Ischemic stroke is a major cause of disability and mortality, making early detection vital for improving patient outcomes and management strategies. Advancements in machine learning (ML) and deep learning (DL) offer powerful tools to enhance diagnostic accuracy. Magnetic Resonance Imaging (MRI) plays a critical role in identifying ischemic strokes. This study focuses on comparing the performance of various ML models—K-Nearest Neighbors (KNN), Decision Tree (DT), and an ensemble KNN-DT model—combined with three DL models (VGG16, VGG19, and ResNet50) for detecting ischemic stroke from MRI images. Utilizing the Kaggle Acute Ischemic Stroke MRI dataset, which includes scans of both stroke-affected and unaffected individuals, we applied different image preprocessing techniques and extracted features using VGG16, VGG19, and ResNet50. Principal Component Analysis (PCA) was employed for feature selection. These selected features were then used to train and evaluate the ML models. Our findings reveal that the ensemble KNN-DT model, especially when paired with ResNet50, consistently outperformed the individual models and other combinations across all performance metrics. These results indicate that integrating DL and ML methods, particularly using ResNet50 with ensemble techniques, can significantly enhance early stroke detection, facilitating quicker and more effective medical interventions.