Introduction: Magnetic resonance imaging (MRI) has become increasingly common for the diagnosis of cardiovascular diseases in recent years. However, this technique remains time‑consuming and prone to human error. These limitations highlight the need for artificial intelligence, particularly deep learning, which has demonstrated strong capabilities in processing medical images. The aim of this study is to present a convolutional neural network (CNN) model for accurate detection of heart disease while reducing reliance on human intervention.
Methods: In this research, a CNN‑based model was designed and implemented for cardiac disease diagnosis. Unlike conventional approaches that process images as single frames or in parallel, this study employed dual‑channel inputs consisting of end‑diastolic and end‑systolic images to better capture cardiac cycle dynamics. The proposed architecture included multiple convolutional and max‑pooling layers, with dropout layers applied to prevent overfitting. Furthermore, fine‑tuning of hyperparameters such as optimizer type and batch size, along with advanced data augmentation techniques, contributed to improved network performance.
Results: Experiments were conducted on the standard ACDC dataset, and model performance was evaluated using established metrics. Findings revealed that the dual‑channel approach achieved significant improvements compared to single‑frame and parallel models. Moreover, comparisons with baseline CNN architectures and traditional machine learning algorithms confirmed the superiority of the proposed method in terms of accuracy and stability.
Conclusion: The dual‑channel approach can serve as an effective and reliable tool for cardiac disease diagnosis. By offering higher accuracy and reducing dependence on human interpretation, this method has the potential to replace parts of conventional, costly diagnostic processes.