Automated assessment of cardiac health from magnetic resonance imaging (MRI) plays an important role in assisting the clinical diagnosis of cardiac problems with great accuracy and in a timely manner; however, it is still challenged by variations in image quality and complex structures of the heart. This paper presents a multi-faceted deep learning model of cardiac image segmentation and classification with a multifaceted preprocessing and modeling pipeline. First, a strong preprocessing method, such as resampling, intensity normalization, and histogram equalization, was used to increase image quality and consistency, and then the process of data augmentation was used to increase the diversity of the dataset and model robustness. For the segmentation phase, an Inverted U-Net-based architecture is created by incorporating Deep Neural Networks (DNN), Graph Neural Networks (GNN), and Long Short-Term Memory (LSTM) networks to properly capture spatial and contextual dependency in the cardiogram images. The segmentation performance was evaluated with accuracy, precision, and dice coefficient, as the Inverted U-Net++ with LSTM showed the best performance, with the maximum accuracy being 0.813 by successfully modeling the complicated structural patterns. In the classification phase, the segmented cardiac images were categorized into normal and abnormal segmentation using Support Vector Machine or SVM, Long Short Term Memory or LSTM, and Convolutional Neural Networks or CNN. Comprehensive evaluation with the help of accuracy, precision, recall, and F1-score proved that LSTM-based classifiers outperform the competing models with the highest classification accuracy of 0.813. The comparison between segmentation and classification stages reveals the effectiveness and strength of the offered framework. Overall, the results show the potential of the proposed approach to provide accurate and reliable automated cardiac health assessment, and is a promising decision support tool for clinical cardiac diagnosis.