Alzheimer’s disease (AD) is mainly a neurodegenerative sickness. The primary characteristics are neuronal atrophy, amyloid deposition, and cognitive, behavioral, and psychiatric disorders. Numerous machine learning (ML) algorithms have been investigated and applied to AD identification over the past decades, emphasizing the subtle prodromal stage of mild cognitive impairment (MCI) to assess critical features that distinguish the disease’s early manifestation and instruction for early detection and treatment. Identifying early MCI (EMCI) remains challenging due to the difficulty in distinguishing patients with cognitive normality from those with MCI. As a result, most classification algorithms for these two groups perform poorly. This paper proposes a hybrid Deep Learning Approach for the early detection of Alzheimer’s disease. A method for early AD detection using multimodal imaging and Convolutional Neural Network with the Long Short-term memory algorithm combines magnetic resonance imaging (MRI), positron emission tomography (PET), and standard neuropsychological test scores. The proposed methodology updates the learning weights, and Adam’s optimization is used to increase accuracy. The system has an unparalleled accuracy of 98.5% in classifying cognitively normal controls from EMCI. These results imply that deep neural networks may be trained to automatically discover imaging biomarkers indicative of AD and use them to identify the illness accurately.

A novel human-inspired metaheuristic framework, named Sastha Pilgrimage Optimization (SPO), mimics the optimization strategies inherent in the Sabarimala pilgrimage structure and the dynamics of pilgrims’ group behaviors. This methodology utilizes the significant characteristics of the practice, including Guruswamy’s selection (Leader), adaptive group interaction, and trade-offs between exploration and exploitation. It features a leader-based decision mechanism that balances individual performance with group harmony, guided by the most experienced participant. The exploration phase aims to cover a vast search space with dynamic updates in position. In contrast, in the exploitation phase, fine-tuning is achieved with adaptive coefficients and Lévy flights that guarantee convergence to global optima. The mathematical model of the algorithm incorporates new performance and distance threshold values, ensuring that route choices align with both individual and group objectives. A new human-inspired design by SPO can overcome the limitations of nonlinear, high-dimensional optimization challenges when compared to standard methods. The algorithm was benchmarked with over ten complex functions and validated on two datasets: the Cardiovascular dataset for feature selection and classification, and the Brain Tumor MRI dataset for image segmentation. These contributions make SPO a scalable, efficient, versatile optimization tool adaptable to domains requiring precise, high-dimensional decision-making.