AI-Assisted Multiclass Brain Tumor Classification from MRI Scans Using EfficientNetB0

Abstract

We present a deep learning framework for multiclass brain tumor classification from MRI scans, with a focus on achieving high diagnostic accuracy through the EfficientNetB0 architecture. The objective of this study addresses the clinical need for automated computer-aided diagnosis systems that can reliably distinguish between glioma, meningioma, pituitary tumors, and normal brain tissue. A dataset of approximately 7,000 MRI images, drawn from public repositories and anonymized hospital databases, was preprocessed through resizing to 224 × 224 pixels, Gaussian filtering, and min-max normalization. Data augmentation techniques—including rotation, flipping, zooming, and contrast adjustment—were applied to improve generalization. We compared four deep learning models: a custom convolutional neural network, VGG16, ResNet50, and EfficientNetB0. The models were trained using the Adam optimizer, categorical cross-entropy loss, and a batch size of 32 over 50 epochs. The dataset was split into 70% training, 15% validation, and 15% testing subsets. Among all architectures, EfficientNetB0 achieved the highest performance, with an accuracy of 98.2%, precision of 97.9%, recall of 98.0%, and an F1-score of 97.9%. The confusion matrix revealed minimal misclassification, with only slight overlap between glioma and meningioma categories. Training and validation curves demonstrated stable convergence and minimal overfitting, confirming robust training dynamics. The primary contribution of this work is the demonstration that a state-of-the-art convolutional network can provide near-perhaps for the first time in this specific multiclass context-diagnostic metrics exceeding 98%, thereby offering a reliable tool for radiologists