Alzheimer’s Prognosis from Neuroimaging Using Optimized Variational Mode Decomposition based on Brown Beer and Transfer Learning

Purpose: Alzheimer’s Disease (AD) is a progressive, chronic disorder of the brain for which there is now no operative treatment. However, there are medications that can slow its growth. In order to stop and control the development of AD, earlier diagnosis of the disease is prototypical.Method: We propose a novel method, which is based on Brown beer optimization, integrated to overcome the issue of the VMD related to the number of modes (K) and the penalty factor (α). This optimal method was helpful in decomposing MRI images into intrinsic modes and capturing complex signal patterns of the brain. Then, transfer learning-based models were used to extract deep features from sub-band images, and feature engineering, such as dimension reduction and feature ranking, was performed on these features before feeding them to a classifier to classify healthy and affected AD images.Result: The proposed method achieved diagnostic accuracy of 97.6%, specificity of 97.4%, and sensitivity of 98.8% in distinguishing AD from healthy controls, outperforming traditional VMD and standard deep learning approaches. Progression prediction models demonstrated a correlation coefficient of 0.89 with clinical outcomes, highlighting robust prognostic capabilities. To validate the model's performance, cross-validation and Grad-CAM are employed.Conclusion: The prognosis for AD is significantly improved by BBO-optimized VMD with transfer learning, providing a scalable and precise instrument for early recognition, tracking, and individualized behavior.