High-precision brain tumor classification from MRI images using an advanced hybrid deep learning method with minimal radiation exposure

Abstract

Background

Accurate identification of brain tumors is critical to improving patient outcomes and minimizing unnecessary radiation exposure from imaging procedures. While Magnetic Resonance Imaging (MRI) is the gold standard for brain tumor detection, manual interpretation remains time-consuming, error-prone, and subject to inter-observer variability.

Objective

This study aims to develop a high-precision, automated MRI-based brain tumor classification model using a hybrid deep learning architecture to reduce diagnostic errors and support radiation exposure minimization strategies.

Methods

A novel hybrid deep learning model was developed by integrating the CE-EEN-B0 and ResGANet architectures. The model incorporates advanced feature selection and ensemble-based learning techniques to enhance classification performance across diverse datasets. The feature vectors extracted from MRI images were benchmarked against state-of-the-art (SOTA) deep learning classifiers, including InceptionV3, Vision Transformer, MobileNet, VGG-SCNet, DenseNet121, and ResNet50.

Results

The proposed hybrid model achieved an accuracy of 99.11 %, with a precision, recall, and F1-Score of 99.6 %. It also attained a specificity of 99.75 %, an error rate of just 0.01, and a Cohen's Kappa score of 99.10, outperforming all benchmark models.

Conclusion

The hybrid CE-EEN-B0-ResGANet model demonstrates high reliability and performance in MRI-based brain tumor classification. Its strong diagnostic metrics support its potential for clinical deployment as an effective, automated tool for aiding radiologists and minimizing unnecessary imaging interventions.