Enhancing Dementia Detection with CNN: A Dragonfly Algorithm-Driven Approach Using Otsu and GLCM

A robust Neural Network based optimal dementia detection technique is presented in this study. Initially, the input brain images are improved in quality with an Adaptive Median Filter based preprocessing to reduce noise and enhance image quality. Following this, segmentation is performed using the Otsu method, which automatically thresholds the image to separate brain regions of interest from the background. For feature extraction, the Gray-Level Co-occurrence Matrix (GLCM) is utilized to capture essential texture characteristics that reflect subtle changes in brain structures indicative of dementia. The extracted features are then classified using a Dragonfly Algorithm-optimized Convolutional Neural Network (CNN), where the optimization process improves the model's classification accuracy by fine-tuning its parameters. Experimental outcomes demonstrate that the presented approach achieves high detection accuracy. In this paper has the comparison between the neural networks shows that the proposed Dragonfly Optimized CNN achieves 91.2% sensitivity and 93.4% for accuracy.