Heart failure with preserved ejection fraction (HFpEF), defined as a left ventricular ejection fraction (LVEF) ≥50% with evidence of elevated left ventricular filling pressures and clinical heart failure1 that is not attributable to a specific underlying cause2, comprises nearly half of all heart failure hospitalizations3. Despite the advances in therapies for heart failure with reduced ejection fraction (HFrEF), medical therapies offer only modest benefit in HFpEF. As a result, HFpEF remains a morbid condition with post-hospitalization mortality rates of 15% at 1 year and up to 75% at 5 years4.Diagnosing HFpEF can be challenging as it is a heterogenous syndrome with comorbidities, including obesity, physical inactivity, insulin resistance, conditions that can themselves cause exertional dyspnea. Multiple underlying pathophysiological phenotypes, including cardiometabolic dysregulation, endothelial dysfunction, left atrial myopathy, and pulmonary vascular disease, exist within the broader spectrum of HFpEF3,5–7. Multimodality imaging, and cardiovascular magnetic resonance (CMR) imaging in particular, is well positioned to aid in the diagnosis of HFpEF and to distinguish between pathophysiological phenotypes8,9.Once the clinical syndrome of heart failure has been identified, HFpEF mimickers, specific diseases that cause the clinical syndrome of heart failure in the setting of normal LVEF, must be excluded, as there may be disease-modifying therapies for alternative diagnoses. CMR has an established role in diagnosing these mimickers and will be discussed in detail herein. Subsequently, a diagnosis of HFpEF may be suggested using scoring systems such as the H2FPEF, HFA-PEFF, and HFpEF-JH score, which incorporate echocardiographic indices of diastolic dysfunction2,10. CMR holds several advantages over echocardiography and there is growing interest in CMR-derived indices of diastolic dysfunction, which may allow for earlier diagnosis and ultimately improved outcomes.This review explores the role of CMR in diagnosing HFpEF, assessing diastolic dysfunction, distinguishing HFpEF from mimicking conditions, and identifying high-risk phenotypes. Advances in CMR-derived hemodynamic assessment, strain analysis, and artificial intelligence applications are also discussed.
CMR in HFpEF Diagnosis and Phenotyping
While echocardiography remains the primary imaging modality, CMR offers superior reproducibility, detailed tissue characterization, and is less impacted by patient body habitus, making it a powerful tool for HFpEF evaluation11. Society guidelines currently recommend CMR for the evaluation of HFpEF when specific cardiomyopathy etiologies are suspected and when poor acoustic windows limit echocardiography1,2,12 (Figure 1). Emerging advancements in CMR techniques may soon improve its clinical application for diagnosing and predicting outcomes in HFpEF.
CMR is the gold standard imaging modality for assessing chamber volumes and ejection fraction. To establish a diagnosis of HFpEF, the European Society of Cardiology guidelines require evidence of structural heart disease, including left ventricular hypertrophy (LVH) and left atrial (LA) enlargement, or diastolic dysfunction12. In a study of obese patients with HFpEF, left ventricle (LV) mass index and LA/LV ratio were among the clinical variables with the strongest association with HFpEF10. CMR provides precise LV mass index measurements, which have been associated with increased risk of adverse outcomes in HFpEF patients (Figure 2). Advanced techniques such as myocardial T1 and T2 mapping allow for direct tissue characterization, improving diagnostic accuracy and risk stratification11.
Strain is a marker of deformation and is a powerful diagnostic and prognostic marker for global and regional LV function. CMR-derived strain can be assessed by several techniques including myocardial tagging, fast stain-encoded CMR (fast-SENC), phase-contrast CMR, and most commonly feature-tracking (FT)-CMR13. Analogous to speckle tracking echocardiography, FT-CMR generates strain values by tracking components of the myocardium throughout the cardiac cycle. While there are few studies investigating CMR-derived strain in HFpEF, global longitudinal strain (GLS) values correlate with invasive diastolic function indices14 and may predict outcomes in HFpEF15. Specifically, fast-SENC has been used to assess LV strain responses to isometric handgrip exercise in HFpEF16. Participants with HFpEF showed heterogenous GLS responses, with greater absolute changes than controls, reflecting underlying disease severity. The diagnostic value of fast-SENC during exercise was limited due to high inter-individual variability. Finally, FT-CMR has been used to evaluate circumferential, radial, and LV rotation indices of torsion, which can be used to distinguish between specific cardiomyopathies17.
CMR enables noninvasive assessment of diastolic function using transmitral and pulmonary venous flow velocities18. Phase-contrast CMR-derived E/e’ ratios correlate with invasively measured filling pressures, providing insights into diastolic dysfunction19. In large study comparing CMR-derived and echocardiography-derived measures of diastolic dysfunction, there was strong agreement between key diastolic parameters, with the highest correlation for E/’e (r =0.9, p < 0.001)20. In HFpEF, transmitral velocities alone do not reliably reflect LV filling pressures, but combining E with early mitral septal tissue velocity (E’) offers a more accurate estimate21. Additionally, CMR-modeled PCWP correlates with RHC values (r = 0.55, p < 0.0001) and outperform TTE in identifying elevated filling pressures22. These data suggest CMR may be a powerful tool in diagnosing HFpEF, though limitations including lower temporal resolution, longer image acquisition time, and challenges associated with arrhythmias remain barriers to widespread clinical adoption.
Elevated left atrial filling pressures, either at rest or during exercise, are common in HFpEF and contribute to LA enlargement, a key marker of diastolic dysfunction and predictor of heart failure outcomes23. As LV relaxation becomes impaired, higher LA pressures are required to maintain LV filling, leading to LA dilatation. Recently, a distinct HFpEF phenotype characterized by disproportionate LA myopathy has been identified24. While echocardiography is recommended for LA volume assessment, it often underestimates LA volumes and CMR provides more accurate, geometry-independent measurements. Notably, several studies suggest LA remodeling may preceded HFpEF symptoms, indicating subtle changes in LA morphology may be used to identify patients at risk for HFpEF25,26.
There is increasing appreciation that measures of LA function may provide clinical value beyond LA volume23,27. The function of the LA can be divided into three components: reservoir (pulmonary venous return during ventricular systole), conduit (between the pulmonary venous system and ventricle during early ventricular diastole), and pump (active augmentation of ventricular filling during late ventricular diastole)23. While not validated for clinical practice, there is interest in using CMR to assess LA performance. Established markers of LA function, including LA ejection fraction can be reliably assessed by CMR. CMR-derived LA ejection fraction correlates with LV dysfunction, exercise intolerance, and heart failure mortality28. Impaired LA reservoir function assessed by CMR is associated with increased hospitalization and mortality risk29. Novel CMR indices of LA function, such as LA circulation transit time30 and left atrioventricular coupling index31 are actively being investigated for their utility in risk stratification. Finally, FT-CMR-derived atrial strain has emerged as a promising marker of LA dysfunction and HFpEF progression29.
Right-heart dysfunction and pulmonary vascular disease are important, high-risk HFpEF phenotypes5. Despite the high prevalence and prognostic value of right heart disease in HFpEF outcomes32, there have been few investigations exploring CMR evaluation of right-heart function in HFpEF33. In one of the only published reports, Aschauer et al. found impaired RVEF by CMR was associated with higher pulmonary artery pressures, lower 6-minute walk distance, and higher rates of cardiovascular events34. These data, consistent with echocardiographic studies, suggest CMR may be able to identify high-risk right- heart dysfunction/pulmonary vascular disease HFpEF phenotype. Novel CMR-indices of pulmonary hypertension have been validated in HFpEF. In a CMR study of HFpEF patients, the ratio of the pulmonary artery in relation to the ascending aorta correlated with mean pulmonary artery pressure35. The use of CMR to identify HFpEF patients with RV dysfunction and pulmonary hypertension is an area of active investigation and future research should focus on integrating CMR-based RV and pulmonary hemodynamic assessments into routine HFpEF evaluation.
In obese individuals, epicardial adipose tissue (EAT) is believed to play a role in HFpEF pathophysiology and is an important prognostic marker. Pathological expansion of EAT increases inflammatory cytokines, including interleukin (IL)-1β, IL-5, and tumor necrosis factor – alpha, which may directly promote fibrosis, adverse remodeling, and ultimately lead to diastolic dysfunction. In line with this hypothesis, EAT is increased in HFpEF36 and is associated with proteomic markers of inflammation, insulin resistance, endothelial dysfunction, and dyslipidemia37. Another proposed mechanism for EAT in HFpEF pathophysiology is by exerting mechanical compressive force resulting in pericardial restriction. Moreover, EAT has prognostic value in identifying high-risk HFpEF phenotypes, as obese patients with HFpEF and increased EAT have higher RA, PA and PCWP at rest and during exercise compared to patients with HFpEF without elevated EAT38.
Notably, most of the work investigating the role of EAT in HFpEF has relied on echocardiography to assess EAT thickness over the right ventricle free wall. CMR allows for assessment of whole heart EAT volume and mass, which may allow for additional insights into HFpEF pathophysiology and refined risk stratification. Notably, Nakamori et al. recently reported that left and right ventricular EAT have distinct effects on left ventricular stiffness in HFpEF—an observation that could not have been achieved with echocardiography alone.39
Myocardial fibrosis and changes in the content and structure of extracellular matrix (ECM) for decades was thought to play a central role in HFpEF has more recently been shown to be a less significant pathophysiological mechanism by which HFpEF develops in the predominant metabolic HFpEF phenotype40,41. Nevertheless, both diffuse and focal myocardial fibrosis are more prevalent in patients with HFpEF compared to healthy controls42. CMR parametric mapping sequences (T1/T2 mapping and extracellular volume [ECV] quantification) allow for myocardial tissue characterization and quantification of ECM. CMR-LGE detects focal fibrosis, while T1 mapping and ECV quantification assess diffuse interstitial fibrosis. In HFpEF, LGE is more prevalent compared to healthy controls and is an independent predictor of HF outcomes42,43. The pattern of LGE can also help to suggest specific etiologies of cardiomyopathy (see “HFpEF Mimickers and CMR”).
The excessive ECM in HFpEF impairs myocyte function and is believed to play an important role in impaired relaxation44 and cardiac sympathetic innervation in HFpEF45. ECV, which in the absence of edema and amyloid protein, primarily reflects expansion of the ECM, can be calculated using a combination of pre- and post-LGE contrast T1 mapping, hematocrit levels, and blood signal T1 values46. CMR-derived ECV correlates with histological ECM expansion in HFpEF47. Moreover, CMR-derived ECV also independently predicts invasively measured LV stiffness in patients with HFpEF48,49, is associated with impaired diastolic dysfunction50, and heart failure outcomes51. More recently, ECV has been shown to correlate with miR-128a-3p expression in patients with HFpEF—a microRNA implicated in cardiac hypertrophy, fibrosis, and functional impairment—highlighting its potential mechanistic role in the pathophysiology of this phenotype52. Notably, myocardial fibrosis may precede the development of HFpEF53. Given these findings, there has been great interest in using ECV as a biomarker of ECM potential maker of HFpEF risk.
Finally, CMR-derived ECV may be useful for identifying HFpEF phenotypes that could inform clinical trial enrollment. A post-hoc analysis of the Aldo-DHF study revealed that HFpEF patients with a biochemical profile indicating high collagen cross-linking were resistant to the benefits of spironolactone54. While additional investigations are needed before CMR-derived ECV is integrated into practice, the Society for Cardiovascular Magnetic Resonance has issued recommendations for the use of T1 mapping and ECV quantification in both clinical practice and research studies55
Stress CMR is a validated technique to evaluate myocardial ischemia and coronary microvascular function. Microvascular dysfunction is prevalent in HFpEF, even in the absence of epicardial coronary artery disease, and is a strong predictor of HF events56. Although microvascular disease can contribute to chronic ischemia, impaired coronary flow reserve, and subsequent fibrosis, it remains uncertain whether it serves as a primary pathophysiological driver of HFpEF or as a consequence of disease progression.
HFpEF Mimickers and CMR
CMR is pivotal in differentiating true HFpEF from infiltrative cardiomyopathies (amyloidosis, sarcoidosis), hypertrophic cardiomyopathy, and pericardial disease. Characteristic LGE patterns and parametric mapping techniques facilitate accurate diagnosis. In a study by Kanagala et al., 27% of patients thought to have HFpEF were found to have alternative diagnoses including hypertrophic cardiomyopathy (HCM)57. Distinguishing between HFpEF and its mimickers is critical for optimal management. The use of CMR to distinguish between specific cardiomyopathies has been reviewed previously9, however, key updates are covered here.
Cardiac amyloidosis (CA) is a restrictive cardiomyopathy caused by amyloid protein deposition in the myocardium, leading to a HFpEF phenotype. The 2022 AHA/ACC/HFSA Guidelines for Management of Heart Failure recommends CA diagnosis be considered when LV thickness ≥14 mm in the context of heart failure symptoms, particularly if discordant low QRS voltage on ECG is present1. Diagnosis can be challenging as highlighted by a study where 14% of HFpEF participants were found to have amyloidosis on endomyocardial biopsy, with nearly half of the cases clinically unsuspected41. CMR plays a crucial role in identifying CA, with a characteristic global subendocardial LGE pattern that indicates the extent of the disease and predicts outcomes58 (Figure 3, 4). Native T1 and ECV mapping help assess amyloid burden, with ECV expansion serving as an early marker even in the absence of LGE58. These imaging modalities are important for distinguishing CA from other hypertrophic conditions and guide confirmatory testing1. While LGE has high sensitivity (85%) and specificity (92%) for CA59, distinguishing between CA subtypes (light-chain and transthyretin), requires further investigation, with certain patterns of LGE offering some insight60. Additionally, T1/T2 mapping and ECV may improve the diagnosis of CA, especially in the early stages before significant LGE is present61,62. Emerging evidence has demonstrated that CMR tissue characterization, including T1 mapping and ECV quantification, are sensitive techniques for monitoring the effects of both established and emerging therapies for transthyretin (TTR)63 and light-chain (AL) amyloidosis64, making CMR a valuable tool for clinical follow-up in these patients.
HCM is a relatively common, genetic disorder marked by LV hypertrophy and potential outflow tract obstruction. CMR is essential for distinguishing HCM from other causes of preserved ejection fraction, offering superior visualization of hypertrophic segment – particularly of the anterolateral free wall and apex – enabling accurate measurement of LV wall thickness (maximal end-diastolic LV wall thickness of ≥15 mm suggests HCM65) (Figure 5). Additionally, LGE scar patterns, although not specific66, aid in risk stratification and predict adverse outcomes such as heart failure67 and SCD68. With the advent of targeted pharmacologic therapies for obstructive hypertrophic cardiomyopathy (HCM), preliminary imaging data from clinical trials investigating selective myosin inhibitors have underscored the potential role of cardiac magnetic resonance (CMR) in detecting structural myocardial changes, including reductions in left ventricular mass, myocardial wall thickness, and native T1 values, without corresponding alterations in late gadolinium enhancement (LGE)-defined scar burden69,70.
Novel CMR Techniques – Hemodynamics and Metabolism
Novel CMR sequences are being developed to assess hemodynamics and metabolic activity, which may enhance HFpEF phenotyping and provide a deeper understanding of HFpEF pathophysiology.
HFpEF often presents subtly with exertional dyspnea as the initial symptom. Resting filling pressures may be normal and elevated filling pressures may only emerge with exercise. Although exercise stress echocardiography has been proposed as a non-invasive alternative to invasive testing, its utility is constrained by suboptimal image quality and false negatives—particularly in HFpEF patients who do not exhibit markers of diastolic dysfunction during exertion71. Consequently, exercise stress echocardiography only offers marginal improvements over resting assessments3. In contrast, real-time (RT)-CMR may overcome these limitations, offering a more robust diagnostic approach. Advances in CMR enable novel RT imaging at high temporal resolutions (20 ms) that is not limited by motion artifacts like conventional CMR sequences. Breath-holding during imaging is not required. In the HFpEF-Stress Trial, RT-CMR enhanced HFpEF diagnosis and LA long axis strain obtained by RT-CMR was found to be independently associated with HFpEF72.
Myocardial stiffness is increased in patients with HFpEF73, HCM74, and CA75. There is interest in assessing myocardial stiffness as a biomarker for HFpEF, however, there are currently no validated imaging techniques that can directly assess myocardial stiffness in vivo. Sheer wave elastography, which relies on an external vibrating source to generate shear waves in the tissue of interest76, is an emerging imaging approach being developed to address this gap. Magnetic resonance elastography (MRE) leverages CMR to capture the vibrational displacement of tissues, which is subsequently processed to generate a precise stiffness map. Recent studies have demonstrated MRE’s feasibility for detecting increased myocardial stiffness in patients with cardiac amyloidosis75, while parallel research in HCM revealed that patients exhibit significantly higher myocardial stiffness compared to healthy controls, with stiffness strongly correlating with LGE74. These encouraging results highlight the potential of MRE as a diagnostic tool, however, additional studies are needed to determine whether MRE has clinical implications for HFpEF diagnosis and prognosis.
Diffusion Tensor (DT)-CMR is a novel technique that allows for characterization of the cardiac microstructure. Diffusion imaging exploits the effect of the myocardial microstructure on the diffusion of water to gain insight into the underlying myocardial organization77. Preliminary studies using DT-CMR have identified microstructural derangements in HCM78 and dilated cardiomyopathy79. These studies suggest that DT-CMR has the potential to identify and differentiate HFpEF phenotypes by revealing microstructural changes in the myocardial tissue. While there is growing interest in using DT-CMR to characterize such changes in HFpEF, more data is needed to determine its clinical impact.
Disturbances in myocardial energy metabolism have been implicated as an important factor in HFpEF pathophysiology7,80. Cardiac magnetic resonance spectroscopy (MRS) is a novel approach for evaluating myocardial metabolism by utilizing the intrinsic magnetic resonance signals from nuclei, including 31Phosphorus (31P) and 1Hydrogen (1H).
31P-MRS assesses myocardial metabolism by measuring phosphorus-containing metabolites, most importantly adenosine triphosphate (ATP) and phosphocreatine (PCr)81. PCr is the principal energy reserve metabolite that sustains a steady flow of ATP in cardiomyocytes. The ratio of PCr/ATP serves as a surrogate marker for available cellular energy stores. In a small study of patient with HFpEF, PCr/ATP ratio was reduced and correlated with impaired LV relaxation80 as well as clinical symptoms of HF82. A stepwise reduction in PCr/ATP ratio parallels the progressive decline in diastolic function in HFpEF82, suggesting PCr/ATP may be a powerful tool for assessing HFpEF disease progression. Finally, 31P-MRS has been used to enhance the mechanistic understanding of therapies in HFpEF. The EMPA-VISION and DoPING-HFpEF trials found that treatment with empagliflozin and trimetazidine did not significantly impact myocardial energetics83,84.
The metabolic profile of HFpEF is an area of active area of investigation and not well defined. Recent studies suggest that fatty acid oxidation is impaired and there is a shift to glucose metabolism7. This preferential metabolic adaptation, favoring glucose utilization, is energy-efficient and oxygen-sparing. While the normal lipid content in the myocardium is minimal (<1%)85, a reduction in fatty acid oxidation results in the accumulation of intracellular lipids, potentially resulting in lipotoxicity and myocyte dysfunction.
1H-MRS enables quantification of myocardial lipid content, including triglycerides81. In a small CMR study, HFpEF patients had double the myocardial triglyceride content compared to controls, with higher triglyceride levels correlating with reduced diastolic strain rate and maximal oxygen consumption86. Notably, HFpEF patients exhibit a greater myocardial fat burden than those with HFrEF87. Further analysis of 1H-MRS data in women suggests that myocardial steatosis may play a mechanistic role in diastolic dysfunction, particularly in those with coronary microvascular dysfunction88. The existing evidence underscores the potential of 1H-MRS as a tool for early HFpEF detection in high-risk individuals and highlights myocardial triglycerides as a promising therapeutic target.
Artificial Intelligence and Machine Learning
Artificial Intelligence (AI), through machine learning (ML), is revolutionizing CMR by automating image analysis, improving diagnostic accuracy, and reducing interpretation and scan time. AI-driven segmentation models, particularly convolutional neural networks, have demonstrated expert-level performance in analyzing CMR images, facilitating large-scale population studies89,90. Automated assessment of LVEF and strain parameters enhances risk stratification and prognosis prediction in HFpEF91. Moreover, a major advancement in AI-driven CMR is virtual native enhancement (VNE), which synthesizes LGE-equivalent images from native T1 and cine imaging without contrast agents92. VNE has shown strong agreement with traditional LGE in hypertrophic cardiomyopathy but has yet to be validated in HFpEF92. This innovation may obviate the need for LGE contrast in the future.
AI also refines parametric mapping techniques, optimizing T1 and T2 mapping by reducing noise and correcting motion artifacts93. These improvements enhance scar imaging and provide a more precise assessment of myocardial fibrosis and edema. While ML data on HFpEF are scarce, there are reports demonstrating the clinical utility of AI models in CMR for deep phenotyping of patients with known or suspected cardiovascular disease, including HCM92, CA94 and CS95. Despite these advancements, there is still insufficient evidence regarding AI’s cost-effectiveness, superiority over human interpretation, and impact on patient outcomes.
Future Directions and Clinical Implications
CMR continues to advance HFpEF assessment by distinguishing phenotypes, guiding targeted therapeutic management, and refining risk stratification (Table 1). Emerging techniques such as multiparametric mapping, real-time hemodynamic imaging, and artificial intelligence-driven analysis will further improve diagnostic precision. Future research should focus on integrating these techniques into clinical workflows and validating their prognostic value in large-scale HFpEF cohorts.
Conclusion
CMR is powerful tool in HFpEF evaluation, offering detailed structural, functional, and tissue characterization (Figure 1). AI and ML are transforming CMR analysis by automating image segmentation, enhancing parametric mapping, and reducing scan interpretation times. As imaging technology evolves, AI-driven CMR will likely play an increasingly central role in HFpEF diagnosis and risk assessment, ultimately improving patient outcomes through precision medicine approaches.