Optimized deep learning architecture for brain tumor classification using improved Hunger Games Search Algorithm.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computational Biology

Background:

• Brain tumors are a severe disease requiring early detection for better patient outcomes.
• Accurate classification of brain tumors is crucial for effective treatment planning.
• Current deep learning models for brain tumor classification face challenges in precision and require expert input.

Purpose of the Study:

• To develop a highly efficient and accurate deep learning model for classifying multiple types of brain tumors.
• To address the limitations of existing models by integrating an improved metaheuristic algorithm for hyperparameter optimization.

Main Methods:

• An optimized residual learning architecture was developed for brain tumor classification.
• An improved Hunger Games Search algorithm (I-HGS), incorporating Local Escaping Operator (LEO) and Brownian motion, was proposed.
• The I-HGS algorithm was used to optimize the hyperparameters of the Residual Network 50 (ResNet50) model, creating the I-HGS-ResNet50.

Main Results:

• The I-HGS algorithm demonstrated superior performance over standard algorithms on CEC'2020 test functions.
• The I-HGS-ResNet50 model achieved high accuracy rates of 99.89%, 99.72%, and 99.88% on three distinct brain MRI datasets.
• Comparative experiments showed the I-HGS-ResNet50 model outperformed established deep learning architectures like VGG16, MobileNet, and DenseNet201.

Conclusions:

• The proposed I-HGS-ResNet50 model offers a significant advancement in automated brain tumor classification.
• The integration of enhanced metaheuristic algorithms with deep learning architectures shows great potential for medical diagnostic tools.
• This approach can lead to more accurate and efficient diagnosis, ultimately improving patient care for brain tumor patients.