Heart failure (HF) remains a major global cause of morbidity and mortality, where early diagnosis is critical for improving patient outcomes. Conventional single-modality approaches often fail to capture the complex and multifactorial nature of HF. This study investigates the feasibility of a late-fusion framework that integrates modality-specific predictions derived independently from cine-MRI, electrocardiographic signals, biomarkers and demographic data for HF prediction. Independent cine-MRI data from 281 patients, ECG recordings from the PTB-XL PhysioNet database and biomarker profiles from 157 patients were retrospectively analyzed as separate modality-specific cohorts. Twenty-five features were extracted and processed. Modality-specific models (Attention U-Net, MLP, XGBoost) were trained separately on pre-extracted features to preserve predictive accuracy while minimizing computational cost. Their outputs were combined through ensemble meta-learning (XGBoost, LightGBM, Random Forest) with sample weighting to handle missing data. The final HF prediction probability was obtained by averaging the outputs across the three meta-learners. The proposed framework achieved competitive diagnostic performance, with 98.00% (95% CI: 94.96–99.45%) accuracy, 97.80% (95% CI: 92.28–99.73%) sensitivity, 98.17% (95% CI: 93.53–99.78%) specificity, an F1-score of 97.80% (95% CI: 93.6–99.8%) and an AUC of 0.978 (95% CI: 0.945–0.996) when evaluated against state-of-the-art methods. The results highlight the potential of late-fusion strategies for integrating independently trained modality-specific predictions, offering a feasible approach for HF risk assessment under heterogeneous data availability.
Wafa Baccouch, Narjes Benameur, Abdulrahman A. Alsayyari et al.· Technologies· 0 citations
Native T1 values in cardiac magnetic resonance (CMR) are influenced by acquisition sequence, physiological factors, and site-specific variability, complicating clinical interpretation. The absence of population-specific reference values further limits diagnostic accuracy. This study aimed to establish native T1 reference ranges in a Tunisian population, evaluate sex- and age-related differences using MOLLI 5(3)3 and SMART1Map sequences, and assess inter-site variability of MOLLI measurements across two centers using identical scanners and protocols. Ninety-four healthy volunteers were enrolled: 64 subjects (24 females, 40 males; 5–77 years) were scanned at Site 1 to establish reference values, and 30 subjects (20 females, 10 males; 19–62 years), frequency-matched by age to a subset of Site 1, were scanned at Site 2 to evaluate site-related variability. MOLLI-derived T1 values showed a significant negative association with age (β = –0.84 ms/year, p = 0.012), whereas SMART1Map-derived values were not associated with age (β = 0.30 ms/year, p = 0.556). SMART1Map consistently yielded higher T1 values than MOLLI (global mean ± SD: 1184 ± 70 ms vs. 1049 ± 55 ms; p < 0.001). Sex-related differences were minimal, with only a small apical difference observed for MOLLI (p = 0.023). Multi-center MOLLI measurements showed similar T1 distributions across sites, with minor differences observed at the mid-ventricular level. These findings provide the first Tunisian-specific native myocardial T1 reference values at 1.5 T, highlighting sequence-dependent differences, age-related effects for MOLLI, minimal sex-related influence, and overall inter-site consistency. These findings support the use of locally derived, sequence-specific reference ranges to improve clinical interpretation.
