Cardiovascular diseases (CVDs) persist a prominentreason of death worldwide, necessitating precise and timely identification for effective treatment. Deep Learning (DL) methods have presentedencouragingoutcomes in numerous medical applications, including cardiac disease diagnosis. However, achieving precise diagnosis often requires the extraction and integration of relevant features from medical imaging data. In this work, we propose a novel methodology for cardiac disease diagnosis using a carefully curated feature set and deep learning models.Our methodology involves the development of a feature extraction pipeline tailored for cardiac imaging data, encompassing both anatomical and functional aspects. This pipeline integrates advanced image processing techniques to extract salient features such as ventricular volume, ejection fraction, myocardial strain, and texture features from cardiac images. These features are then utilized to construct a comprehensive feature set capturing diverse aspects of cardiac morphology and function.We employ various deep learning architectures, including recurrent neural networks (RNNs) and convolutional neural networks (CNNs), to study the complex associationsinside the feature set and accurately classify different cardiac diseases. The prototypes are trained on a huge dataset comprising diverse cardiac imaging modalities and pathological conditions to ensure robustness and generalization.Evaluation of our proposed approach on independent test datasets demonstrates superior performance compared to conventional methods and existing deep learning models. The proposed feature set modelling framework not only achieves high diagnostic accuracy but also provides insights into the underlying physiological processes contributing to cardiac pathology.