immunotherapy response in glioma patients[J/OL]. Sci Rep, 2024, 14: 4173 [2024-09-12]. https://pubmed.ncbi.nlm.nih.gov/38378721/. DOI: 10.1038/s41598-024-54643-3.
[41]
LIU H W, ZHANG W, ZHANG Y H, et al. Mime: a flexible machine-learning framework to construct and visualize models for clinical characteristics prediction and feature selection[J/OL]. Comput Struct Biotechnol J, 2024, 23: 2798-2810 [2024-09-12]. https://pubmed.ncbi.nlm.nih.gov/39055398/. DOI: 10.1016/j.csbj.2024.06.035.

PREV Feasibility study of cardiac magnetic resonance four-dimensional flow imaging to evaluate early left ventricular diastolic dysfunction in patients with hypertensive heart disease
NEXT The value of CT and MRI features of pancreatic neuroendocrine neoplasm in predicting the pathological grade
  


LINK


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