Dr. Sujata Dash is a Professor and IEEE Senior Member in the Department of Information Technology at Nagaland University, India. She has over three decades of academic and research experience, including as a Visiting Professor at the University of Manitoba, Canada. Her research interests include Artificial Intelligence, Machine Learning, Deep Learning, Bioinformatics, NLP, Metaheuristics, and IoT. She has published extensively in international journals, books, and conferences and has supervised numerous research scholars.

Benchmarking Metaheuristic-Based Feature Selection for Multi-Class Imbalanced Medical Image Classification                                                                                                                                                           Abstract: Medical image classification is crucial for intelligent healthcare systems. However, the performance of these classifications is often hindered by high-dimensional feature spaces and significant class imbalance in medical datasets. Most existing studies use a single metaheuristic optimisation technique and primarily focus on binary classification problems, with limited research on systematic benchmarking under multi-class imbalanced conditions. To address this issue, this study proposes a comprehensive benchmarking framework to evaluate metaheuristic-based feature selection techniques specifically designed for multi-class imbalanced medical image classification. The proposed framework integrates deep feature extraction using a pretrained ResNet50 convolutional neural network, along with imbalance-aware preprocessing and metaheuristic optimization. Initially, medical images from benchmark datasets, including the BRATS brain tumor MRI images, COVID-19 radiography images, and ECG arrhythmia datasets, are preprocessed and transformed into deep feature embeddings. To enhance the representation of minority classes, the Synthetic Minority Oversampling Technique (SMOTE) is applied before optimizing the features. Subsequently, five metaheuristic algorithms, Genetic Algorithm (GA), Particle Swarm Optimization (PSO), Ant Colony Optimization (ACO), Differential Evolution (DE), and Grey Wolf Optimizer (GWO), are utilized to identify optimal discriminative feature subsets. The optimized feature subsets are assessed using machine learning classifiers, including Random Forest and Support Vector Machine. This evaluation considers various performance metrics, such as accuracy, precision, recall, F1-score, runtime complexity, and feature reduction rate. Additionally, we conduct convergence analysis and comparative benchmarking to evaluate the optimization behavior, computational efficiency, and generalization capability of each algorithm across different imbalance scenarios. The proposed study aims to identify effective optimization strategies for medical image classification that accounts for class imbalance. It seeks to reduce computational complexity while enhancing the prediction performance for minority classes. The findings will offer foundational insights for the future development of hybrid explainable deep learning methods and metaheuristic optimization frameworks in intelligent healthcare systems. Keywords: Medical image classification, metaheuristic optimization, feature selection, class imbalance, deep learning, ResNet50, BRATS, Grey Wolf Optimizer, computational intelligence.