An efficient Alzheimer’s disease detection using optimized hybrid light GBM-based CNN

Alzheimer’s disease is a serious neurological disorder that destroys brain tissue, resulting in dementia or irreversible memory loss. Many people lose their lives to this incurable disease every year. On the other hand, early identification is essential to slowing its rapid growth. Recent studies encountered difficulties with overfitting, low accuracy, and the laborious nature of manual diagnosis. In order to resolve this, a hybrid light GBM-based CNN using Parabuteo Finch Optimization is developed for effective Alzheimer’s disease identification. The combined approach of multi-level Bayesian fuzzy clustering (BFC) texture-based segmentation aims to accurately distinguish and identify specific regions corresponding to grey matter, white matter, and cerebrospinal fluid within brain MRI images. The hybrid light GBM-based DCNN provides a standardized hybridization that enables the model to harness both structured and image data. This integration enhances the capacity to capture a comprehensive set of features relevant to Alzheimer’s disease detection, offering a more robust approach. The Parabuteo Finch Optimization (PFO) integration creates a versatile and efficient optimization strategy for diverse scenarios. Furthermore, the PFO algorithm enhances the overall performance and adaptability of the model by optimizing its key parameters for improved results. The research evaluates outcomes through metrics encompassing an accuracy of 94.48%, the Mathew correlation coefficient (MCC) of 0.91, the negative predictive value (NPV) of 0.92, the positive predictive value (PPV) of 0.95, and the threat score of 0.85 based on 90% of training. These results signify superior performance, establishing the effectiveness of the proposed model in comparison to alternative approaches.