This retrospective study included 405 patients with HCM who underwent CMR and were hospitalized at Nanjing Gulou Hospital between January 2019 and December 2024. The main admission diagnoses of these patients covered categories: chest pain, dyspnea, palpitations, syncope, or routine follow-up for HCM. The study protocol was approved by the Institutional Ethics Committee (2024-551-01) with a waiver for informed consent.

Inclusion Criteria (Based on the 2024 American Heart Association guidelines [21]): (a) Septal thickness $\ge$15 mm (or $\ge$13 mm in the presence of a family history of HCM, confirmed by imaging modalities including echocardiography, CMR, or computed tomography (CT)); (b) Age $\ge$18 years.

Exclusion Criteria: (a) Age 18 years; (b) Incomplete clinical or imaging data; (c) Poor image quality, defined as significant artifacts (e.g., motion artifacts, susceptibility artifacts), missing frames, or insufficient resolution that impeded accurate assessment of cardiac structures; (d) Severe comorbidities, including active infection, malignancy, renal or hepatic dysfunction, acute myocardial infarction; (e) Infiltrative or systemic cardiomyopathies and other conditions associated with secondary hypertrophy, including refractory hypertension, aortic stenosis, and other infiltrative cardiomyopathies.

AF was diagnosed according to the 2023 American College of Cardiology/American Heart Association/American College of Chest Physicians/Heart Rhythm Society (ACC/AHA/ACCP/HRS) AF diagnostic guidelines [25]. The diagnostic criteria were: (a) ECG findings: A single-lead ECG recording for $\ge$30 seconds or a 12-lead ECG for $\ge$10 seconds demonstrating abnormal atrial waveforms, characterized by the absence of P waves, presence of irregular fibrillatory waves (f-waves) with variable size, shape, and timing, accompanied by irregular RR intervals; (b) AF status: Only patients diagnosed with AF during the current hospital admission were included, without further subclassification into paroxysmal or persistent forms. Patients with a pre-existing AF diagnosis prior to the current admission were excluded to focus on new-onset AF during hospitalization.

CMR scans were performed during hospitalization using a 3.0-T MRI system (Philips Healthcare, Best, Netherlands) in accordance with standardized imaging protocols recommended by the Society for Cardiovascular Magnetic Resonance (SCMR) [26]. All scans were completed during sinus rhythm to ensure accurate measurement of cardiac volume and functional parameters. Scans were acquired at end-expiration using vector ECG-gated techniques with the patient in the supine position. A balanced turbo field echo (B-TFE) movie sequence was used to capture two-, triple-, and four-lumen long-axis views of the left ventricle (LV), as well as successive short-axis slices. For T1 mapping imaging, a breath-hold modified MOLLI (Modified Look-Locker Inversion Recovery) sequence was employed to acquire three short-axis slices at the basal, midventricular, and apical levels. Enhanced T1 measurements were performed 10–15 minutes after intravenous administration of 0.2 mmol/kg gadolinium-based contrast agent (Magnevist, Bayer Healthcare, Berlin, Germany). Images were then translated to the hospital’s Picture Archiving and Communication System (PACS). All CMR analyses were independently conducted by two experienced radiologists using PACS and CVI42 software (version 6.1, Circle Cardiovascular Imaging, Alberta, Canada) (Fig. 1).

Fig. 1.

Typical cardiac magnetic resonance imaging (CMRI) presentation in patients with hypertrophic cardiomyopathy (HCM). This figure shows representative cardiac MRI images and contoured images obtained from patients with HCM, utilizing cinematic MRI technology to precisely depict the structural and functional characteristics of the heart in various anatomical planes. Where (a–d) correspond to the short-axis (SAX), two-chamber (2CH), three-chamber (3CH), and four-chamber (4CH) views of the original cinematic MRI, respectively; (e–h) corresponding to the SAX, 2CH, 3CH, and 4CH views of the cinematic MRI contoured in the original image, respectively.

Cardiac structure: right atrial diameter (RAD) anteroposterior, LA diameter (LAD) anteroposterior, left ventricular wall thickness (LVWT), and interventricular septum thickness (IVST) were measured using the short cine image.

Functional parameters: MAPSE (assessing left ventricular longitudinal systolic function, including inferior, anterior, lateral, and septal walls) and TAPSE (assessing right ventricular systolic function)—were measured using two- and four-chamber cine images.

Other parameter: LV and RV overall strain—global circumferential strain (GCS), global longitudinal strain (GLS), and global radial strain (GRS)—measured by FT-CMR. Pre-contrast T1 values, post-contrast T1 values, and extracellular volume (ECV) were obtained using the tissue T1 mapping function (see Supplementary Table 1 for the specific parameter list).

In order to avoid the limitations of traditional models that rely solely on 2D ultrasound, this study integrated clinical baseline data with multi-dimensional CMR parameters to develop a predictive model through the following steps:

All statistical analyses were conducted using R software (version 4.3.2, R Foundation for Statistical Computing, Vienna, Austria). Continuous variables were summarized as mean $\pm$ standard deviation (Mean $\pm$ SD) for normally distributed data, or as median (M) and interquartile range (Q1–Q3) for non-normally distributed data. Categorical variables were presented as counts and percentages. Normality was assessed using the Shapiro-Wilk test. Between-group comparisons were performed using independent sample t-tests or Mann-Whitney U tests for continuous variables, and chi-square or Fisher’s exact tests for categorical variables, as appropriate. All statistical tests were two-sided, with p 0.05 considered statistically significant.

This study enrolled 405 patients with HCM, and the participant screening process was illustrated in (Fig. 2). The baseline characteristics of the AF and non-AF groups are as follows: (1) Significant differences were observed between the two groups regarding gender, age, eosinophil count, aspartate aminotransferase (AST), lactate dehydrogenase (LDH), direct bilirubin, cholinesterase, urea, creatinine, estimated glomerular filtration rate (eGFR), free triiodothyronine (FT3), and free thyroxine (FT4) (all p 0.05) (Table 1); For CMR parameters, significant intergroup differences between AF and non-AF groups were observed in all measured indices except LVWT, left ventricular end-systolic volume (LVESV), and right ventricular end-systolic volume (RVESV) (all p 0.05; Table 2). The incidence of AF showed no significant difference between the training and validation sets (21.4% vs. 20.8%, p $>$ 0.05), with comparable baseline demographics and CMR metrics across groups (p values 0.052–1.000) (Supplementary Table 4). Notable exceptions were MAPSE Septal (p = 0.035), MAPSE Anterior (p = 0.016), and MAPSE Inferior (p = 0.034) (Supplementary Table 4). Such intergroup variations in segmental MAPSE potentially compromised the model’s generalizability in the validation set; however, LASSO regression was applied to exclude MAPSE features with significant intergroup differences and minimal AF relevance, thereby mitigating the model’s dependency on training-set-specific distributions and enhancing its clinical applicability.

Fig. 2.

Flow chart for patient selection. AF, Atrial Fibrillation.

In recent years, the field of AF risk prediction in cardiomyopathy has witnessed substantial advancements. Multiple studies [28, 29, 30, 31] highlight the pivotal role of CMR in evaluating left heart structure/function, predicting AF onset, and guiding clinical management. Marstrand et al. [32] have confirmed that reduced left ventricular ejection fraction (LVEF) is strongly associated with increased AF risk in HCM, a finding mechanistically linked to hemodynamic imbalance and atrial remodeling secondary to systolic dysfunction. Additionally, Schmidt et al. [33] reveal a correlation between LVESV and AF via Mendelian randomization analysis. Building on these findings, our study introduced LVESV as a sensitive marker of early systolic reserve in HCM, overcoming LVEF’s limitation of relying on dynamic ejection fraction changes to enable earlier detection of systolic dysfunction. Notably, left atrial enlargement serves as a key prognostic factor for AF recurrence after ablation, with MaxLAV acting as an independent predictor (OR = 1.016, 95% CI: 1.004–1.029, p = 0.010) [13]—consistent with den Uijl et al. [34], who identify left atrial volume (LA) as the strongest independent predictor of AF recurrence (OR = 1.019, p = 0.018). By integrating LVESV with atrial remodeling indices, our study addresses a critical gap in traditional models (e.g., HCM-AF calculator) [4, 5] by incorporating CMR-derived ventricular-atrial coupling markers, thus providing a more comprehensive AF risk assessment framework.

The characteristic asymmetric left ventricular hypertrophy in HCM significantly impairs interventricular septum function. As a core parameter for evaluating left ventricular longitudinal systolic function [14], MAPSE septal exhibits distinct abnormalities in HCM. Doesch et al. [15] have demonstrated that septal MAPSE is significantly lower in HCM patients compared to healthy controls (1.57 $\pm$ 0.24 vs. 1.31 $\pm$ 0.25; p = 0.012). Our multiparametric CMR analysis further revealed that septal MAPSE served as an independent predictor of AF (OR = 0.850, 95% CI: 0.736–0.976, p = 0.023) in the overall HCM cohort, with differential predictive values in HOCM and HNCM subtypes—findings that underscore the utility of dynamic MAPSE changes for AF risk stratification.

This study innovatively integrated RAD anteroposterior and TAPSE into the predictive model, highlighting the right heart’s independent role in AF pathogenesis. Prabhu et al. [35] show that electrical/structural remodeling in the RA correlates with LA remodeling in AF patients, supporting interatrial crosstalk. Our data showed that RAD anteroposterior was a significant AF predictor (OR = 1.82, 95% CI 1.13–3.01, p = 0.016), while TAPSE independently predicted AF risk (OR = 0.919, 95% CI 0.852–0.987, p = 0.022). Mechanistically, reduced TAPSE reflects impaired right ventricular-pulmonary artery (RV-PA) coupling, which exacerbates RA pressure load and triggers AF [36]—a mechanism supported by a retrospective study showing that lower TAPSE/PASP ratios increase AF risk [19]. Doesch et al. [15] further validate the association between decreased TAPSE and right atrial dilation in AF patients. This integration of right heart parameters confirms the RV-PA coupling$\to$RA load$\to$AF pathway, offering a novel paradigm for comprehensive risk stratification in HCM.

HOCM patients exhibit significantly restricted left ventricular diastolic function due to elevated left ventricular outflow tract gradient (LVOTG $\ge$30 mmHg). This hemodynamic abnormality drives left atrial dilation and fibrosis by increasing left atrial afterload [18, 34]. This study demonstrated that LAEF, a core indicator of left atrial active emptying capacity, is independently associated with AF occurrence (OR = 0.900, p 0.001), possibly involving compensatory decline in left atrial contractile function with electromechanical delay [6, 37]. Notably, hemodynamic disorders caused by interventricular septal hypertrophy and altered mitral valve spatial relationships in HOCM significantly reduce MAPSE Septal [23, 38]. In contrast, HNCM patients lack significant left ventricular outflow tract obstruction but still exhibit impaired diastolic function and mild hemodynamic abnormalities due to myocardial hypertrophy and ventricular remodeling [24, 39]. These alterations can lead to persistent elevation in right atrial pressure, promoting right atrial dilation and fibrosis, thereby reducing RAEF [24, 39, 40]. This “mild hemodynamic disturbance—right atrial remodeling” pathway is consistent with the mechanism proposed by Croon et al. [40], who reported that RAEF combined with RVEF influences AF risk in genetic variant carriers of HCM. Conversely, HOCM patients demonstrate more pronounced left atrial remodeling due to elevated afterload, while right atrial dysfunction may be partially compensated by adaptive mechanisms [21, 38]. These findings collectively highlight the need for obstruction status-based differential clinical assessment: HNCM requires focused monitoring of right atrial structure and function, as mild hemodynamic disturbances gradually accumulate into right atrial remodeling and AF; HOCM necessitates prioritizing optimization of left atrial-right ventricular hemodynamic coupling to interrupt atrial electrophysiological abnormalities from pressure overload. This mechanism-based stratified assessment strategy provides precise imaging and functional targets for personalized AF prevention in HCM patients.

The HCM-AF Risk Score we developed offers clinicians a straightforward tool to stratify AF risk in HCM patients during standard CMR assessments. By integrating parameters already routinely captured in most CMR protocols—such as LVESV and RAEF—this score requires no extra scanning time or additional costs to implement, enhancing its real-world usability. We see particular value in applying this score across three key clinical scenarios: First, it could help flag high-risk patients who might benefit from more rigorous rhythm monitoring (e.g., prolonged Holter monitoring or implantable loop recorders) to catch AF earlier. Second, it may support shared decision-making when anticoagulation prophylaxis is being considered for patients with ambiguous clinical indications. Third, it could refine pre-procedural risk assessment ahead of surgeries like myectomy, where undiagnosed AF might significantly influence perioperative care plans. Of course, broader adoption hinges on future validation. Prospective testing in multi-center cohorts and research exploring how this score impacts actual patient outcomes will be critical to integrating it into routine clinical practice.

First, the study’s single-center retrospective design may limit the external validity of the conclusions. Further validation through multicenter prospective studies is required to confirm generalizability. Second, the absence of long-term follow-up data means that the model’s clinical prognostic value remains unestablished. Future studies incorporating longitudinal follow-up are necessary to evaluate its predictive utility for clinical outcomes. Third, the exclusion of patients with pre-existing AF and the focus on in-hospital new-onset AF may limit the generalizability of our findings to the broader HCM population with prevalent or historical AF. Fourth, this study did not include data on mitral regurgitation or tricuspid regurgitation, which are known to be associated with atrial fibrillation and are particularly relevant in obstructive HCM due to their impact on left atrial remodeling and hemodynamic load. Future studies incorporating valvular assessment would enhance the comprehensiveness of AF risk prediction models.