<p>This paper proposes a novel and integrated approach for the detection of Parkinson’s Disease (PD) using brain imaging, combining deep learning with metaheuristic optimization techniques to enhance diagnostic accuracy and efficiency. The process begins with preprocessing of T1W1 and DT1 brain scans, followed by feature extraction, where shape features are derived using Hu moments and texture features are captured using the Gray Level Co-occurrence Matrix (GLCM). To optimize the feature selection, the Adaptive Spotted Hyena Algorithm (ASHA) is applied, ensuring that the most relevant and informative features are utilized for classification. The selected features are then fed into an EfficientNet-based deep learning model for PD classification. The integration of ASHA for feature selection significantly reduces the dimensionality of the data, thereby minimizing computational complexity while preserving critical diagnostic information. This approach not only improves classification accuracy but also enhances the model’s robustness and scalability in clinical settings. The methodology offers a comprehensive framework for the early detection of PD, addressing the need for reliable, efficient, and scalable diagnostic tools that can assist healthcare professionals in making informed decisions. By providing an optimized, automated solution for PD diagnosis, this work contributes to the development of advanced AI-based systems that can support clinical decision-making, facilitate early intervention, and improve patient outcomes. Furthermore, the proposed approach opens up avenues for future research, such as the integration of multimodal data sources and further optimization of the feature selection and classification processes.</p>