Heart disease or Cardiovascular diseases (CVDs) are a leading cause of mortality worldwide, necessitating the development of robust predictive models for early diagnosis and intervention. This paper explores the application of advanced Convolutional Neural Network (CNN) architectures ResNet, EfficientNet, DenseNet, InceptionV4, and SqueezeNet in predicting heart disease. By leveraging the strengths of these state-of-the-art models, we aim to enhance the accuracy and reliability of heart disease prediction systems. The proposed methodology involves a detailed process of data preprocessing, model training, and performance evaluation. Key metrics such as accuracy, sensitivity, specificity, F1 score, and Receiver Operating Characteristic - Area Under the Curve (ROC-AUC) are used to assess the effectiveness of each CNN model. Our experimental results demonstrate that DenseNet outperforms the other models, achieving an accuracy of 93%, with balanced performance across all other evaluation metrics. ResNet and EfficientNet also exhibit strong performance, highlighting their potential in clinical applications. The findings underscore the importance of selecting appropriate CNN architectures to maximize diagnostic accuracy in heart disease prediction. This study not only advances the understanding of CNNs in healthcare applications but also provides a foundation for future research into more models and their real-world deployment. The paper concludes by discussing the inferences of the results and suggests directions for future work, including the incorporation of additional data sources, development of hybrid models, and the study of model explainability to enhance trust in clinical decision support systems.