BACKGROUND:Cardiac amyloidosis (CA) is an underdiagnosed yet treatable cause of heart failure in which timely diagnosis is essential to initiate life-prolonging therapies. While artificial intelligence (AI)-based tools using transthoracic echocardiography (TTE), electrocardiography, or electronic health records have demonstrated promise for CA detection, most rely on single data sources. We aimed to evaluate whether integrating clinical, laboratory, and TTE biomarkers improves the performance of an existing TTE-based AI model for CA detection. METHODS:We developed and tested a combined AI echo-clinical model (AI-ECM) incorporating demographics, laboratory biomarkers, and TTE parameters into a previously validated TTE-only AI model (Us2.Ca). Model training and internal validation were performed using the Amyloidosis Imaging International Consortium, a global multiethnic registry comprised of 727 patients with CA and 316 controls, including 202 with suspected transthyretin-CA with negative diagnostic evaluation and 114 patients with biopsy-proven extracardiac light chain amyloidosis without cardiac involvement. Ground truth CA diagnosis was adjudicated per consensus criteria. AI-ECM and Us2.Ca performance was assessed using area under the curve, accuracy, sensitivity, and specificity. RESULTS:In building the AI-ECM, feature importance analysis showed that having the Us2.Ca prediction scores, relative wall thickness, sex, and estimated glomerular filtration rate contributed most to performance. The AI-ECM demonstrated superior performance (area under the curve, 0.94; accuracy, 90%; sensitivity, 93%; specificity, 85%) compared with the Us2.Ca (area under the curve, 0.89; accuracy, 80%; sensitivity, 76%; specificity, 91%; P=0.006). While the Us2.Ca model classification was indeterminate in 9% of the cases, the AI-ECM allowed classification of all cases. The AI-ECM improved sensitivity for light chain-CA detection and maintained high accuracy across subtypes and control groups. CONCLUSIONS:A multiparametric AI model integrating basic clinical, laboratory, and TTE data with the deep learning Us2.Ca improved performance for CA detection over Us2.Ca alone. This approach represents a step toward scalable, AI-guided precision diagnostics for CA in diverse populations.

INTRODUCTION:To externally validate an FDA-approved artificial intelligence (AI) tool for detecting heart failure with preserved ejection fraction (HFpEF) using echocardiographic video clips, comparing its performance to invasive haemodynamic criteria in a real-world referral cohort with unexplained dyspnoea. METHODS:We retrospectively analysed 47 patients who underwent transthoracic echocardiography (TTE) and right heart catheterization (RHC), including 28 with both rest and exercise haemodynamics. The AI model evaluated apical 4-chamber video data and classified outputs as HFpEF, no HFpEF, or non-diagnostic, without any other clinical data. The primary outcome was invasively defined HFpEF: pulmonary capillary wedge pressure (PCWP) ≥15 mmHg at rest or ≥25 mmHg with exercise. Secondary analysis used a PCWP/cardiac output (CO) slope >2 mmHg/L/min. We assessed sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV). RESULTS:Among patients with exercise RHC (n = 28), 71% met haemodynamic HFpEF criteria. The AI tool demonstrated sensitivity of 30%, specificity of 88%, PPV of 86%, and NPV of 33%. Using the PCWP/CO slope (n = 25), specificity and PPV were 100%, sensitivity 27%, and NPV 16%. AI-positive patients had significantly higher resting PCWP (20 vs 15 mmHg, P = .029), mean PA pressure (29 vs 24 mmHg, P = .02), and PVR (2.1 vs 1.3 WU, P = .002). In patients with indeterminate H2FPEF scores (n = 25), the AI correctly identified 80% of those with invasively confirmed HFpEF. Model performance was consistent across TTE-RHC intervals <365 and <90 days. CONCLUSIONS:This AI model demonstrated high specificity and positive predictive value (PPV) for detecting HFpEF, reliably identifying patients with more advanced haemodynamic abnormalities. Its performance remained robust across variable intervals between transthoracic echocardiography (TTE) and right heart catheterization (RHC), and in patients with indeterminate clinical scores. Due to limited sensitivity, the tool is best utilized to enrich identification of patients with clearly abnormal haemodynamics rather than to exclude HFpEF, particularly in early or borderline cases. While broader use as a screening tool is promising, prospective validation studies are necessary to confirm its utility in general populations.

BACKGROUND:Cardiovascular magnetic resonance (CMR) accurately measures left ventricular (LV) ejection fraction (EF) and late gadolinium enhancement (LGE), but is not commonly used for diastolic function. This study combined unsupervised machine learning and diastolic function with cine imaging to identify prognostic groups. METHODS:CMR studies from 391 patients with LVEF ≥50%, and no LGE were included. Using LV cine diastolic time-volume curves, early peak filling rate (E-PFR) indexed to stroke volume index (SVI), early to late diastolic filling ratio (E/A) ratio, and deceleration time (Decelt) were measured, along with left atrial volume index (LAVI) and LV mass index (LVMI). K-means clustering grouped individuals using only these parameters. Primary outcome was a composite of heart failure or ventricular arrhythmia admission, heart transplant, mechanical support, and cardiac death. RESULTS:Median follow-up was 5 years with 30 events. Three clusters were identified with event rates of 2.6% (group 1), 8.5% (group 2), and 17.2% (group 3). Time-volume curves demonstrated a triphasic pattern with group 2 having lower E-PFR (6.5 vs 9.7 and 8.4 SVI/s for groups 1 and 3, respectively, p < 0.001), lower E/A (1.2 vs 2.4 and 1.9 for groups 1 and 3, respectively, p < 0.001), and longer Decelt (223 vs 79 and 105 ms for groups 1 and 3, respectively, p < 0.001). Group 3 had increased remodeling based on higher LAVI (51.4 vs 26.6 and 27.5 mL/m2 for groups 1 and 2, respectively, p < 0.001) and LVMI (78.7 vs 48.7 and 50.6 g/m2 for groups 1 and 2, respectively, p < 0.001). Compared to group 1, groups 2 and 3 had hazard ratios (HR) of 3.21 (95% confidence interval [CI] 1.06-9.68, p = 0.038) and 2.52 (95% CI 1.06-4.47, p = 0.002) for events, respectively. Adjusting for comorbidities, only group 3 had a significant HR of 2.55 (95% CI 1.42-4.57, p = 0.002) compared to group 1. CONCLUSION:CMR cine-derived diastolic parameters can add prognostic value despite normal EF and absence of LGE.

This review summarizes current applications of artificial intelligence (AI) in multimodality cardiac imaging for the evaluation of valvular heart disease (VHD). The prevalence of VHD continues to rise, placing increasing demands on cardiovascular imaging and longitudinal management. AI systems have been applied across echocardiography, cardiac computed tomography (CCT), and cardiac magnetic resonance (CMR) to automate image classification, segmentation, disease detection, and severity assessment. The most mature AI models have centered on transthoracic echocardiography (TTE), where deep learning (DL) frameworks enable whole-study interpretation and preliminary report generation. Applications in CCT and CMR remain in earlier stages but show promise for segmentation, tissue characterization, and pre-procedural planning. AI has the potential to enhance the accuracy, reproducibility, and efficiency of imaging-based VHD assessment. Key challenges remain around generalizability, transparency, and clinical integration. Multidisciplinary collaboration is essential to ensure that AI complements, rather than replaces, human expertise.

AIMS:While pre-defined reference shapes have been used to assess morphological changes in the left ventricle, standardized methods for evaluating right ventricular (RV) remodelling are lacking. This study aimed to develop and test a new 3D echocardiography (3DE)-based method for quantifying RV shape in a large cohort of healthy individuals and across various disease states. METHODS AND RESULTS:3DE-derived RV mesh models were reconstructed in 1043 healthy subjects from the World Alliance of Societies of Echocardiography (WASE) study and in 581 patients with severe aortic stenosis, heart failure with reduced ejection fraction (HFrEF), post-heart transplantation, severe primary mitral regurgitation (MR), atrial secondary tricuspid regurgitation (A-STR), tetralogy of Fallot (TOF), and pulmonary hypertension (PH). To assess global RV shape, hemi-sphericity volume ratio (HSVR) and hemi-conicity angle (HCA) were calculated, where a higher HSVR and a more acute HCA reflect more spherical and conical shapes, respectively. In the WASE population, females had more spherical RVs, whereas males had more conical RVs (P = 0.028). Considering age, younger females had more conical RVs, while older individuals in both sexes showed spherical remodelling (P < 0.05). Comparing disease groups with WASE controls, MR, HFrEF, and A-STR patients had more spherical RVs compared with controls (both P < 0.001), while PH and TOF patients showed conical remodelling (both P < 0.001). In A-STR, a more conical remodelling was associated with adverse clinical outcomes. CONCLUSION:The proposed 3DE-based method comprehensively characterizes RV geometry, demonstrating demographic variation in healthy individuals and disease-specific alterations in patients, with important prognostic implications.

The use of artificial intelligence (AI) in echocardiography has been growing continuously, particularly in chamber quantification and disease detection. Although studies have validated different AI approaches for automated quantification of cardiac size and function, it is unknown whether different AI software tools provide concordant measurements. Accordingly, we aimed to: (1) evaluate the inter-software variability of the automated, AI-based measurements, and (2) compare their variability to that of conventional measurements performed by expert readers.Echocardiographic images from 116 randomly selected subjects, including 66 patients with cardiac pathology and 50 healthy volunteers were analyzed to obtain the following: (1) linear measurements: end-diastolic left ventricular internal diameter (LVIDd), basal right ventricular diameter (RVDd), and LV posterior wall thickness (LVPWd), (2) Doppler measurements (e’ lateral, e’ septal, mitral inflow E/A ratio), and (3) end-diastolic and end-systolic left ventricular volumes (LVEDV, LVESV), ejection fraction (LVEF), and end-systolic left atrial volume (LAESV). These measurements were performed by: (1) two fully automated AI-based software tools (Us2.AI and Philips), and (2) two expert readers using guidelines recommended methodology. Inter-software and inter-observer variability were assessed by percent absolute difference (PAD) and intraclass correlation coefficients (ICC) between corresponding measurements.AI measurements differed from each other, with the differences reaching significance for 6/10 parameters, while conventional measurements by experts significantly differed from each other for 3/10 parameters. Overall, the variability between AI tools was similar to the conventional methodology, with smaller PAD and higher ICC values noted for 6/10 and 2/10 parameters, respectively. Although different commercial AI-based tools for automated analysis of echocardiographic images do not provide identical results, the variability in the majority of these measurements was similar to or better than that of experts using the conventional methodology. Further studies are needed to ascertain these findings for other AI algorithms and larger numbers of readers and parameters.

An ultrasound artifact is a feature in an ultrasound image that does not accurately represent the true anatomy or pathology. Cardiac ultrasound artifacts are common and inevitable findings as they originate from the physical properties of ultrasound. Additionally, artifacts may occur due to interference from external equipment and devices producing ultrasound waves. This document provides a uniform and structured approach to managing ultrasound artifacts, including the appearance of the artifact on the image, the mechanism behind the artifact generation, the clinical impact of the artifact on the diagnosis and management of the patient, examples of real cases, and how the artifact can be avoided or mitigated. In addition to true artifacts, we also discuss a series of artifact-like phenomena. Everyone involved in performing or interpreting cardiac ultrasound should be familiar with artifacts and their potential for misdiagnosis, which in some instances may lead to serious clinical consequences. Despite continued improvements in ultrasound imaging technologies, artifacts remain common in all echocardiographic modes, including two-dimensional, spectral, and color Doppler, as well as three-dimensional echocardiography.

Guidelines for transthoracic echocardiographic examination recommend the acquisition of multiple video clips from different views of the heart, resulting in a large number of clips. Typically, automated methods, for instance disease classifiers, either use one clip or average predictions from all clips. Relying on one clip ignores complementary information available from other clips, while using all clips is computationally expensive and may be prohibitive for clinical adoption. To select the optimal subset of clips that maximize performance for a specific task (image-based disease classification), we propose a method optimized through reinforcement learning. In our method, an agent learns to either keep processing view-specific clips to reduce the disease classification uncertainty, or stop processing if the achieved classification confidence is sufficient. Furthermore, we propose a learnable attention-based aggregation method as a flexible way of fusing information from multiple clips. The proposed method obtains an AUC of 0.91 on the task of detecting cardiac amyloidosis using only 30

Cardiac amyloidosis (CA) is a rare cardiomyopathy, with typical abnormalities in clinical measurements from echocardiograms such as reduced global longitudinal strain of the myocardium. An alternative approach for detecting CA is via neural networks, using video classification models such as convolutional neural networks. These models process entire video clips, but provide no assurance that classification is based on clinically relevant features known to be associated with CA. An alternative paradigm for disease classification is to apply models to quantitative features such as strain, ensuring that the classification relates to clinically relevant features. Drawing inspiration from this approach, we explicitly constrain a transformer model to the anatomical region where many known CA abnormalities occur- the myocardium, which we embed as a set of deforming points and corresponding sampled image patches into input tokens. We show that our anatomical constraint can also be applied to the popular self-supervised learning masked autoencoder pre-training, where we propose to mask and reconstruct only anatomical patches. We show that by constraining both the transformer and pre-training task to the myocardium where CA imaging features are localized, we achieve increased performance on a CA classification task compared to full video transformers. Our model provides an explicit guarantee that the classification is focused on only anatomical regions of the echo, and enables us to visualize transformer attention scores over the deforming myocardium.

La ecocardiografía es una herramienta clave en la evaluación cardíaca, pero enfrenta desafíos como la subjetividad en la interpretación y la dependencia de la calidad de las imágenes. La IA, a través de técnicas como el machine learning y el deep learning, permite la automatización de procesos como la adquisición de imágenes, la cuantificación de parámetros (fracción de eyección, volúmenes ventriculares) y la sospecha de patologías (miocardiopatía hipertrófica, amiloidosis cardíaca). Además, la IA facilita la educación médica mediante simuladores que ofrecen retroalimentación en tiempo real. A pesar de sus ventajas, persisten desafíos como la falta de transparencia en los algoritmos ("caja negra"), la necesidad de validación externa y la dependencia de la calidad de las imágenes. En conclusión, la IA no reemplazará al ecocardiografista, pero se convertirá en una herramienta indispensable para optimizar el diagnóstico y la toma de decisiones clínicas.

Artificial intelligence (AI) is transforming echocardiography, ushering in an era of improved diagnostic precision, efficiency and patient care. In this Review, we present an in-depth exploration of AI applications in echocardiography, highlighting the latest advances, practical implementations and future directions. We discuss the integration of AI throughout the echocardiographic workflow, from image acquisition and analysis to interpretation. We outline the potential of AI to automate routine measurements and calculations, enable task shifting, recognize disease-specific patterns and uncover new phenogroups that might surpass current diagnostic classifications. Moreover, we address the aspects needed to create trustworthy AI systems, through careful validation, navigating regulatory requirements and upholding ethical standards, thereby presenting a balanced perspective on the advantages and limitations of this rapidly evolving technology. Through an examination of current AI applications, clinical studies and technological breakthroughs, we offer a comprehensive understanding of the evolving role of AI in the future of echocardiography and its capacity to advance cardiovascular care, while also acknowledging the current limitations of the widespread clinical implementation of AI-supported echocardiography. In this Review, the authors explore the state of artificial intelligence (AI) applications in echocardiography, discussing the integration of AI throughout the echocardiographic workflow and highlighting the clinical implementation, challenges and future directions of AI-supported echocardiography.

AIMS:Left ventricular (LV) pressure-volume (PV) loops, LV strain tensors, and intraventricular pressure gradients (IVPG) provide physiological information on cardiovascular performance and the interaction between LV and arterial system. Given that acquisition of PV loops and IVPG require invasive measurements, there is interest in the development of new noninvasive tools. This work aims to (i) demonstrate the application of noninvasive methods based on three-dimensional echocardiography (3DE) for describing PV loops and haemodynamic performance in terms of IVPG, in conjunction to LV strain tensors and (ii) to determine sex-, age-, and race-related normative values. METHODS AND RESULTS:This work is based on 3DE data from the World Alliance of Societies of Echocardiography study (1403 normal subjects). It applies physics-based techniques to construct noninvasive PV loops, assess blood propulsion using IVPG and 3D deformation using LV strain tensors. Sex- and age-related differences in strain and PV loop are limited to some parameters only, while haemodynamic performance in terms of IVPG does not vary with sex and is reduced with aging. Asian populations are characterized by smaller hearts, higher EF, strain, and PV loop scores than Whites, which in turn have similar LV size to Blacks and higher EF, strain, and PV loop scores, while IVPG was more similar across the populations. CONCLUSION:This study analysed a large normal population with physics-based methods that allow a deep mechanical and haemodynamic analysis of 3DE with 70% feasibility and provided reference values for a series of advanced parameters and their dependency on sex, age, and race.

Background: Transthyretin amyloid cardiomyopathy (ATTR-CM) is an important cause of heart failure (HF) characterized by increased left ventricular (LV) wall thickness that can be challenging to distinguish from other LV hypertrophic processes. Myocardial radiomic texture analysis using echocardiographic images may identify subtle differences in tissue structure, invisible to the human eye, that result from amyloid deposition. Our aim was to train a machine learning model utilizing echo-derived radiomics features to identify amyloid cardiomyopathy. Methods: Retrospective multi-center study of echocardiographic images from 749 patients: 200 with ATTR-CM and 549 with non-amyloid HF. Conventional radiomic features such as histogram, two-dimensional, gray-level co-occurrence matrix, and gray-level run-length matrix, in addition to the novel application of chi-square, gray-level gradient matrix, and Laws’ texture features, were extracted from parasternal short-axis views at 3 cardiac levels (base, mid, apex). Data preprocessing via median imputation and z-score standardization was performed. A machine learning pipeline was developed and implemented using a Random Forest (RF) classifier with Least Absolute Shrinkage and Selection Operator (LASSO) in Python. A balanced training set was randomly subsampled (150 ATTR-CM and 150 non-amyloid HF) with hyperparameter tuning. Model performance was evaluated on an independent testing set of real-world prevalence comprising the remaining 50 ATTR-CM and 399 non-amyloid HF samples (conferring 11.1% ATTR-CM prevalence). Results: Using the realistic low-prevalence testing dataset, the machine learning model achieved a sensitivity of 86%, specificity of 92%, positive predictive value (PPV) of 57.3%, and negative predictive value (NPV) of 98.1%. The F1 score was 0.92, respectively. Overall accuracy was 0.91. The area under the receiver operating characteristic curve (ROC-AUC) was 0.938. Analysis of features incorporated into the model demonstrated that all 3 LV levels contributed and that novelly applied features sensitive to various local texture patterns (lines, edges,&spots) highly contributed to discrimination performance. Conclusion: We have developed a novel tool using echocardiographic radiomics that differentiates ATTR-CM from non-amyloid HF with high precision. The high NPV has clinical utility to exclude ATTR-CM, while further refinement using available demographics should increase performance.

En la última década, la ecocardiografía tridimensional ha experimentado una enorme evolución tecnológica. El primer hito importante fue el desarrollo de los transductores transtorácicos matriciales, que reemplazaron la reconstrucción 3D tediosa y lenta a partir de la adquisición consecutiva de imágenes en múltiples planos, y dieron como resultado conjuntos de datos de volumen casi en tiempo real. Otro hito importante más reciente fue el desarrollo de transductores de muestreo completo para ETE 3D en tiempo real, que en los últimos cinco años se han utilizado ampliamente en la clínica. Una ventaja importante de esta tecnología incluye una excelente calidad de imagen independientemente del tipo de vida del paciente, facilidad de uso e imágenes visualmente impactantes, fáciles de interpretar, que proporcionan información clínica novedosa. Además, la rápida aparición de procedimientos percutáneos para el tratamiento de enfermedades estructurales del corazón, como la reparación de la válvula mitral o el cierre de fugas perivalvulares, demuestra que el éxito de estos procedimientos depende mucho de la guía por ETE 3D. En la Tabla 1 se detallan las áreas de investigación activa y las áreas que aún no se han explorado. Anticipamos que, en el futuro, una mayor miniaturización de los transductores de ETE 3D podría permitir que esta tecnología se expanda a pequeños pacientes pediátricos, lo cual tendría un gran impacto sobre los resultados de reparaciones intracardíacas complejas. Es más, la optimización de imágenes 3D con contraste haría que esta tecnología fuera útil en pacientes “técnicamente” difíciles y también en las pruebas de eco estrés 3D. El desarrollo de transductores vasculares 3D con frecuencias de captura de imágenes más altas mejoraría las capacidades actuales de diagnosticar la enfermedad arteriosclerótica carotídea, ya que posibilitaría una evaluación más sencilla de la carga de la enfermedad. Todos estos desarrollos futuros fortalecerán las bases de la E3D incrementando y expandiendo más su utilidad clínica en nuevos territorios.