A novel adaptive dual swarm intelligence based image quality enhancement approach with the modified SegNet -RBM-based Alzheimer Segmentation and classification

• Published in: Multimedia tools and applications Vol. 83; no. 10; pp. 29261 - 29288
• Main Authors: Anitha, Raju, Dasari, Durga Bhavani, Vivek, P. Sandalya Sai, Kakarla, Naga Madhavi Latha, Kumar, M. Sirish
• Format: Journal Article
• Language: English
• Published: New York Springer US 01.03.2024; Springer Nature B.V
• Subjects: Adaptive algorithms; Alzheimer's disease; Artificial neural networks; Classification; Computer Communication Networks; Computer Science; Data collection; Data Structures and Information Theory; Error analysis; Geometric transformation; Heuristic methods; Image enhancement; Image filters; Image quality; Machine learning; Magnetic resonance imaging; Medical imaging; Multimedia Information Systems; Neural networks; Noise reduction; Optimization; Optimization algorithms; Performance evaluation; Performance measurement; Recurrent neural networks; Root-mean-square errors; Special Purpose and Application-Based Systems; Swarm intelligence; Restricted Boltzmann Machine; SegNet; Adaptive salp swarm optimization algorithm; Adaptive wind driven optimization algorithm; Alzheimer's disease
• ISSN: 1573-7721; 1380-7501; 1573-7721
• DOI: 10.1007/s11042-023-16486-4

In this research, a novel metaheuristic-based approach for enhancing the image quality of Magnetic resonance imaging (MRI) scans for Alzheimer’s disease (AD) classification and segmentation is proposed. The proposed approach involves four phases: data collection, preprocessing, segmentation, and classification. The dataset used in this study is collected from an open-source internet platform and contains four classes of AD. To improve the quality of the collected images, various enhancement techniques are employed, such as local entropy-weighted histogram equalization using an adaptive wind-driven optimization algorithm and guided filter-based image denoising using an adaptive Salp Swarm Optimization algorithm. In addition to this, geometric transformations, image augmentation, noise reduction, and grayscale conversion are also utilized. The proposed segmentation model employs a drop-block in Segmentation Network (SegNet) for regularized feature learning, resulting in highly efficient pixel-wise segmentation. A modified loss function is also introduced for fine-tuning the performance of SegNet. The classification layer of SegNet utilizes a Restricted boltzmann machine (RBM) model for classifying AD. Performance evaluation of the proposed approach is done by considering performance metrics such as accuracy, precision, recall, F1-measure, Root means square error (RMSE), and mean absolute error (MAE). The proposed approach outperforms existing techniques such as Convolutional neural network (CNN), Recurrent neural network (RNN), Long short-term memory (LSTM), Understanding networks (UNET), and SegNet, demonstrating the effectiveness of the proposed metaheuristic-based image quality enhancement approach. The implementation of the proposed approach is carried out using Matlab.