| Echocardiography is a critical imaging modality for assessing cardiac structure and function, particularly in evaluating the left ventricle. The left ventricular ejection fraction (LVEF) is a key metric for diagnosing heart health, yet challenges persist in accurately predicting extreme values. This study harnesses a diverse range of deep and machine learning techniques to cardiac function classification and enhance LVEF prediction from echocardiographic images, specifically focusing on apical four-chamber views. Enhancement techniques applied on specific frames extracted from the EchoNet-Dynamic dataset to optimize feature representation. By employing advanced convolutional neural networks (CNNs), including EfficientNetB0, VGG19, InceptionV3, NASNetLarge, and ResNet-50, the study seeks to maximize the efficacy of feature extraction. Among these models, ResNet-50 demonstrated the highest classification accuracy, forming the foundation for subsequent analysis. The comprehensive feature representations derived from ResNet-50 were then utilized in a classification learner, integrating methods such as Support Vector Machines (SVM), optimized SVM, Cosine K-Nearest Neighbors (KNN), and Gaussian Naive Bayes to further enhance classification performance achieving 89%. Additionally, these features were implemented within a Gaussian Process Regression (GPR) framework, yielding impressive predictive accuracy for LVEF, characterized by a Root Mean Square Error (RMSE) of 2, Mean Absolute Error (MAE) of 1.32 and an R² value of 0.92. Validation of the model's performance using a newly collected dataset reinforces its applicability in clinical settings. After achieving 87.88% classification accuracy, a regression model was constructed to predict ejection fraction values, yielding a high R-squared value of 0.88, and MAE was found to be 1.55, and the RMSE was 3.563, highlighting the model's reliability in producing accurate predictions. These results provide important insights into the assessment of cardiac health and address notable gaps in the literature regarding LVEF extremes. The results highlight the effectiveness of enhanced frame extraction in advancing cardiac function assessment through sophisticated echocardiographic analysis. | Abstract |