Purpose <p>Heart disease remains one of the leading causes of mortality worldwide, and its early detection is critical for reducing fatal outcomes, especially in regions with limited medical expertise and diagnostic infrastructure. This study aims to develop an accurate and robust heart disease detection framework by leveraging multi-modal ECG data processing, optimized feature selection, and an ensemble learning strategy to enhance early diagnostic performance.</p> Methods <p>This paper proposes CardioOptiEnsembleNet, a novel heart disease detection model integrating advanced signal preprocessing, feature extraction, optimization, and ensemble classification. ECG signals are preprocessed using Independent Component Analysis (ICA) to remove artifacts, while Notch Filtering is applied to eliminate powerline interference. Feature extraction is performed using Continuous Wavelet Transform (CWT) to capture time–frequency characteristics and Long Short-Term Memory (LSTM) networks to model temporal dependencies in ECG time-series data. An attention-based feature fusion mechanism integrates heterogeneous features, followed by a ChimpGaz hybrid optimization algorithm, which combines Gazelle and Chimp optimization strategies to select the most relevant features. Finally, classification is carried out using an ensemble of Random Forest (RF), K-Nearest Neighbors (KNN), and Support Vector Machine (SVM) models, with the hybrid optimizer further tuning SVM parameters. The entire framework is implemented in MATLAB, and performance is evaluated using standard classification metrics.</p> Results <p>Experimental evaluation demonstrates that the proposed CardioOptiEnsembleNet significantly improves heart disease detection performance compared to individual classifiers and non-optimized models. The attention-driven feature fusion and ChimpGaz-based optimization effectively reduce redundancy and enhance discriminative power, leading to improved classification accuracy, sensitivity, specificity, and overall robustness. The optimized ensemble consistently outperforms baseline approaches across all evaluation metrics.&#xa0;</p> Conclusion <p>The proposed CardioOptiEnsembleNet provides an effective and reliable framework for early heart disease detection by integrating optimized feature selection, attention-based fusion, and ensemble learning. The hybrid ChimpGaz optimization algorithm plays a key role in enhancing diagnostic accuracy and model generalization. This approach has strong potential for deployment in clinical decision-support systems, particularly in resource-constrained healthcare.</p>