Deep learning-based prognostic prediction in ischemic heart failure using SPECT myocardial perfusion imaging

Abstract

Aim: This study aimed to develop and validate a multitask deep learning model, termed the information fusion model for single-photon emission computed tomography (SPECT)-based prognosis (IF-SPECT), for predicting all-cause mortality in patients with ischemic heart failure (IHF).

Methods: We retrospectively enrolled 356 patients with IHF who underwent resting-state SPECT myocardial perfusion imaging at Beijing Anzhen Hospital between 2017 and 2022. The cohort was divided into a training set (n = 213) and a validation set (n = 143). IF-SPECT uses a 3D U-Net backbone with a structure-aware module to simultaneously perform left ventricular segmentation (auxiliary task) and survival prediction (main task). A deep learning-derived risk score was generated from 3D SPECT images and then integrated with clinical variables (age, sex, body mass index, and NYHA_3_4) in the final prognostic model.

Results: IF-SPECT achieved the highest C-index in the validation cohort (0.724; 95%CI: 0.621-0.814), outperforming both the clinical and radiomics models in paired bootstrap testing. In the validation cohort, patients classified as high-risk by IF-SPECT had a significantly higher risk of all-cause mortality than those in the low-risk group (HR: 5.01; 95%CI: 2.17-11.59; P < 0.001). Exploratory subgroup analyses suggested that IF-SPECT further restratified patients within the low HM_TPD subgroup into distinct prognostic categories.

Conclusion: IF-SPECT provides a promising approach for prognostic risk stratification in IHF by combining a deep learning-derived imaging risk score with routine clinical variables.

Keywords

Ischemic heart failure / SPECT / deep learning / all-cause mortality / survival analysis

References

[1] Ran J,Wang J.Global, regional, and national burden of heart failure and its underlying causes, 1990-2021: results from the global burden of disease study 2021.Biomark Res2025;13:16 PMCID:PMC11755835

[2] Tsao CW,Almarzooq ZI.Heart disease and stroke statistics—2023 update: a report from the american heart association.Circulation2023;147:8

[3] Mcmurray JJ,Inzucchi SE.Dapagliflozin in patients with heart failure and reduced ejection fraction.N Engl J Med2019;381:1995-2008

[4] Heidenreich PA,Aguilar D.2022 ACC/AHA/HFSA guideline for the management of heart failure.J Card Fail2022;28:e1-e167

[5] Mcdonagh TA,Adamo M.2021 ESC guidelines for the diagnosis and treatment of acute and chronic heart failure: developed by the Task Force for the diagnosis and treatment of acute and chronic heart failure of the European Society of Cardiology (ESC). With the special contribution of the Heart Failure Association (HFA) of the ESC.Eur J Heart Fail2022;24:4-131

[6] Goldberger JJ,Boineau R.Risk stratification for sudden cardiac death: a plan for the future.Circulation2014;219:4

[7] Pocock SJ,Mcmurray JJ.Predicting survival in heart failure: a risk score based on 39 372 patients from 30 studies.Eur Heart J2013;34:1404-13

[8] Hachamovitch R,Shaw LJ.Incremental prognostic value of myocardial perfusion single photon emission computed tomography for the prediction of cardiac death: differential stratification for risk of cardiac death and myocardial infarction.Circulation1998;97:535-43

[9] Amini M,Hajianfar G.Machine learning-based diagnosis and risk classification of coronary artery disease using myocardial perfusion imaging SPECT: a radiomics study.Sci Rep2023;13:14920 PMCID:PMC10493219

[10] Leslie WD,Goertzen A.Prediction of 2-year major adverse cardiac events from myocardial perfusion scintigraphy and clinical risk factors.J Nucl Cardiol2022;29:1956-63 PMCID:PMC8551291

[11] Miller RJ.Artificial intelligence in nuclear cardiology: an update and future trends.Semin Nucl Med2024;54:648-57

[12] Betancur J,Motlagh M.Deep learning for prediction of obstructive disease from fast myocardial perfusion SPECT.JACC Cardiovascular Imaging2018;11:1654-63 PMCID:PMC6135711

[13] Kusumoto D,Hashimoto M.A deep learning-based automated diagnosis system for SPECT myocardial perfusion imaging.Sci Rep2024;14:13583 PMCID:PMC11169468

[14] Miller RJ,Lemley M.Artificial intelligence-enhanced perfusion scoring improves the diagnostic accuracy of myocardial perfusion imaging.J Nucl Med2025;66:648-53 PMCID:PMC11960614

[15] Hajianfar G,Sabouri M.Artificial intelligence-powered coronary artery disease diagnosis from SPECT myocardial perfusion imaging: a comprehensive deep learning study.Eur J Nucl Med Mol Imaging2025;52:3019-35 PMCID:PMC12162751

[16] Papandrianos NI,Feleki A,Papageorgiou EI.Deep learning exploration for SPECT MPI polar map images classification in coronary artery disease.Ann Nucl Med2022;36:823-33

[17] Zahiri N,Razavi-Ratki S.Deep learning analysis of polar maps from SPECT myocardial perfusion imaging for prediction of coronary artery disease.Res Sq2021;

[18] Qiao X,Meng H.Multi-modality deep learning-based [68Ga]Ga-DOTA-FAPI-04 PET polar map generation: potential value in detecting reactive fibrosis after myocardial infarction.Eur J Nucl Med Mol Imaging2024;51:3944-59

[19] Ko C,Huang C.Polar map-free 3D deep learning algorithm to predict obstructive coronary artery disease with myocardial perfusion CZT-SPECT.Eur J Nucl Med Mol Imaging2022;50:376-86

[20] Du Y,Lin C.Deep learning-based obstructive coronary artery disease prediction from myocardial perfusion SPECT.Eur J Nucl Med Mol Imaging2025;53:2686-96

[21] Gao Z,Kang Y.Improving the prognostic evaluation precision of hospital outcomes for heart failure using admission notes and clinical tabular data: multimodal deep learning model.J Med Internet Res2024;26:e54363 PMCID:PMC11099809

[22] Wang J,Duan S.Multimodal data-driven prognostic model for predicting long-term prognosis in patients with ischemic cardiomyopathy and heart failure with preserved ejection fraction after coronary artery bypass grafting: a multicenter cohort study.J Am Heart Assoc2024;13:e036970 PMCID:PMC11681600

[23] Huang S,Seyyedi S,Lungren MP.Fusion of medical imaging and electronic health records using deep learning: a systematic review and implementation guidelines.NPJ Digit Med2020;3:136 PMCID:PMC7567861

[24] Dhar T,Borra S.Challenges of deep learning in medical image analysis—improving explainability and trust.IEEE Trans Technol Soc2023;4:68-75

[25] Murti DMP,Wibawa AP.K-nearest neighbor (K-NN) based missing data imputation. In 2019 5th International Conference on Science in Information Technology (ICSITech); 2019 Oct 23-24; Yogyakarta, Indonesia; pp. 83-8.

[26] Ronneberger O,Brox T.U-net: convolutional networks for biomedical image segmentation. In: Navab N, Hornegger J, Wells WM, Frangi AF, editors. Medical image computing and computer-assisted intervention - MICCAI 2015. Cham: Springer International Publishing; 2015, pp. 234-41.

[27] Isensee F,Kohl SAA,Maier-Hein KH.nnU-Net: a self-configuring method for deep learning-based biomedical image segmentation.Nat Methods2020;18:203-11

[28] Hu C,Zheng Z.AI-based large-scale screening of gastric cancer from noncontrast CT imaging.Nat Med2025;31:3011-9

[29] Hu Y,Zhou Y.AI-based diagnosis of acute aortic syndrome from noncontrast CT.Nat Med2025;31:3832-44 PMCID:PMC12618251

[30] Hu J,Sun G.Squeeze-and-excitation networks. In 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition; 18-23 June 2018; Salt Lake City, UT, USA.

[31] Woo S,Lee J.CBAM: convolutional block attention module. In: Ferrari V, Hebert M, Sminchisescu C, Weiss Y, editors. Computer vision - ECCV 2018. Cham: Springer International Publishing; 2018. pp. 3-19.

[32] Katzman JL,Cloninger A,Jiang T.DeepSurv: personalized treatment recommender system using a Cox proportional hazards deep neural network.BMC Med Res Methodol2018;18:24 PMCID:PMC5828433

[33] Cox DR.Regression models and life-tables.J R Stat Soc Series B Stat Methodol1972;34:187-202

[34] Lambin P,Leijenaar R.Radiomics: extracting more information from medical images using advanced feature analysis.Eur J Cancer2012;48:441-6 PMCID:PMC4533986

[35] Tibshirani R.Regression shrinkage and selection via the lasso.J R Stat Soc Series B Stat Methodol1996;58:267-88

[36] Selvaraju RR,Das A,Parikh D.Grad-CAM: visual explanations from deep networks via gradient-based localization. In Proceedings of the 2017 IEEE International Conference on Computer Vision (ICCV). Venice, Italy; 22-29 October 2017; pp. 618-26.

[37] Van der Maaten L.Visualizing data using t-SNE.J Mach Learn Res2008;9:2579-605Available from: https://www.jmlr.org/papers/v9/vandermaaten08a.html [Last accessed on 20 Jul 2026]

[38] Harrell FE,Pryor DB,Rosati RA.Evaluating the yield of medical tests.JAMA1982;247:2543-6

[39] Heagerty PJ,Pepe MS.Time-dependent ROC curves for censored survival data and a diagnostic marker.Biometrics2004;56:337-44

[40] Wiefels C,Craviolatti PS.Artificial intelligence in nuclear cardiology: technical perspectives, strategic directions, and recommendations from an IAEA expert working group.Semin Nucl Med2026;56:119-31