A Hybrid ML–CNN Heart Disease Prediction System using Clinical Features and ECG Images with user- Friendly Risk Assessment

One of the main causes of death worldwide is still heart disease, and serious complications can be avoided with early detection. The literature demonstrates that convolutional neural networks (CNNs) exhibit remarkable performance on ECG images, whereas machine learning (ML) models are effective for structured clinical data. Nevertheless, the majority of earlier research treats these modalities separately and hardly ever concentrates on deployable, user-friendly hybrid systems. This study suggests a novel hybrid heart-disease risk prediction framework, which highlights the predictive value of behavioural and clinical parameters, and which shows high- accuracy ECG classification using CNNs. Using simplified user- provided inputs (age, blood pressure tendency, cholesterol level, chest discomfort, and lifestyle factors), the system incorporates (i) an ML-based clinical risk estimator and (ii) an optional tendency, cholesterol, weariness, chest pain, and lifestyle factors), and (ii) an optional CNN model-based ECG-image classifier that is trained on a balanced ECG dataset. Even in the absence of an ECG, an accurate assessment is made possible to produce a final risk score. An interactive Streamlit application is used to deploy the entire pipeline for real-time use. According to experimental results, the CNN model achieves 95.5% validation accuracy on ECG classification, while ML models reached up to 91.6% accuracy on clinical data. The hybrid design offers a useful decision-support tool for early heart disease assessment, enhances interpretability, and is in line with current research trends. This study shows how well behavioural, clinical, and ECG-based deep learning features can be combined to produce reliable cardiac risk predictions.