Share：
Share this content in WeChat
X
[Chinese] [PDF] 4 1
• Special Focus •
Application of radiomics prognostic models based on cardiac magnetic resonance in patients with heart failure with reduced ejection fraction
GAO Yifeng  ZHOU Zhen  CHEN Yan  LI Weibo  LI Shuang  ZHAO Shifeng  XU Lei 

DOI:10.12015/issn.1674-8034.2025.11.002.


[Abstract] Objective To develop and validate a prognostic model for heart failure with reduced ejection fraction (HFrEF) patients through the integration of cardiac magnetic resonance (CMR) cine-based radiomics with clinical and imaging characteristics.Materials and Methods This study retrospectively enrolled 503HFrEF patients diagnosed according to guidelines and undergoing CMR between January 2018 and April 2023. Clinical baseline data, laboratory results, electrocardiograms, and echocardiographic parameters were collected as part of electronic health records (EHR), with follow-up for adverse cardiovascular events, including cardiac death, heart failure rehospitalization, and cardiac transplantation. All patients went through standardized CMR examination. The unsupervised nnU-Netv2 algorithm was employed to extract functional parameters from the CMR cine sequences as imaging features. Additionally, radiomic features were obtained from the same sequences with an open-source software package. After intra- and inter-group consistency testing, features were reduced via minimum redundancy maximum relevance analysis. Classifier with the best performance was selected to build the model. Models combining radiomics with imagingclinical data and standalone radiomics models were developed. The predictive power of the model was assessed by area under the curve (AUC), precision, recall, and F1- score.Results After applying stringent inclusion and exclusion criteria, a total of 389 patients with HFrEF were enrolled for model development. Of the patients followed for a median of 1041 days (IQR: 212, 1238), 87 (22.4%) experienced the endpoint. The median survival time was 495 days (IQR: 8, 1900). Twelve clinical features were identified via univariable Cox regression, which included NYHA class Ⅲ/Ⅳ and BNP. Subsequently, feature selection and dimensionality reduction yielded a final set of four imaging and nine radiomic features. Ensemble learning (EL) demonstrated optimal performance across the models. Superior prognostic performance was attained by the combined radiomics and imaging features model generated by EL classifier, which yielded an AUC of 0.789, an accuracy of 81.6%, a precision of 72.5%, a recall of 71.6%, and an F1-score of 72.0%.Conclusions This study leveraged non-contrast CMR cine to innovatively develop a radiomics-based prognostic models with relatively good predictive performance. Model's predictive efficiency was further enhanced by integrating clinical and cardiac functional imaging features.
[Keywords] heart failure；magnetic resonance imaging；cardiac magnetic resonance；radiomics；prognostic prediction

GAO Yifeng1   ZHOU Zhen1   CHEN Yan1   LI Weibo1   LI Shuang1   ZHAO Shifeng2   XU Lei1*  

1 Department of Radiology, Beijing Anzhen Hospital, Capital Medical University Beijing 100029, China

2 School of Artificial Intelligence, Capital Normal University, Beijing 100048, China

Corresponding author: XU L, E-mail: leixu2001@hotmail.com

Conflicts of interest   None.

Received  2025-09-16
Accepted  2025-11-03
DOI: 10.12015/issn.1674-8034.2025.11.002
DOI:10.12015/issn.1674-8034.2025.11.002.

[1]
Chinese Society of Cardiology, Chinese Medical Association;Chinese College of Cardiovascular Physician;Chinese Heart Failure Association of Chinese Medical Doctor Association;Editorial Board of Chines, et al. Chinese guidelines for the diagnosis and treatment of heart failure 2024[J]. Chin J Cardiol, 2024, 52(3): 235-275. DOI: 10.3760/cma.j.cn112148-20231101-00405.
[2]
MURPHY S P, IBRAHIM N E, JANUZZI J L. Heart failure with reduced ejection fraction: a review[J]. JAMA, 2020, 324(5): 488-504. DOI: 10.1001/jama.2020.10262.
[3]
SHAH K S, XU H L, MATSOUAKA R A, et al. Heart failure with preserved, borderline, and reduced ejection fraction 5-year outcomes[J]. J Am Coll Cardiol, 2017, 70(20): 2476-2486. DOI: 10.1016/j.jacc.2017.08.074.
[4]
AVERBUCH T, SULLIVAN K, SAUER A, et al. Applications of artificial intelligence and machine learning in heart failure[J]. Eur Heart J Digit Health, 2022, 3(2): 311-322. DOI: 10.1093/ehjdh/ztac025.
[5]
HEIDENREICH P A, BOZKURT B, AGUILAR D, et al. 2022 AHA/ACC/HFSA guideline for the management of heart failure: executive summary: a report of the American college of cardiology/American heart association joint committee on clinical practice guidelines[J/OL]. Circulation, 2022, 145(18): e876-e894 [2025-09-15]. https://pubmed.ncbi.nlm.nih.gov/35363500/. DOI: 10.1161/CIR.0000000000001062.
[6]
ADAMO L, ROCHA-RESENDE C, PRABHU S D, et al. Reappraising the role of inflammation in heart failure[J]. Nat Rev Cardiol, 2020, 17(5): 269-285. DOI: 10.1038/s41569-019-0315-x.
[7]
CREA F. Heart failure: from pathophysiology to deep learning-based outcome prediction[J]. Eur Heart J, 2023, 44(8): 629-632. DOI: 10.1093/eurheartj/ehad065.
[8]
BORLAUG B A, NG A C T. Cardiac magnetic resonance to enhance phenotypic characterization of HFpEF[J]. JACC Cardiovasc Imaging, 2020, 13(10): 2129-2131. DOI: 10.1016/j.jcmg.2020.06.020.
[9]
RUSSO V, LOVATO L, LIGABUE G. Cardiac MRI: technical basis[J]. Radiol Med, 2020, 125(11): 1040-1055. DOI: 10.1007/s11547-020-01282-z.
[10]
KARAMITSOS T D, FRANCIS J M, MYERSON S, et al. The role of cardiovascular magnetic resonance imaging in heart failure[J]. J Am Coll Cardiol, 2009, 54(15): 1407-1424. DOI: 10.1016/j.jacc.2009.04.094.
[11]
RAO R A, JAWAID O, JANISH C, et al. When to use cardiovascular magnetic resonance in patients with heart failure[J]. Heart Fail Clin, 2021, 17(1): 1-8. DOI: 10.1016/j.hfc.2020.09.001.
[12]
LIN K, SARNARI R, CARR J C, et al. Cine MRI-derived radiomics features of the cardiac blood pool: periodicity, specificity, and reproducibility[J]. J Magn Reson Imaging, 2023, 58(3): 807-814. DOI: 10.1002/jmri.28572.
[13]
BAEßLER B, ENGELHARDT S, HEKALO A, et al. Perfect match: radiomics and artificial intelligence in cardiac imaging[J/OL]. Circ Cardiovasc Imaging, 2024, 17(6): e015490 [2025-09-15]. https://pubmed.ncbi.nlm.nih.gov/38889216/. DOI: 10.1161/CIRCIMAGING.123.015490.
[14]
POLIDORI T, DE SANTIS D, RUCCI C, et al. Radiomics applications in cardiac imaging: a comprehensive review[J]. Radiol Med, 2023, 128(8): 922-933. DOI: 10.1007/s11547-023-01658-x.
[15]
CHANG S, HAN K, SUH Y J, et al. Quality of science and reporting for radiomics in cardiac magnetic resonance imaging studies: a systematic review[J]. Eur Radiol, 2022, 32(7): 4361-4373. DOI: 10.1007/s00330-022-08587-9.
[16]
DURMAZ E S, KARABACAK M, OZKARA B B, et al. Radiomics-based machine learning models in STEMI: a promising tool for the prediction of major adverse cardiac events[J]. Eur Radiol, 2023, 33(7): 4611-4620. DOI: 10.1007/s00330-023-09394-6.
[17]
FAHMY A S, ROWIN E J, JAAFAR N, et al. Radiomics of late gadolinium enhancement reveals prognostic value of myocardial scar heterogeneity in hypertrophic cardiomyopathy[J]. JACC Cardiovasc Imaging, 2024, 17(1): 16-27. DOI: 10.1016/j.jcmg.2023.05.003.
[18]
SZABO L, SALIH A, PUJADAS E R, et al. Radiomics of pericardial fat: a new frontier in heart failure discrimination and prediction[J]. Eur Radiol, 2024, 34(6): 4113-4126. DOI: 10.1007/s00330-023-10311-0.
[19]
ALIS D, YERGIN M, ASMAKUTLU O, et al. The influence of cardiac motion on radiomics features: radiomics features of non-enhanced CMR cine images greatly vary through the cardiac cycle[J]. Eur Radiol, 2021, 31(5): 2706-2715. DOI: 10.1007/s00330-020-07370-y.
[20]
DENG J, ZHOU L T, LI Y Y, et al. Integration of cine-cardiac magnetic resonance radiomics and machine learning for differentiating ischemic and dilated cardiomyopathy[J]. Acad Radiol, 2024, 31(7): 2704-2714. DOI: 10.1016/j.acra.2024.03.032.
[21]
CHEN Y C, HSING S C, CHAO Y P, et al. Clinical relevance of the LVEDD and LVESD trajectories in HF patients with LVEF < 35[J/OL]. Front Med (Lausanne), 2022, 9: 846361 [2025-09-15]. https://pubmed.ncbi.nlm.nih.gov/35646999/. DOI: 10.3389/fmed.2022.846361.
[22]
DEMIR A R, CELIK O, USTÜNDAĞ S, et al. Relationship between late gadolinium enhancement and ventricular repolarization parameters in heart failure patients with reduced ejection fraction[J]. Arq Bras Cardiol, 2021, 117(4): 678-687. DOI: 10.36660/abc.20200149.
[23]
ZHANG N, YANG G, GAO Z F, et al. Deep learning for diagnosis of chronic myocardial infarction on nonenhanced cardiac cine MRI[J]. Radiology, 2019, 291(3): 606-617. DOI: 10.1148/radiol.2019182304.
[24]
LI X, XU Y W, CHEN X Y, et al. Prognostic value of enhanced cine cardiac MRI-based radiomics in dilated cardiomyopathy[J/OL]. Int J Cardiol, 2025, 418: 132617 [2025-09-15]. https://www.internationaljournalofcardiology.com/article/S0167-5273(24)01239-7/abstract. DOI: 10.1016/j.ijcard.2024.132617.
[25]
ZHOU D, ZHU L Y, WU W C, et al. A novel cardiac magnetic resonance-based personalized risk stratification model in dilated cardiomyopathy: a prospective study[J]. Eur Radiol, 2024, 34(6): 4053-4064. DOI: 10.1007/s00330-023-10415-7.
[26]
FLATHER M D, SHIBATA M C, COATS A J S, et al. Randomized trial to determine the effect of nebivolol on mortality and cardiovascular hospital admission in elderly patients with heart failure (SENIORS)[J]. Eur Heart J, 2005, 26(3): 215-225. DOI: 10.1093/eurheartj/ehi115.
[27]
SU J J, SU K K, SONG Y P, et al. Clinical characteristics and prognosis of heart failure with preserved ejection fraction across diverse ejection fraction ranges[J/OL]. Rev Cardiovasc Med, 2024, 25(5): 177 [2025-09-15]. https://pubmed.ncbi.nlm.nih.gov/39076487/. DOI: 10.31083/j.rcm2505177.
[28]
MCALISTER F A, EZEKOWITZ J A, ARMSTRONG P W. Heart failure treatment and the art of medical decision making[J]. Eur J Heart Fail, 2019, 21(12): 1510-1514. DOI: 10.1002/ejhf.1655.
[29]
POCOCK S J, FERREIRA J P, GREGSON J, et al. Novel biomarker-driven prognostic models to predict morbidity and mortality in chronic heart failure: the EMPEROR-Reduced trial[J]. Eur Heart J, 2021, 42(43): 4455-4464. DOI: 10.1093/eurheartj/ehab579.
[30]
NAKAO Y M, NAKAO K, NADARAJAH R, et al. Prognosis, characteristics, and provision of care for patients with the unspecified heart failure electronic health record phenotype: a population-based linked cohort study of 95262 individuals[J/OL]. EClinicalMedicine, 2023, 63: 102164 [2025-09-15]. https://pubmed.ncbi.nlm.nih.gov/37662516/. DOI: 10.1016/j.eclinm.2023.102164.
[31]
MCDONAGH T A, METRA M, ADAMO M, et al. 2021 ESC Guidelines for the diagnosis and treatment of acute and chronic heart failure[J]. Eur Heart J, 2021, 42(36): 3599-3726. DOI: 10.1093/eurheartj/ehab368.
[32]
WILSTRUP C, CAVE C. Combining symbolic regression with the Cox proportional hazards model improves prediction of heart failure deaths[J/OL]. BMC Med Inform Decis Mak, 2022, 22(1): 196 [2025-09-15]. https://pubmed.ncbi.nlm.nih.gov/35879758/. DOI: 10.1186/s12911-022-01943-1.
[33]
SUN R, WANG X, JIANG H Y, et al. Prediction of 30-day mortality in heart failure patients with hypoxic hepatitis: Development and external validation of an interpretable machine learning model[J/OL]. Front Cardiovasc Med, 2022, 9: 1035675 [2025-09-15]. https://pubmed.ncbi.nlm.nih.gov/36386374/. DOI: 10.3389/fcvm.2022.1035675.
[34]
LONCAR G, OBRADOVIC D, THIELE H, et al. Iron deficiency in heart failure[J]. ESC Heart Fail, 2021, 8(4): 2368-2379. DOI: 10.1002/ehf2.13265.
[35]
LANSER L, POELZL G, MESSNER M, et al. Imbalance of iron availability and demand in patients with acute and chronic heart failure[J/OL]. J Am Heart Assoc, 2024, 13(9): e032540 [2025-09-15]. https://pubmed.ncbi.nlm.nih.gov/38639356/. DOI: 10.1161/JAHA.123.032540.
[36]
ROHDE L E, ZIMERMAN A, VADUGANATHAN M, et al. Associations between New York heart association classification, objective measures, and long-term prognosis in mild heart failure: a secondary analysis of the PARADIGM-HF trial[J]. JAMA Cardiol, 2023, 8(2): 150-158. DOI: 10.1001/jamacardio.2022.4427.
[37]
ZHANG H W, WANG Y R, HU B, et al. Using machine learning to develop a stacking ensemble learning model for the CT radiomics classification of brain metastases[J/OL]. Sci Rep, 2024, 14(1): 28575 [2025-09-15]. https://pubmed.ncbi.nlm.nih.gov/39562670/. DOI: 10.1038/s41598-024-80210-x.
[38]
ZHAO S L, WANG J, JIN C T, et al. Stacking ensemble learning-based [18F] FDG PET radiomics for outcome prediction in diffuse large B-cell lymphoma[J]. J Nucl Med, 2023, 64(10): 1603-1609. DOI: 10.2967/jnumed.122.265244.
[39]
KOBAYASHI K, MIYAKE M, TAKAHASHI M, et al. Observing deep radiomics for the classification of glioma grades[J/OL]. Sci Rep, 2021, 11(1): 10942 [2025-09-15]. https://pubmed.ncbi.nlm.nih.gov/34035410/. DOI: 10.1038/s41598-021-90555-2.

PREV Review on the value of cardiac magnetic resonance in the early assessment of cardiovascular diseases
NEXT Value of in vivo cardiac diffusion tensor imaging in characterizing the myocardial microstructure of hypertrophic cardiomyopathy
  


LINK


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