Identifying functional outcomes after ischemic stroke (IS) early and accurately is essential for improving patient care and customizing rehabilitation programs. In this article, we compare several deep learning approaches, with a particular focus on convolutional neural networks (CNNs), recurrent neural networks (RNNs), and multimodal fusion models. The dataset enables a comprehensive approach to predictive modeling by integrating structured clinical data (such as vital signs and patient histories) with medical imaging (MRI and CT scans) obtained from clinical trials and publicly accessible sources. Key performance indicators, including accuracy, sensitivity, specificity, and the Area Under the Receiver Operating Characteristic Curve (AUC-ROC), are used to evaluate the models. The findings demonstrate that multimodal fusion approaches significantly outperform standalone CNN and RNN models, achieving accuracy rates above 90% and AUC-ROC values as high as 0.9. These results highlight the potential of integrating multiple data modalities to enhance prediction performance. Interpretability techniques, such as SHapley Additive exPlanations (SHAP) and Gradient-weighted Class Activation Mapping (Grad-CAM), are employed to improve the clinical relevance of the models. These methods provide insights into the models’ decision-making processes, fostering trust among medical professionals and facilitating their adoption in clinical settings. This work paves the way for more individualized and effective treatment strategies. By developing AI-driven tools for stroke outcome prediction, it ultimately aims to improve patient outcomes and optimize rehabilitation programs.