Hypertrophic cardiomyopathy (HCM) is the most common cause of sudden cardiac death (SCD) in the young and is a significant cause of sudden death in adults (Maron, 2004). The disease, with an incidence of 1 in 500, presents with progressive myocardial fibrosis which can create substrates for ventricular arrhythmias (VAs) leading to SCD in patients who are typically asymptomatic (Galati et al., 2016; Olivotto et al., 2012). Implantable cardioverter defibrillator (ICD) deployment, a procedure that carries risk of potential complications and morbidity, is used as primary prevention of SCD due to VA in patients with HCM (Lambiase et al., 2016; Jayatilleke et al., 2004). However, current risk stratification criteria outlined by the American College of Cardiology Foundation (ACCF)/American Heart Association (AHA) and European Society of Cardiology (ESC) fail to accurately identify all patients at risk for SCD, leading to suboptimal rates of appropriate ICD implantation (Gersh et al., 2011; O’Mahony et al., 2014; Schinkel et al., 2012). Thus, many HCM patients receive ICDs without deriving any health benefits, while others are not adequately protected. The development of accurate means to stratify SCD risk due to VA in HCM patients for guidance of ICD deployment is an important unmet clinical need.

Cardiac magnetic resonance (CMR) imaging with late gadolinium enhancement (LGE) has unparalleled capability in the detection and quantification of scar and dense fibrosis (Prakosa et al., 2014). In HCM, myocardial fibrosis takes the form of both dense (focal) and diffuse fibrosis, with histopathological evidence showing diffuse fibrosis as the hallmark feature of the disease (Galati et al., 2016). Diffuse fibrosis, however, is not well captured by standard LGE-CMR. Instead, postcontrast T1 mapping, a parametric imaging modality, has been used to visualize diffuse fibrosis in patients with HCM (Chu et al., 2017; Ellims et al., 2012). We have previously developed a computational modeling approach (virtual heart) to predict SCD risk due to VA in postinfarction patients (Arevalo et al., 2016). We hypothesized that a new personalized virtual-heart technology, one that entails constructing fusion electrophysiological models based on the distribution of both dense and diffuse fibrosis, as acquired by the two different CMR modalities, would be predictive of the propensity of the HCM-remodeled substrate to VAs and could thus be used to assess SCD risk due to VA in this patient population.

The goal of this study is to create a personalized virtual-heart approach based on the combination of postcontrast T1 mapping and LGE-CMR and to employ it (1) to analyze how HCM-specific remodeling promotes arrhythmogenesis and (2) in a targeted strategy to forecast risk of VA in HCM patients. In a proof-of-concept patient cohort, we assess the predictive capability of the approach as compared to that of other clinical metrics for VA risk prediction in HCM.

The new approach to analyzing arrhythmogenic propensity in HCM patients developed here involved creating three-dimensional (3D) patient-specific electrophysiological ventricular models based on fusing data from LGE-CMR and postcontrast T1 mapping. Each model thus represented the personalized distribution of focal fibrosis (scar) and diffuse fibrosis. VA inducibility in each HCM patient’s substrate was probed to determine VA risk for the patient and to understand the mechanisms of arrhythmogenesis, and specifically, the contribution of the individualized diffuse fibrosis distribution, which is rarely assessed in these patients. Conceptual overview of our approach to analyzing the arrhythmogenic propensity of HCM patient hearts is presented in Figure 1A.

Figure 1

Download asset Open asset

Assessment of HCM structural remodeling using LGE-T1 geometrical models

To reconstruct the geometrical model of each patient’s heart, LGE-CMR and postcontrast T1 mapping images were combined, creating a personalized LGE-T1 fusion model of HCM ventricular geometry and structural remodeling. Figure 1B presents the ‘fusing’ process, in which an initial reconstruction of ventricular geometry and scar/fibrosis was performed from the LGE-CMR images using standard ‘one-size-fits-all’ thresholds, and then the relaxation times from the short-axis T1 map were used to define personalized signal intensity thresholds to delineate areas of diffuse fibrosis and scar (see Methods for detailed description). The personalized thresholds were unique to each patient. The additional personalization of the geometrical model furnished by the usage of the T1 mapping data ensured a comprehensive representation of the individualized structural remodeling in each patient’s heart.

Once the geometrical models were reconstructed, they were analyzed to determine whether the level (amount) and/or distribution of structural remodeling discriminate between patients with and without clinical VA. The level of regional hypertrophy was first assessed, as measured by the wall thickness of the heart models. No statistically significant difference in regional hypertrophy was found at the septum (p = 0.61), anterior wall (p = 0.84), posterior wall (p = 0.94), and apex (p = 0.73) between heart models of patients with and without clinical VA, as shown in Table 2. These results indicated that the level of hypertrophy does not discriminate between arrhythmogenic and nonarrhythmogenic substrates in HCM patients.

Table 2

	Patients without VA, n = 13	Patients with VA, n = 13	p value
Wall thickness (mean ± standard deviation [SD])
Septum, mm	11.3 ± 8.4	13.1 ± 9.2	0.61
Anterior, mm	11.3 ± 7.4	10.7 ± 7.3	0.84
Posterior, mm	11.1 ± 7.2	11.3 ± 7.3	0.94
Apex, mm	7.9 ± 5.3	7.2 ± 4.8	0.73

1. p values were calculated using Student’s t-test (p ≤ 0.05 considered statistically significant).

Figure 2A and B presents a comparison between geometrical heart models of two patients (one with clinical VA and another without) reconstructed by combining LGE-CMR with T1 mapping, and by using LGE-CMR only. In the latter models, the accepted ‘one-size-fits-all’ thresholds of three and five times the SD of the low-intensity mean were used to reconstruct dense fibrosis (scar) and diffuse fibrosis distributions (see Methods). In the former models, patient-specific thresholds from the T1 mapping were used to delineate dense and diffuse fibrosis. As evident from the figure, using patient-specific signal intensity thresholding from the T1 map resulted in a significantly higher amount of diffuse fibrosis in these two models (42.9 ± 3.4% vs 9.8 ± 0.1%).

Figure 2

Download asset Open asset

Figure 2—source data 1

Spreadsheet including source data underlying Figure 2.

https://cdn.elifesciences.org/articles/73325/elife-73325-fig2-data1-v1.xlsx

Download elife-73325-fig2-data1-v1.xlsx

For all HCM LGE-T1 fusion models, the average threshold for diffuse fibrosis, T_Diffuse, was 1.1 ± 0.7, significantly different from the corresponding LGE ‘one-size-fits-all value’, 3 SD. The average threshold for dense fibrosis, T_Dense, was 5.1 ± 0.5, not a significant change from the original 5 SD. The personalized threshold adjustment did not therefore result in a significant change in the amount of dense fibrosis for LGE-T1 models compared to LGE-only models (averages of 3.8 ± 2.3 vs 3.2 ± 1.3, p = 0.30). However, it resulted in a significant change in diffuse fibrosis across all models, as illustrated in Figure 2C (40.5 ± 9.4% for LGE-T1 vs 8.9 ± 1.7% for LGE only, p < 0.0001).

No statistical differences were found in the amounts of diffuse fibrosis between LGE-T1 models with and without clinical VA (p = 0.53, confidence interval [CI: 36.8, 44]) and between LGE-only models with and without clinical VA (p = 0.94, CI: [8.25, 9.53]; Figure 1B); also, no statistical difference was found in the amount of scar (3.7 ± 2.2% vs 3.8 ± 2.5%, p = 0.91 for LGE-T1 and 3.4 ± 1.2% vs 3.0 ± 1.5%, p = 0.53 for LGE-only models). These results indicate that the imaging characteristics of HCM structural remodeling, as visualized by the combination of LGE-CMR and T1 mapping, cannot be used to discriminate between patients who will and will not develop clinical VA.

For each geometrical model used in this study, the amount of diffuse fibrosis in each LGE-T1 and LGE-only model (Figure 2—source data 1).

Assessment of propensity to VA in HCM LGE-T1 virtual-heart models

Once the geometrical models of all HCM patients were reconstructed, electrophysiological models were generated and used to assess the individualized propensity to VA by pacing from distributed ventricular sites, representing potential ectopy. Full detail is in Methods. A total of 182 simulations ([26 patient heart] × [7 pacing locations]) were performed to probe propensity to VA induction in this cohort. To be able to better understand the role of T1-based diffuse fibrosis in arrhythmogenesis, we also repeated the simulations with LGE-only models.

Figure 3 presents reentrant arrhythmias induced (from sites marked with stars) in three LGE-T1 virtual hearts from patients with known clinical VAs. In all three cases, a single VA morphology was induced. In Figure 3, left, the VA localized in a region of interdigitated diffuse and dense fibrosis. In Figure 3, middle, there was a figure-of-eight reentry in a transmural region of diffuse fibrosis. In Figure 3, right, the VA shown was induced from two different pacing sites, one in the basal lateral and another in the inferoseptal wall and persisted also in a region of interdigitated diffuse and dense fibrosis.

Figure 3

Download asset Open asset

For each geometrical model that reentry was induced, the number of unique VA morphologies and amount of diffuse fibrosis in each LGE-T1 and LGE-only model (Figure 4—source data 1).

Of the 26 LGE-T1 models, 14 were found inducible for VA in simulations. In contrast, only 12 LGE-only models were found inducible, indicating that the presence of diffuse fibrosis leads to increased VA inducibility. Figures 4 and 5 explore the mechanistic contributions to increased VA vulnerability in models with T1-based diffuse fibrosis.

Figure 4

Download asset Open asset

Figure 4—source data 1

Spreadsheet including source data underlying Figure 4.

https://cdn.elifesciences.org/articles/73325/elife-73325-fig4-data1-v1.xlsx

Download elife-73325-fig4-data1-v1.xlsx

Figure 5

Download asset Open asset

Figure 5—source data 1

Spreadsheet including source data underlying Figure 5.

https://cdn.elifesciences.org/articles/73325/elife-73325-fig5-data1-v1.xlsx

Download elife-73325-fig5-data1-v1.xlsx

Figure 4A presents the number of unique VA morphologies induced by the pacing protocol. LGE-T1 models had a total of 24 unique VAs induced in them (out of total 32 VA episodes induced in the LGE-T1 models); in each model, there were between 1 and 3 different VA morphologies. LGE-only models had a total of 15 unique VAs (with a total of 17 VA episodes induced in these models), with only 1 or 2 distinct VA morphologies induced per model. These results indicate that the presence of diffuse fibrosis as reconstructed from T1 mapping increases the number of unique VAs in each substrate. Figure 4B correlates the amount of diffuse fibrosis and the number of unique VAs (R = 0.40, p = 0.157) in LGE-T1 inducible models. The moderate negative correlation indicates that the distribution of diffuse fibrosis is more important than its amount as the mechanism of VA inducibility in the HCM-remodeled substrate. Figure 4C presents two bullseye plots with the 7 AHA regions in which pacing sites were located; shown are the number of pacing sites in each segment that elicited VAs in LGE-T1 and LGE-only inducible models. In the LGE-T1 models, out of the 98 pacing sites (7 pacing sites per each of the 14 inducible models), 32 (33%) resulted in VA induction. In contrast, out of 84 pacing sites in the 12 LGE-only inducible models, 17 (20%) resulted in VA induction. Thus, the presence of T1-based diffuse fibrosis renders the substrate inducible from a larger number of ectopic locations, contributing to the overall increased vulnerability to VA. Interestingly, the sector with the pacing sites that induced most VAs (mid-anteroseptal) and that with least (basal inferolateral) were the same in LGE-T1 and LGE-only models, indicating that the additional T1-based diffuse fibrosis localizes to the sectors with arrhythmogenic substrate in the LGE-only models. Overall, the distribution of pacing sites is the same (with small exception in the basal regions), but the number of sites per sector increased with the presence of diffuse fibrosis.

Figure 5 explores the contribution of T1-based diffuse fibrosis to VA inducibility by comparing arrhythmogenesis in individual models. Panel A shows, for the 13 patients with clinical VAs, the number of distinct VAs per patient model. First, LGE-T1 modeling documented correctly VA occurrence in the digital substrates of 11 of the 13 patients with clinical VAs (compared to 9 LGE-only correct VA predictions); thus, the representation of T1-based diffuse fibroses increased the fidelity of the HCM virtual-heart approach. Second, while the number of unique VA morphologies per inducible virtual heart increased in LGE-T1 models as compared to LGE only (consistent with data in Figure 4A), with LGE-T1 models having maximum three unique VAs (vs 2 in LGE only) in the cohort, this plot points to interesting VA dynamics in individual substrates. Patients 5, 6, and 9 had the same level of arrhythmogenicity of the substrate (1 VA induced) regardless of the presence of T1-based diffuse fibrosis. An example of VA dynamics in these models is shown in Figure 5B (patient 6). The additional diffuse fibrosis did not alter the location or direction of VA reentry; the reentry occurred in a region of dense scar in both types of models. Although the T1-based diffuse fibrosis (36.2%) in this patient augments the existing diffuse fibrosis of the LGE-only model (9.6%), there are no additional VA morphologies. However, the activation dynamics were altered, with propagation being less organized.

In the reminder of the clinical-VA patients in Figure 5A, the presence of T1-based diffuse fibrosis resulted in the occurrence of additional VA(s), on top of that (those) also present in the LGE-only model. Furthermore, there was not a strict correspondence between the VAs in the LGE-only models and the equivalent ones in the LGE-T1 virtual heart in terms of VA locations and dynamics. This indicates that the arrhythmogenic substrate changes in a global fashion when T1-based diffuse fibrosis in considered. An example is presented in Figure 5C (patient 9).

As VAs in HCM patients can occur under different circumstances and be documented by different means, there is no invasive clinical mapping data for these patients regarding the location and morphologies of the clinical VAs. However, two patients in this cohort underwent clinical electrophysiology studies that identified episodes of VA. The simulated VAs in the LGE-T1 models matched the locations of the clinical VAs as documented by the chart review.

It remains unclear why patients 1 and 2 had noninducible LGE-T1 substrates despite having clinical VAs. The amount of diffuse fibrosis in these two substrates (42.4 ± 12.2%) was similar to that in the 11 inducible substrates (41.2 ± 10.9%); similar was the finding regarding dense scar. Additionally, the maximum wall thickness and the thickness distribution fall into the same ranges as the averages of the cohort. It is possible that there have been other factors, including electrophysiological remodeling that the fusion substrate-based modeling approach presented here cannot capture. Despite incorrect prediction in 2 out of 26 patients in the HCM cohort, the VA risk prediction capabilities of the LGE-T1 virtual-heart approach significantly surpassed those of any current clinical risk assessment approaches, as detailed in the next section.

For each geometrical model with clinical VA, the number of unique VA morphologies in each LGE-T1 and LGE-only model (Figure 5—source data 1).

In this study, we presented a new personalized virtual-heart approach for assessing arrhythmia risk in patients with HCM, which could be used in guiding clinical decisions for prophylactic ICD implantation. Our technology uses multiscale computational models of patients’ hearts reconstructed on the basic of the fusion of imaging data from LGE-CMR and T1 mapping. With the inclusion of information from postcontrast T1 mapping, a quantitative and parametric imaging modality, extensive diffuse fibrotic remodeling, which is a hallmark of HCM, is adequately represented. Here, this is done by adjusting the diffuse fibrosis intensity-based thresholds in model construction based on the T1 maps, while preserving the identification of dense scar from LGE-MRI. These are the first personalized heart models created with data from different types of CMR; previous personalized modeling approaches for VA assessment have utilized LGE-MRI scans in reconstructing model geometry/structure (Prakosa et al., 2014; Cartoski et al., 2019; Arevalo et al., 2016). Once constructed, the mechanistic personalized electrophysiological models were used to analyze how HCM-induced remodeling, and specifically the presence of diffuse fibrosis promotes arrhythmogenesis. Finally, the capability of our approach to forecast future VA events was assessed in the cohort of 26 HCM patients.

The presence of diffuse fibrosis has been suggested previously as a potential factor in increased risk of VA, in addition to focal scar. Previous studies have found associations between diffuse fibrosis and VA in nonischemic dilated cardiomyopathies (Nakamori et al., 2018) and mitral valve prolapse (Bui et al., 2017). Additionally, diffuse ventricular fibrosis has been found to increase left atrial pressure, and may be a marker of atrial fibrillation recurrence postablation (Begg et al., 2020). However, its contribution to arrhythmogenic propensity in HCM patients has never been assessed before. This study found that the presence of T1-based diffuse fibrosis resulted in the occurrence of new VAs, in addition to those arising from scar (as assessed by signal heterogeneities in LGE-MRI). Furthermore, T1-based diffuse fibrosis distribution rendered the substrate inducible from a larger number of ectopic locations, contributing to the overall increased vulnerability to VA. The arrhythmogenic propensity of the ventricular substrate with diffuse fibrosis is patient specific; it is the amount of fibrosis and its distribution (shape and location) that determine inducibility of arrhythmia.

In this retrospective proof-of-concept HCM study, the personalized LGE-T1 virtual-heart technology demonstrated excellent performance in forecasting future VA events in HCM patients, achieving 84.6%, 76.9%, and 80.1% sensitivity, specificity, and accuracy, respectively. It outperformed both risk models used in current clinical practice, the ACCF/AHA and ESC models. Indeed, all 26 HCM patients in our study were deemed at high risk for SCD by the ACCF/AHA criteria and received ICDs for primary prevention, but only 13 patients, that is 50% of the cohort, actually experienced VA (appropriate ICD firing). Should our LGE-T1 virtual-heart technology be proven to be a superior risk predictor in larger clinical studies, it would advance the management of patients with this complex disease, helping to ensure that those at high risk for VA are adequately protected by ICDs and that unnecessary ICD implantations and the associated device complications are minimized.

The HCM virtual-heart technology’s ability to comprehensively evaluate substrate arrhythmogenicity, as probed by rapid pacing delivered at a number of uniformly distributed ventricular locations, is paramount to its superior performance. Even when using only LGE-CMR in model construction, which reliably detects focal scar (dense fibrosis) but underestimates the amount of nonischemic fibrotic remodeling, our technology still offers VA risk assessment that is superior to the clinical risk models. However, the use of T1 maps in model construction confers a higher level of personalization in each patient’s heart model as compared to LGE only (i.e. personalized thresholds for segmentation), which ultimately translates into superior predictive capability.

HCM is a genetic disease that progresses throughout the life of the patient, and a cardiac event might be a phenotypic expression of the disease at any point of time. Therefore, we envision that in the clinical application of our technology, patients would be reimaged at different time points and risk assessment repeated to account for changes in arrhythmia susceptibility over time as the diseased heart remodels.

The technology developed here charts a new direction in the use of biophysically detailed heart modeling in the prognosis of rhythm disorders. A number of different imaging modalities used in patient assessment such as positron emission tomography or single-photon emission computerized tomography could also be integrated with LGE-CMR to construct hybrid heart computational models. Combining such computational approaches with machine learning techniques (Shade et al., 2020a; Shade et al., 2021) will enable the incorporation of additional patient clinical data, such as genetic information, phenotypic characterization, time series, such as electrocardiography, and fibrotic distribution in the diagnosis and treatment of complex cardiac diseases with adequately sized patient cohorts.

Where possible (Figures 2, 4, 5), the raw numerical data used to generate the figures have been provided. The segmented MR images of the left ventricles used to generate the finite-element meshes for the computational models used in this study are available on Dryad (https://doi.org/10.5061/dryad.bk3j9kdd0). Patient-derived data, including the MR images, related to this article are not publicly available to respect patient privacy. Interested parties wishing to obtain these data for non-commercial reuse should contact the corresponding author via email. Upon all reasonable requests for access to these data, the corresponding author will work to pursue negotiation of a Data Transfer and Use Agreement with the requesting party, administrators at the requesting party's institution, Johns Hopkins University and Hospital, and the relevant Institutional Review Boards. Documentation and instructions on the use of the openCARP cardiac electrophysiology simulator and meshalyzer visualization software (both available via https://opencarp.org/) can be used to precisely reproduce the computational protocol applied to patient-specific left ventricular models in this study.

1. 1. O'Hara R

   (2021) Dryad Digital Repository

   Left ventricular MRI segmentations.

   https://doi.org/10.5061/dryad.bk3j9kdd0

1. 1. Arevalo HJ
   2. Vadakkumpadan F
   3. Guallar E
   4. Jebb A
   5. Malamas P
   6. Wu KC
   7. Trayanova NA

   (2016) Arrhythmia risk stratification of patients after myocardial infarction using personalized heart models

   Nature Communications 7:11437.

   https://doi.org/10.1038/ncomms11437

   • PubMed
   • Google Scholar
2. 1. Bayer JD
   2. Blake RC
   3. Plank G
   4. Trayanova NA

   (2012) A novel rule-based algorithm for assigning myocardial fiber orientation to computational heart models

   Annals of Biomedical Engineering 40:2243–2254.

   https://doi.org/10.1007/s10439-012-0593-5

   • PubMed
   • Google Scholar
3. 1. Begg GA
   2. Swoboda PP
   3. Karim R
   4. Oesterlein T
   5. Rhode K
   6. Holden AV
   7. Greenwood JP
   8. Shantsila E
   9. Lip GYH
   10. Plein S
   11. Tayebjee MH

   (2020) Imaging, biomarker and invasive assessment of diffuse left ventricular myocardial fibrosis in atrial fibrillation

   Journal of Cardiovascular Magnetic Resonance 22:13.

   https://doi.org/10.1186/s12968-020-0603-y

   • PubMed
   • Google Scholar
4. 1. Bui AH
   2. Roujol S
   3. Foppa M
   4. Kissinger KV
   5. Goddu B
   6. Hauser TH
   7. Zimetbaum PJ
   8. Ngo LH
   9. Manning WJ
   10. Nezafat R
   11. Delling FN

   (2017) Diffuse myocardial fibrosis in patients with mitral valve prolapse and ventricular arrhythmia

   Heart 103:204–209.

   https://doi.org/10.1136/heartjnl-2016-309303

   • PubMed
   • Google Scholar
5. 1. Cartoski MJ
   2. Nikolov PP
   3. Prakosa A
   4. Boyle PM
   5. Spevak PJ
   6. Trayanova NA

   (2019) Computational Identification of Ventricular Arrhythmia Risk in Pediatric Myocarditis

   Pediatric Cardiology 40:857–864.

   https://doi.org/10.1007/s00246-019-02082-7

   • PubMed
   • Google Scholar
6. 1. Chu LC
   2. Corona-Villalobos CP
   3. Halushka MK
   4. Zhang Y
   5. Pozzessere C
   6. Kamel IR
   7. Pozios I
   8. Van Der Geest RJ
   9. Gai N
   10. Abraham RM
   11. Abraham TP
   12. Bluemke DA
   13. Zimmerman SL

   (2017) Structural and Functional Correlates of Myocardial T1 Mapping in 321 Patients With Hypertrophic Cardiomyopathy

   Journal of Computer Assisted Tomography 41:653–660.

   https://doi.org/10.1097/RCT.0000000000000564

   • PubMed
   • Google Scholar
7. 1. Coppini R
   2. Ferrantini C
   3. Yao L
   4. Fan P
   5. Del Lungo M
   6. Stillitano F
   7. Sartiani L
   8. Tosi B
   9. Suffredini S
   10. Tesi C
   11. Yacoub M
   12. Olivotto I
   13. Belardinelli L
   14. Poggesi C
   15. Cerbai E
   16. Mugelli A

   (2013) Late sodium current inhibition reverses electromechanical dysfunction in human hypertrophic cardiomyopathy

   Circulation 127:575–584.

   https://doi.org/10.1161/CIRCULATIONAHA.112.134932

   • PubMed
   • Google Scholar
8. 1. Ellims AH
   2. Iles LM
   3. Ling L
   4. Hare JL
   5. Kaye DM
   6. Taylor AJ

   (2012) Diffuse myocardial fibrosis in hypertrophic cardiomyopathy can be identified by cardiovascular magnetic resonance, and is associated with left ventricular diastolic dysfunction

   Journal of Cardiovascular Magnetic Resonance 14:76.

   https://doi.org/10.1186/1532-429X-14-76

   • PubMed
   • Google Scholar
9. 1. Ellims AH
   2. Iles LM
   3. Ling L
   4. Chong B
   5. Macciocca I
   6. Slavin GS
   7. Hare JL
   8. Kaye DM
   9. Marasco SF
   10. McLean CA
   11. James PA
   12. du Sart D
   13. Taylor AJ

   (2014) A comprehensive evaluation of myocardial fibrosis in hypertrophic cardiomyopathy with cardiac magnetic resonance imaging: linking genotype with fibrotic phenotype

   European Heart Journal. Cardiovascular Imaging 15:1108–1116.

   https://doi.org/10.1093/ehjci/jeu077

   • PubMed
   • Google Scholar
10. 1. Galati G
    2. Leone O
    3. Pasquale F
    4. Olivotto I
    5. Biagini E
    6. Grigioni F
    7. Pilato E
    8. Lorenzini M
    9. Corti B
    10. Foà A
    11. Agostini V
    12. Cecchi F
    13. Rapezzi C

    (2016) Histological and Histometric Characterization of Myocardial Fibrosis in End-Stage Hypertrophic Cardiomyopathy: A Clinical-Pathological Study of 30 Explanted Hearts

    Circulation. Heart Failure 9:e003090.

    https://doi.org/10.1161/CIRCHEARTFAILURE.116.003090

    • PubMed
    • Google Scholar
11. 1. Gersh BJ
    2. Maron BJ
    3. Bonow RO
    4. Dearani JA
    5. Fifer MA
    6. Link MS
    7. Naidu SS
    8. Nishimura RA
    9. Ommen SR
    10. Rakowski H
    11. Seidman CE
    12. Towbin JA
    13. Udelson JE
    14. Yancy CW
    15. American College of Cardiology Foundation/American Heart Association Task Force on Practice Guidelines

    (2011) 2011 ACCF/AHA Guideline for the Diagnosis and Treatment of Hypertrophic Cardiomyopathy: a report of the American College of Cardiology Foundation/American Heart Association Task Force on Practice Guidelines. Developed in collaboration with the American Association for Thoracic Surgery, American Society of Echocardiography, American Society of Nuclear Cardiology, Heart Failure Society of America, Heart Rhythm Society, Society for Cardiovascular Angiography and Interventions, and Society of Thoracic Surgeons

    Journal of the American College of Cardiology 58:e212–e260.

    https://doi.org/10.1016/j.jacc.2011.06.011

    • PubMed
    • Google Scholar
12. 1. Iles L
    2. Pfluger H
    3. Phrommintikul A
    4. Cherayath J
    5. Aksit P
    6. Gupta SN
    7. Kaye DM
    8. Taylor AJ

    (2008) Evaluation of diffuse myocardial fibrosis in heart failure with cardiac magnetic resonance contrast-enhanced T1 mapping

    Journal of the American College of Cardiology 52:1574–1580.

    https://doi.org/10.1016/j.jacc.2008.06.049

    • PubMed
    • Google Scholar
13. 1. Jayatilleke I
    2. Doolan A
    3. Ingles J
    4. McGuire M
    5. Booth V
    6. Richmond DR
    7. Semsarian C

    (2004) Long-term follow-up of implantable cardioverter defibrillator therapy for hypertrophic cardiomyopathy

    The American Journal of Cardiology 93:1192–1194.

    https://doi.org/10.1016/j.amjcard.2004.01.057

    • PubMed
    • Google Scholar
14. 1. Lambiase PD
    2. Gold MR
    3. Hood M
    4. Boersma L
    5. Theuns DAMJ
    6. Burke MC
    7. Weiss R
    8. Russo AM
    9. Kääb S
    10. Knight BP

    (2016) Evaluation of subcutaneous ICD early performance in hypertrophic cardiomyopathy from the pooled EFFORTLESS and IDE cohorts

    Heart Rhythm 13:1066–1074.

    https://doi.org/10.1016/j.hrthm.2016.01.001

    • PubMed
    • Google Scholar
15. 1. Maron BJ

    (2004) Hypertrophic cardiomyopathy: an important global disease

    The American Journal of Medicine 116:63–65.

    https://doi.org/10.1016/j.amjmed.2003.10.012

    • PubMed
    • Google Scholar
16. 1. Mewton N
    2. Liu CY
    3. Croisille P
    4. Bluemke D
    5. Lima JAC

    (2011) Assessment of myocardial fibrosis with cardiovascular magnetic resonance

    Journal of the American College of Cardiology 57:891–903.

    https://doi.org/10.1016/j.jacc.2010.11.013

    • PubMed
    • Google Scholar
17. 1. Nakamori S
    2. Bui AH
    3. Jang J
    4. El-Rewaidy HA
    5. Kato S
    6. Ngo LH
    7. Josephson ME
    8. Manning WJ
    9. Nezafat R

    (2018) Increased myocardial native T₁ relaxation time in patients with nonischemic dilated cardiomyopathy with complex ventricular arrhythmia

    Journal of Magnetic Resonance Imaging 47:779–786.

    https://doi.org/10.1002/jmri.25811

    • PubMed
    • Google Scholar
18. 1. Olivotto I
    2. Cecchi F
    3. Poggesi C
    4. Yacoub MH

    (2012) Patterns of disease progression in hypertrophic cardiomyopathy: an individualized approach to clinical staging

    Circulation. Heart Failure 5:535–546.

    https://doi.org/10.1161/CIRCHEARTFAILURE.112.967026

    • PubMed
    • Google Scholar
19. 1. O’Hara T
    2. Virág L
    3. Varró A
    4. Rudy Y

    (2011) Simulation of the undiseased human cardiac ventricular action potential: model formulation and experimental validation

    PLOS Computational Biology 7:e1002061.

    https://doi.org/10.1371/journal.pcbi.1002061

    • PubMed
    • Google Scholar
20. 1. O’Mahony C
    2. Jichi F
    3. Pavlou M
    4. Monserrat L
    5. Anastasakis A
    6. Rapezzi C
    7. Biagini E
    8. Gimeno JR
    9. Limongelli G
    10. McKenna WJ
    11. Omar RZ
    12. Elliott PM
    13. Hypertrophic Cardiomyopathy Outcomes Investigators

    (2014) A novel clinical risk prediction model for sudden cardiac death in hypertrophic cardiomyopathy (HCM risk-SCD)

    European Heart Journal 35:2010–2020.

    https://doi.org/10.1093/eurheartj/eht439

    • PubMed
    • Google Scholar
21. 1. Plank G
    2. Zhou L
    3. Greenstein JL
    4. Cortassa S
    5. Winslow RL
    6. O’Rourke B
    7. Trayanova NA

    (2008) From mitochondrial ion channels to arrhythmias in the heart: computational techniques to bridge the spatio-temporal scales

    Philosophical Transactions. Series A, Mathematical, Physical, and Engineering Sciences 366:3381–3409.

    https://doi.org/10.1098/rsta.2008.0112

    • PubMed
    • Google Scholar
22. 1. Prakosa A.
    2. Malamas P
    3. Zhang S
    4. Pashakhanloo F
    5. Arevalo H
    6. Herzka DA
    7. Lardo A
    8. Halperin H
    9. McVeigh E
    10. Trayanova N
    11. Vadakkumpadan F

    (2014) Methodology for image-based reconstruction of ventricular geometry for patient-specific modeling of cardiac electrophysiology

    Progress in Biophysics and Molecular Biology 115:226–234.

    https://doi.org/10.1016/j.pbiomolbio.2014.08.009

    • PubMed
    • Google Scholar
23. 1. Prakosa A
    2. Arevalo HJ
    3. Deng D
    4. Boyle PM
    5. Nikolov PP
    6. Ashikaga H
    7. Blauer JJE
    8. Ghafoori E
    9. Park CJ
    10. Blake RC
    11. Han FT
    12. MacLeod RS
    13. Halperin HR
    14. Callans DJ
    15. Ranjan R
    16. Chrispin J
    17. Nazarian S
    18. Trayanova NA

    (2018) Personalized virtual-heart technology for guiding the ablation of infarct-related ventricular tachycardia

    Nature Biomedical Engineering 2:732–740.

    https://doi.org/10.1038/s41551-018-0282-2

    • PubMed
    • Google Scholar
24. 1. Prassl AJ
    2. Kickinger F
    3. Ahammer H
    4. Grau V
    5. Schneider JE
    6. Hofer E
    7. Vigmond EJ
    8. Trayanova NA
    9. Plank G

    (2009) Automatically generated, anatomically accurate meshes for cardiac electrophysiology problems

    IEEE Transactions on Bio-Medical Engineering 56:1318–1330.

    https://doi.org/10.1109/TBME.2009.2014243

    • PubMed
    • Google Scholar
25. 1. Salvador M
    2. Fedele M
    3. Africa PC
    4. Sung E
    5. Dede’ L
    6. Prakosa A
    7. Chrispin J
    8. Trayanova N
    9. Quarteroni A

    (2021) Electromechanical modeling of human ventricles with ischemic cardiomyopathy: numerical simulations in sinus rhythm and under arrhythmia

    Computers in Biology and Medicine 136:104674.

    https://doi.org/10.1016/j.compbiomed.2021.104674

    • PubMed
    • Google Scholar
26. 1. Schinkel AFL
    2. Vriesendorp PA
    3. Sijbrands EJG
    4. Jordaens LJLM
    5. ten Cate FJ
    6. Michels M

    (2012) Outcome and complications after implantable cardioverter defibrillator therapy in hypertrophic cardiomyopathy: systematic review and meta-analysis

    Circulation. Heart Failure 5:552–559.

    https://doi.org/10.1161/CIRCHEARTFAILURE.112.969626

    • PubMed
    • Google Scholar
27. 1. Shade JK
    2. Ali RL
    3. Basile D
    4. Popescu D
    5. Akhtar T
    6. Marine JE
    7. Spragg DD
    8. Calkins H
    9. Trayanova NA

    (2020a) Preprocedure Application of Machine Learning and Mechanistic Simulations Predicts Likelihood of Paroxysmal Atrial Fibrillation Recurrence Following Pulmonary Vein Isolation

    Circulation. Arrhythmia and Electrophysiology 13:e008213.

    https://doi.org/10.1161/CIRCEP.119.008213

    • PubMed
    • Google Scholar
28. 1. Shade JK
    2. Cartoski MJ
    3. Nikolov P
    4. Prakosa A
    5. Doshi A
    6. Binka E
    7. Olivieri L
    8. Boyle PM
    9. Spevak PJ
    10. Trayanova NA

    (2020b) Ventricular arrhythmia risk prediction in repaired Tetralogy of Fallot using personalized computational cardiac models

    Heart Rhythm 17:408–414.

    https://doi.org/10.1016/j.hrthm.2019.10.002

    • PubMed
    • Google Scholar
29. 1. Shade JK
    2. Prakosa A
    3. Popescu DM
    4. Yu R
    5. Okada DR
    6. Chrispin J
    7. Trayanova NA

    (2021) Predicting risk of sudden cardiac death in patients with cardiac sarcoidosis using multimodality imaging and personalized heart modeling in a multivariable classifier

    Science Advances 7:eabi8020.

    https://doi.org/10.1126/sciadv.abi8020

    • PubMed
    • Google Scholar
30. 1. ten Tusscher K
    2. Panfilov AV

    (2006) Alternans and spiral breakup in a human ventricular tissue model

    American Journal of Physiology. Heart and Circulatory Physiology 291:H1088–H1100.

    https://doi.org/10.1152/ajpheart.00109.2006

    • PubMed
    • Google Scholar
31. 1. Vigmond EJ
    2. Weber dos Santos R
    3. Prassl AJ
    4. Deo M
    5. Plank G

    (2008) Solvers for the cardiac bidomain equations

    Progress in Biophysics and Molecular Biology 96:3–18.

    https://doi.org/10.1016/j.pbiomolbio.2007.07.012

    • PubMed
    • Google Scholar

Author details

1. Ryan P O'Hara

   Department of Biomedical Engineering, Johns Hopkins University, Baltimore, United States

   Contribution

   Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Software, Validation, Visualization, Writing - original draft, Writing - review and editing

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-7636-1799

2. Edem Binka

   Division of Pediatric Cardiology, Johns Hopkins School of Medicine, Baltimore, United States

   Contribution

   Conceptualization, Data curation, Investigation, Methodology, Project administration, Resources, Supervision

   Competing interests

   No competing interests declared

3. Adityo Prakosa

   Department of Biomedical Engineering, Johns Hopkins University, Baltimore, United States

   Contribution

   Conceptualization, Data curation, Investigation, Methodology, Software, Supervision

   Competing interests

   No competing interests declared

4. Stefan L Zimmerman

   Department of Radiology, Johns Hopkins School of Medicine, Baltimore, United States

   Contribution

   Data curation, Resources

   Competing interests

   No competing interests declared

5. Mark J Cartoski

   Division of Pediatric Cardiology, Nemours/Alfred I. duPont Hospital for Children, Wilmington, United States

   Contribution

   Conceptualization, Data curation, Investigation, Methodology, Resources

   Competing interests

   No competing interests declared

6. M Roselle Abraham

   Division of Cardiology, University of California, San Francisco, San Francisco, United States

   Contribution

   Data curation, Resources

   Competing interests

   No competing interests declared

7. Dai-Yin Lu

   Division of Cardiology, University of California, San Francisco, San Francisco, United States

   Contribution

   Data curation, Resources

   Competing interests

   No competing interests declared

8. Patrick M Boyle

   Department of Bioengineering, University of Washington, Seattle, United States

   Contribution

   Conceptualization, Investigation, Methodology, Project administration, Software

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0001-9048-1239

9. Natalia A Trayanova

   Department of Biomedical Engineering, Johns Hopkins University, Baltimore, United States

   Contribution

   Conceptualization, Formal analysis, Funding acquisition, Project administration, Resources, Supervision, Writing - original draft, Writing - review and editing

   For correspondence

   ntrayanova@jhu.edu

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-8661-063X

• 1,664

  views

• 225

  downloads

• 23

  citations

Views, downloads and citations are aggregated across all versions of this paper published by eLife.