2026/05/01 by Benoit Caullery, Céline Fouard, Antoine Simon +2 · 1 voice
Medicine · #Inflammatory Myopathies and Dermatomyositis #IgG4-Related and Inflammatory Diseases #Pericarditis and Cardiac Tamponade
paper · doi:10.1093/ehjci/jeag111
openalex publication_date 2026/05/01 · openalex created_date 2026/05/09 · openalex updated_date 2026/07/22
Endomyocardial biopsy (EMB) remains a key diagnostic tool when myocardial tissue characterization is expected to alter management, particularly in suspected inflammatory cardiomyopathy or isolated cardiac sarcoidosis.1 Its value is especially high when histological confirmation may support immunosuppressive treatment or clarify the diagnosis in the setting of persistent troponin elevation, ventricular dysfunction, or unexplained arrhythmias. However, conventional EMB has an important limitation: diagnostic yield may be low when myocardial abnormalities are focal, patchy, or remote from standard biopsy sites.2 In such situations, the challenge is not only to identify the diseased substrate before the procedure, but also to translate this information into an actionable roadmap in the catheterization laboratory. We describe a workflow based on routine late gadolinium enhancement cardiovascular magnetic resonance (LGE-CMR), patient-specific three-dimensional modelling, and fluoroscopic roadmapping. The principle is to transform routine clinical LGE images into a patient-specific model that contains the predefined target and can be displayed in fluoroscopic working views during biopsy. This strategy is intended to improve spatial targeting rather than replace procedural expertise. It is especially useful when multimodality imaging reveals a focal myocardial abnormality and when tissue confirmation is likely to influence treatment. In our experience, the method is applicable not only to conspicuous septal lesions but also to smaller substrates that may be difficult to reach using fluoroscopy alone.3 An important practical point is that the workflow does not require a dedicated high-resolution 3D LGE acquisition. The planning model is generated from routine clinical LGE-CMR data. This matters because routine LGE datasets are frequently anisotropic and have limited through-plane resolution, which restricts their direct volumetric use. The segmentation strategy was therefore developed specifically to recover meaningful three-dimensional information from standard clinical acquisitions.4 In the present implementation, the LGE-positive target remains manually delineated, since the workflow is not yet a fully automated bedside tool. At this stage, manual delineation remains the most robust way to ensure that the final target truly corresponds to the abnormal substrate identified on LGE-CMR.3,4 Its innovative aspect lies in the segmentation strategy used to extract clinically meaningful three-dimensional information from routine LGE datasets. Rather than relying on a simple binary myocardial mask, the method first generates a probability map of healthy myocardium using a dedicated 2D MultiResUnet. This probability map is then concatenated with the original LGE dataset and processed by a transformer-based 3D U-Net architecture.4 An adapted contour-aware loss function is used to improve myocardial boundary delineation. This design addresses a major limitation of routine LGE-CMR, namely, substantial slice thickness and anisotropy. The probability map enriches the input of the second network, while the transformer-based 3D stage helps preserve spatial relationships despite limited z-axis resolution. The result is a patient-specific three-dimensional myocardial model that integrates ventricular geometry and the predefined LGE target.3,4 Once generated, the patient-specific 3D model is displayed using the same or closely matched oblique projections as those used during the biopsy procedure. The goal is not strict real-time fusion but a practical visual correspondence between fluoroscopic anatomy and the pre-identified CMR lesion. This step helps determine whether a right- or left-sided biopsy approach is more appropriate and increases operator confidence in catheter and bioptome positioning.1 A key procedural point is that targeted EMB should not rely on a single standard setup. Instead, the access route and support system should be adapted to the location of the imaging-defined target. For right ventricular targets, especially septal lesions, we favour a transfemoral venous approach using an Agilis™ NxT Steerable Introducer (Abbott) to improve directional control and contact stability. In this setting, the steerable sheath provides a stable and highly manoeuvrable platform for precise sampling. Biopsy is then performed using a 110-cm endomyocardial bioptome (BD2C, ab medica SAS, France). This configuration is particularly useful for septal targets, where subtle angulation differences may determine whether the bioptome reaches the intended lesion or only a neighbouring normal region. For left ventricular biopsy, we use a 7-F guiding catheter through either femoral or radial arterial access, depending on target location, vascular anatomy, and patient size. A JR4 90-cm guiding catheter is preferred for septal, inferior, posterior, lateral, and apical targets, whereas an AL1 90-cm guiding catheter is preferred for anterior targets. A rotating haemostatic Y-adapter is systematically used to allow flushing, pressure monitoring, and careful air management throughout the procedure. For left and right ventricular access, the ventricle is first entered with a pigtail catheter. A guidewire exchange is then performed, and the pigtail is replaced by the guiding catheter selected according to the target location. Before tissue acquisition, ventriculography is repeated in RAO and LAO views to recapitulate the fluoroscopic roadmapping projections and confirm target orientation relative to the catheter trajectory. In practice, both right- and left-sided procedures include ventriculography in matched projections to reproduce the same chamber shape and working incidences as the 3D model, thereby facilitating concordance between the roadmap and live fluoroscopy.1 During biopsy, the pre-procedural 3D model is displayed in matching views while EMB is performed under standard fluoroscopy. The objective is to reduce the mismatch between the biopsy site and the imaging-defined lesion. This is most valuable when the substrate is focal, patchy, or poorly represented by fluoroscopic landmarks alone. The roadmap may also support multimodality integration: when another imaging modality, such as FDG-PET, shows concordant abnormality in the same region, this overlap may further strengthen the targeting strategy in suspected inflammatory cardiomyopathy or cardiac sarcoidosis.3 Systemic anticoagulation is achieved with 5000 IU of unfractionated heparin during left ventricular EMB. Biopsy sampling is performed only after confirmation of an activated clotting time of approximately 250 s. This is particularly important because thromboembolic risk must be minimized, although the same careful anticoagulation monitoring is applied to targeted biopsy procedures in general. At the end of the procedure, transthoracic echocardiography is performed systematically as an immediate safety check. After right ventricular biopsy, echocardiography is used primarily to exclude pericardial effusion. After left ventricular biopsy, it is used to exclude both pericardial effusion and new or worsening mitral regurgitation. When transfemoral access is used, vascular closure is performed with a FemoSeal™ vascular closure device (Terumo) according to local practice. A representative case is that of a 74-year-old woman presenting with heart failure, severely reduced left ventricular ejection fraction, and persistent troponin elevation in a clinical context suggestive of overlap between systemic sclerosis and inflammatory cardiomyopathy. Cardiac magnetic resonance showed focal septal late gadolinium enhancement, increased T2 signal, and elevated extracellular volume, consistent with active inflammatory myocardial involvement. To optimize diagnostic yield, a patient-specific three-dimensional ventricular model was generated from the routine CMR dataset and displayed in fluoroscopic working views to define the most appropriate projections and biopsy approach. A targeted EMB was then performed under fluoroscopic guidance with simultaneous display of the CMR-derived roadmap. Histopathological analysis showed a moderate interstitial inflammatory infiltrate predominantly composed of lymphocytes and histiocytes, without myocyte necrosis, thereby supporting the diagnosis of inflammatory cardiomyopathy. Histological confirmation had a direct therapeutic consequence, leading to immunosuppressive treatment and subsequent clinical improvement (Figure 1). Targeted myocardial biopsy guided by CMR-based 3D roadmapping. (A) Cardiac magnetic resonance late gadolinium enhancement (LGE) short-axis view showing focal septal myocardial hyperenhancement, consistent with inflammatory myocardial involvement. (B) Right ventriculography in right anterior oblique (RAO) 60° projection. (C) Patient-specific three-dimensional cardiac model reconstructed from CMR data, with the targeted myocardial region highlighted in red. (D) Fusion of live fluoroscopic imaging during endomyocardial biopsy with the CMR-derived 3D model, demonstrating accurate positioning of the biopsy forceps within the targeted myocardial region. (E) Histopathological examination (haematoxylin–eosin–saffron staining) showing preserved myocardial architecture with a moderate interstitial inflammatory infiltrate predominantly composed of lymphocytes and histiocytes. This workflow has several practical strengths. First, it is based on routine LGE-CMR rather than a dedicated 3D high-resolution sequence, which supports broader clinical applicability. Second, it provides a tangible way to translate tissue characterization into procedural guidance. Third, it can help the operator choose not only where to biopsy, but also how to biopsy, by adapting the access route and support material to the target location. Finally, it may renew interest in advanced LGE-based targeting of EMB in focal inflammatory cardiomyopathies and related entities, where non-targeted biopsy is expected to have limited sensitivity.2 Several limitations should nevertheless be acknowledged. This workflow is currently best suited to expert centres familiar with advanced CMR, post-processing, and EMB. The present implementation still includes manual delineation of the LGE-positive target, which means that the workflow is not yet fully automated. Lesion accessibility also varies according to substrate distribution: some septal lesions are relatively straightforward to target, whereas smaller, more lateral, or less extensive lesions may remain technically challenging despite roadmapping. In addition, the current report illustrates a practical workflow rather than a standardized, prospectively validated targeting platform. It should therefore be regarded as a clinically useful proof of concept rather than a universally established method. Routine LGE-CMR can be transformed into a patient-specific three-dimensional roadmap for targeted EMB without requiring a dedicated 3D high-resolution acquisition. By combining routine clinical imaging, dedicated segmentation, fluoroscopic recapitulation, and target-adapted procedural material, this workflow may improve targeting of focal myocardial abnormalities and support histological diagnosis in selected patients. Beyond lesion visualization alone, its practical contribution is to bridge the gap between pre-procedural tissue characterization and the technical reality of biopsy in the catheterization laboratory. Benoit Caullery [Conceptualization, Writing—original draft (lead)], Celine Fouard [Resources (equal)], Mireille Garreau [Resources (equal)], Antoine Simon [Resources (equal)], and Gilles Barone-Rochette (Conceptualization [equal], Writing—review & editing [lead]) None. The data underlying this study are not publicly available due to intellectual property restrictions and software-related confidentiality constraints. Access to the data is therefore limited, as the proprietary software used for data processing and analysis is not publicly accessible for patent and confidentiality reasons.