Routine Heart MRI Reveals Hidden Damage in a Common, Elusive Form of Heart Failure

Heart failure with preserved ejection fraction, known as HFpEF, has long been one of cardiology’s most frustrating puzzles. Patients arrive breathless and exhausted, yet their hearts pump a seemingly normal fraction of blood with every beat. Now a new clinical analysis suggests that the answer may be hiding in plain sight, inside a routine cardiac magnetic resonance scan that thousands of patients already undergo every day.

The study, published in Frontiers in Cardiovascular Medicine, analyzed 126 adults evaluated between January 2023 and December 2025 at a tertiary teaching hospital in China. The cohort included 72 patients with adjudicated HFpEF, 27 hypertensive controls without heart failure, and 27 healthy volunteers. All participants had available 1.5-Tesla cine cardiovascular magnetic resonance studies, the foundational balanced steady-state free-precession sequence that captures the beating heart in contiguous short-axis slices and standard long-axis views without any contrast agent.

The technical premise is elegant. Cine CMR quantifies ventricular volumes, myocardial mass, and global function with high reproducibility and without geometric assumptions. The researchers performed blinded analysis of left ventricular volumes, ejection fraction, left ventricular mass index, left atrial volume index, right ventricular ejection fraction, and feature-tracking global longitudinal strain. Two independent readers reanalyzed 25 randomly selected studies, yielding intraclass correlation coefficients of 0.95 for left atrial volume index and 0.91 for strain, with within-subject coefficients of variation of just 4.8 and 6.2 percent.

The results were striking. Ejection fraction was preserved and statistically indistinguishable across all three groups, hovering near 59 to 61 percent. Yet HFpEF patients carried a completely different cardiac signature. Their left atrial volume index averaged 52.7 milliliters per square meter, compared with 39.6 in hypertensive controls and 31.8 in healthy volunteers. Their global longitudinal strain was markedly less negative at −14.2 percent, versus −17.1 and −19.0 in the comparator groups, indicating impaired longitudinal deformation of the subendocardial fibers that ejection fraction cannot reveal.

Standardized effect sizes told the same story with unusual force. Against healthy volunteers, left atrial volume index and strain showed Cohen’s d values of 2.02 and 2.06, among the largest imaging effect sizes reported in this syndrome. Even against the more demanding hypertensive comparator, both markers remained large at 1.21. Left ventricular mass index was also significantly elevated in HFpEF, at 76.8 grams per square meter versus 68.7 and 58.5, reflecting the hypertrophic remodeling that chronic pressure overload imposes on the myocardium.

Diagnostic discrimination followed the biology. Strain alone achieved an apparent area under the curve of 0.85, and left atrial volume index 0.83, substantially exceeding mass index at 0.74. A multiparametric model combining all three markers, fitted with Firth-penalized logistic regression to handle the quasi-complete separation produced by comorbidities, reached an apparent AUC of 0.87. After 1,000-sample bootstrap internal validation, the optimism-corrected AUC settled at 0.84 with a calibration slope of 0.88, indicating only modest overfitting. Notably, the model did not significantly outperform strain alone, a finding the authors interpret as a call for restraint rather than a deployable diagnostic rule.

The most clinically consequential sensitivity analysis restricted the comparator to hypertensive controls, the population that most closely resembles real diagnostic uncertainty. Discrimination fell to 0.77 for left atrial volume index, 0.79 for strain, and 0.81 for the model, confirming that performance against a mixed comparator including healthy volunteers is optimistic. The authors argue that external validation should preferentially enrich comparator cohorts with treated hypertension, obesity, atrial fibrillation, and unexplained exertional dyspnea rather than unselected healthy individuals.

The serial component of the study may prove the most provocative. Forty-six of the 72 HFpEF patients completed repeat cine CMR at approximately six months after individualized treatment optimization, which could include diuretic adjustment, blood pressure control, sodium-glucose cotransporter 2 inhibition, rhythm management, and comorbidity-directed care. Over that interval, left atrial volume index fell by a mean of 4.8 milliliters per square meter, a 9.4 percent relative reduction with a standardized paired effect size of 1.85, and strain improved by 1.3 percentage points. Both changes exceeded the prespecified minimal detectable change thresholds of 2.8 and 1.2, while ejection fraction remained static.

The authors are careful about what this means. Because treatment was individualized rather than randomized and the substudy included no untreated comparator, the six-month changes cannot be attributed to any single intervention and may be influenced in part by regression to the mean. They describe the findings as evidence that cine CMR can detect selective short-term temporal changes in atrial size and longitudinal deformation, rather than definitive proof of treatment-induced reverse remodeling. Completers and non-completers did not differ significantly in age, sex, body mass index, biomarker levels, or baseline imaging markers, arguing against major attrition bias.

A post hoc stratification added a further layer. When the researchers defined a cine CMR phenotype as fulfillment of at least two of three Youden-derived thresholds, 56 of the 72 adjudicated HFpEF patients, or 77.8 percent, exhibited the imaging signature. Patients with the phenotype carried a markedly higher biomarker burden, with median NT-proBNP of 892 versus 462 picograms per milliliter, and among those completing paired follow-up they showed a significantly larger reduction in left atrial volume index over six months. The authors envision cine-derived markers as supportive phenotyping tools in patients already undergoing CMR, and as an adjunct after inconclusive echocardiography, particularly when acoustic windows are poor or HFA-PEFF scoring is discordant. They do not propose replacing the echocardiography-first pathway or comprehensive multiparametric CMR when tissue characterization is indicated. External validation in multicenter cohorts remains essential before any clinical application, but the message from this cohort is clear: the routine cine sequence, already acquired in most cardiac magnetic resonance laboratories, can reveal the structural and functional signature that ejection fraction alone misses.

Subject of Research: Cardiovascular Medicine

Article Title: Routine cine cardiovascular magnetic resonance for clinical phenotyping and serial assessment of heart failure with preserved ejection fraction: a retrospective single-center clinical analysis

Article References: Wang, J., Xie, Y., Zeng, C., Nie, L., & Yang, Y. (2026). Routine cine cardiovascular magnetic resonance for clinical phenotyping and serial assessment of heart failure with preserved ejection fraction: a retrospective single-center clinical analysis. Frontiers in Cardiovascular Medicine, 13, Article 1965415. https://doi.org/10.3389/fcvm.2026.1965415

Image Credits: AI Generated

DOI: 10.3389/fcvm.2026.1965415

Keywords: cardiovascular magnetic resonance, HFpEF, heart failure, cine imaging, global longitudinal strain, left atrial volume index, cardiac imaging, Routine, cine, cardiovascular, magnetic, resonance