B baseline performance is established using an R(2+1)D video backbone with LSTM aggregation trained from Kinetics-400 pretrained weights, demonstrating strong discriminative performance for cardiac functional assessment and LV dysfunction classification, while early cardiotoxicity prediction from a single pre-therapy video remains a significant open problem for the community.
Abstract
Therapy-induced cardiotoxicity is the leading non-oncological cause of treatment interruption in breast cancer patients, yet early, automated risk stratification from routine cardiac imaging remains an unsolved problem. We present EchoRisk, the first curated, multicentre, longitudinal echocardiography dataset with explicit cardiotoxicity labels, released as the primary technical reference for the EchoRisk-MICCAI 2026 challenge. The dataset comprises 422 patients enrolled in the EU-funded CARDIOCARE prospective study across five European sites, yielding 2,159 echocardiography videos across 1,123 clinical exams acquired at up to five longitudinal timepoints, alongside a dedicated cohort of 280 patients with baseline imaging for early cardiotoxicity prediction. Three clinically grounded tasks are defined: automated estimation of left ventricular ejection fraction from cine video (Task 1), classification of LV dysfunction from longitudinal imaging (Task 2), and early prediction of therapy-induced cardiotoxicity from pre-therapy baseline echocardiography alone (Task 3). For each task we specify the evaluation protocol, primary and secondary metrics, and ranking procedure. We establish baseline performance using an R(2+1)D video backbone with LSTM aggregation trained from Kinetics-400 pretrained weights, demonstrating strong discriminative performance for cardiac functional assessment and LV dysfunction classification, while early cardiotoxicity prediction from a single pre-therapy video remains a significant open problem for the community. The dataset, evaluation code, and baseline implementations are publicly available to serve as a benchmark for further collaboration, comparison, and the creation of task-specific architectures in cardio-oncology.
Echocardiography is the cornerstone for risk stratification, diagnosis, and monitoring of cancer therapy–related cardiac dysfunction (CTRCD)(1). Artificial intelligence (AI)–guided echocardiography has shown high accuracy and reliability in diverse cardiac populations and may reduce variability while improving workflow efficiency(2, 3). However, this technology has not yet been validated in a dedicated cohort of patients with cancer.
To evaluate the accuracy and reliability of AI-guided echocardiography in assessing left ventricular ejection fraction (LVEF) and additional parameters, compared with conventional echocardiography, in a cardio-oncology population.
This study included patients identified retrospectively from a cardio-oncology registry. Studies, that had already been analysed manually by expert sonographers and reported using AGFA PACS system, were uploaded to the US2.ai platform for automated analysis. The primary outcome was the level of agreement (LoA) between AI-guided and standard echocardiography for LVEF. Secondary outcomes included LoA for additional echocardiographic parameters and LoA between AI- LVEF and 3D LVEF. The performance of the deep learning (DL) algorithm in identifying LVEF <50% was evaluated using the area under the receiver operating characteristic curve (ROC-AUC). Subgroup analyses were performed in predefined populations clinically relevant in cardio-oncology.
A total of 282 patients were included. Mean age was 60 ± 16 years, and 61% were women. Breast cancer was the most frequent malignancy (30.5%), followed by haematological malignancies (16.7%) and gastrointestinal tumours (10.6%). Manual median 2D LVEF was 60% (IQR: 55-64) and AI-derived LVEF was 59.2% (IQR: 53-64) showing good agreement and correlation (bias: −0.138, SD: 5.38, 95% LoA: −10.7 to 10.4, ICC: 0.791, 95% CI: 0.742–0.831, Spearman ρ: 0.718,), Table 1. Comparison between 3D echocardiography LVEF and AI-derived 2D LVEF showed similar agreement with narrower limits (bias: −0.13, 95% LoA: −9.51 to 9.26). The DL algorithm accurately identified LVEF <50% (ROC-AUC: 0.918, 95% CI: 0.875–0.961), Figure 1. Subgroup analyses demonstrated consistent agreement in patients with breast cancer, body mass index >30, prior radiotherapy and pericardial effusion.
In a large real-world cardio-oncology cohort, AI-guided echocardiography demonstrated strong agreement with conventional echocardiography for LVEF assessment and high accuracy for detecting clinically relevant LV dysfunction. Performance was consistent across key subgroups, supporting the feasibility, reliability, and potential clinical value of integrating DL-based analysis into routine cardio-oncology echocardiographic workflows.Agreement between manual and AI-echo AUC-ROC curve for LVEF<50%
M. Andres, V. Maharajan, M. C. Llamedo et al.· European Heart Journal, Supp...· 0 citations
Anthracyclines remain cornerstone agents in oncology, yet their cardiotoxic potential poses a substantial clinical challenge. Up to 30% of treated patients develop some degree of cardiac dysfunction, with overt heart failure occurring in 2–5% of cases. Current surveillance relies on serial ejection fraction measurements, which detect damage only after significant myocardial injury has occurred. Histopathological evidence consistently shows that the subendocardial layer suffers earliest and most severely from anthracycline exposure—often weeks before any decline in global ventricular function becomes apparent. This temporal gap represents a missed opportunity for timely cardioprotective intervention.
We designed the CARDIAC-STRAIN study to determine whether layer-specific strain analysis by cardiac magnetic resonance can identify subclinical cardiotoxicity substantially earlier than conventional echocardiographic surveillance, potentially enabling earlier initiation of cardioprotective treatment.
This prospective single-centre diagnostic cohort study will recruit 120 consecutive patients scheduled for anthracycline-based chemotherapy (sample size calculated to detect 20% sensitivity difference, power 80%, α=0.05). Eligible participants are aged 18–75 years with preserved baseline ejection fraction (≥50%) and no prior anthracycline exposure or known cardiomyopathy. Each patient undergoes blinded comprehensive cardiac evaluation at four timepoints: baseline, after the fourth chemotherapy cycle, four weeks post-treatment, and at six months' follow-up. The protocol includes 1.5T cardiac MRI with cine sequences, native and post-contrast T1 mapping, T2 mapping, and late gadolinium enhancement. Layer-specific strain is quantified at subendocardial, midmyocardial, and subepicardial levels using dedicated feature-tracking software. Parallel assessments include three-dimensional echocardiography and cardiac biomarkers (troponin, NT-proBNP). Primary endpoints: ejection fraction decline >10% to below 50%, decline >15% with preserved function, or layer-specific strain deterioration >15% from baseline.
Study hypothesis: We hypothesize that layer-specific strain analysis will detect subclinical myocardial injury approximately 2–4 weeks earlier than conventional ejection fraction monitoring, with significantly improved diagnostic sensitivity compared to standard surveillance protocols.
Expected outcomes: If layer-specific strain analysis proves capable of reliably identifying subclinical cardiotoxicity before irreversible damage occurs, this approach may help shift clinical practice from heart failure treatment to early prevention in cardio-oncology. The study received ethics committee approval in November 2025, and patient recruitment is underway.
N. Kavelashvili, N. Sharashidze, F. Schiedat et al.· European Heart Journal, Supp...· 0 citations
Background Cardiac computed tomography angiography (CCTA) has evolved beyond anatomical stenosis assessment into a comprehensive platform for cardiovascular and cardiometabolic risk stratification. Advances in postprocessing and artificial intelligence now enable automated quantification of multiple imaging biomarkers from a single acquisition, including coronary plaque characteristics, CT-derived fractional flow reserve (FFR-CT), epicardial adipose tissue (EAT), pericoronary adipose tissue (PCAT), and hepatic steatosis. Purpose In this narrative review, we synthesize current imaging biomarkers, evaluate their individual and combined prognostic value, and propose a conceptual multimarker framework for cardiovascular risk stratification—recognizing that several domains remain investigational and are not yet ready for routine, biomarker-guided management. Key findings Quantitative plaque analysis identifies high-risk features — including low-attenuation plaque, positive remodeling, and napkin-ring sign — that independently predict major adverse cardiovascular events (MACE) beyond stenosis severity. FFR-CT carries a Class 2a guideline recommendation for intermediate lesions and demonstrates superior vessel-level diagnostic accuracy compared with SPECT and comparable performance to PET in head-to-head trials. EAT volume and density independently predict incident coronary heart disease, atrial fibrillation, and all-cause mortality across large prospective cohorts. Pericoronary fat attenuation index (FAI) reflects local coronary inflammation, independently predicts MACE after adjustment for conventional risk factors and coronary calcium, and decreases in response to high-dose statin therapy. Hepatic steatosis, identifiable from the same noncontrast acquisition used for calcium scoring, is associated with a 64% increased odds of cardiovascular events in a meta-analysis exceeding 34,000 adults and predicts both plaque progression and high-risk plaque features in longitudinal registries. Emerging multimarker models combining these domains demonstrate incremental discriminatory value beyond individual imaging biomarkers or traditional clinical risk scores. Conclusion CCTA provides a pragmatic, multidimensional framework that integrates anatomic, functional, inflammatory, and metabolic information from a single noninvasive examination. While standardization, longitudinal validation, and equitable representation in research cohorts remain unresolved challenges, ongoing advances in AI-driven image analysis and multiomics integration may, if validated in prospective outcome studies, support the future translation of quantitative CCTA imaging biomarkers into more personalized cardiovascular care.
R. Mora, K. Irannejad, N. Abbas et al.· Frontiers in Cardiovascular...· 0 citations
PURPOSE
Durable LVAD therapy improves survival for advanced heart failure, yet adverse outcomes remain common. We evaluated whether combining pre-implant echocardiography with routinely available Electronic Health Record (EHR) data yields clinically useful post-LVAD risk predictions to improve patient selection and perioperative management.
METHODS
In this retrospective study (2015-2022), pre-implant apical four-chamber echocardiograms were processed via raw loops and U-Net segmentation. CNN embeddings were integrated with PCA-reduced EHR variables including demographics, laboratories, and hemodynamics. Survival models, including Cox proportional hazards and random survival forests, were trained on multimodal inputs. Performance was validated using stratified 5-fold cross-validation, targeting a primary endpoint of time to death or missed follow-up. Saliency mapping was utilized to ensure clinical interpretability of the model's features.
RESULTS
Multimodal models achieved higher discrimination than single-modality models, with segmented-echo inputs outperforming raw videos (mean C-index of 0.711). Saliency mapping identified clinically coherent predictors: right ventricular and septal geometry on imaging, alongside renal, hepatic, and nutritional status from the EHR.
CONCLUSIONS
Integrating pre-implant echocardiography with EHR data enhances risk stratification survival for LVAD candidates. This multimodal approach identifies high-risk phenotypes, specifically right-heart and systemic frailty, providing a framework for personalized clinical decision support and future multicenter validation.
Gabriel Farias Cacao, Dongping Du, Nandini Nair· International Journal of Art...· 0 citations
Background: Excess epicardial adipose tissue (EAT) is associated with cardiovascular-kidney-metabolic (CKM) dysfunction, but its assessment has traditionally required advanced imaging. We tested whether AI-enhanced echocardiography could enable scalable phenotyping of adverse epicardial adiposity and identify individuals at increased cardiometabolic risk. Methods: We developed PanAdipo, a video-based deep learning model in 1,114,441 videos from 28,797 studies across the Yale-New Haven Health System (YNHHS; 2016-2022), using expert reader annotations of prominent EAT (2.5% of studies). PanAdipo was evaluated in four cohorts: a temporally distinct YNHHS TTE set (n=4,588), an emergency department point-of-care ultrasound cohort across YNHHS (n=10,957), the geographically distinct MIMIC-IV cohort (n=4,549), and the community-based Multi-Ethnic Study of Atherosclerosis (MESA Exam 6, n=2,740). Analyses examined (i) discrimination of prominent EAT; (ii) independence from conventional echocardiographic outputs; (iii) spatial explainability using gradient-weighted class activation mapping; (iv) correspondence with paired cardiac CT-derived body composition phenotypes (n=5,594); and (v) age-, sex-, and BMI-independent associations with cardiometabolic biomarkers and incident metabolic disease. Results: In the held-out health system test set, PanAdipo discriminated prominent EAT with an AUROC of 0.91 (95% CI, 0.88-0.94), exceeding conventional measures of cardiac function and structure. In explainability analyses, the model's attention localized to the epicardial area across views and throughout the cardiac cycle. On paired cardiac CT imaging, the PanAdipo score correlated most strongly with epicardial adiposity (Spearman's rho=0.75; P<0.001) with weaker correlations with other adipose and non-adipose compartments and only modest correlations with BMI across cohorts (rho=0.19-0.40). In MESA, greater PanAdipo scores were independently associated with higher HOMA-IR and triglycerides, associations that persisted among normoglycemic participants. Higher PanAdipo scores were also associated with newly documented metabolic disease, including MASLD/MASH, after adjustment for BMI (HRpooled 1.25 [1.08-1.44] per 1-SD increment in log[PanAdipo]). Conclusions: AI-enabled echocardiography provides a scalable, view-agnostic biomarker that characterizes adverse epicardial adiposity and is associated with cardiometabolic dysfunction, highlighting a novel role for echocardiography in CKM risk stratification.
Arya Aminorroaya, A. Coppi, R. Mcnamara et al.· medRxiv· 0 citations
GLS and LVEF were the most sensitive echocardiographic parameters for differentiating between groups with and without cardiotoxicity, showing consistent changes throughout the study.
R. D. De Sousa, V. Fonseca, M. Henriques et al.· European Heart Journal, Supp...· 0 citations