Bacterial Foraging Optimization-boosted convolutional neural network for brain tumor detection using MRI images

Accurate brain tumor classification from magnetic resonance imaging (MRI) is essential for early diagnosis and treatment planning; however, existing automated approaches often face challenges related to feature redundancy, computational complexity, and limited training data. In this study, we propose a hybrid Stacked Network Model (SNM) that combines deep learning-based feature extraction, Bacterial Foraging Optimization (BFO)-based feature selection, and traditional machine learning classifiers for binary brain tumor classification. Using an augmented MRI dataset comprising 1,574 tumor-positive and 1,572 tumor-negative images, deep features were extracted from four pre-trained convolutional neural networks, namely ResNet50, Xception, InceptionV3, and DenseNet121. The optimized features were subsequently classified using Logistic Regression, Support Vector Classifier, Random Forest, Decision Tree, XGBoost, K-Nearest Neighbor, and Naive Bayes classifiers. Among all configurations, the ResNet50-BFO-LR pipeline achieved the best performance, attaining 99.81% accuracy, precision, recall, and F1-score, while reducing the feature dimension from 2,048 to 991 features. The proposed framework also demonstrated efficient inference with an average processing time of 0.11 s per image. These findings suggest that the integration of deep feature extraction, BFO-based optimization, and traditional machine learning classifiers provides an accurate and computationally efficient solution for MRI-based brain tumor classification, although further validation on external multicenter datasets is required before clinical deployment.