Automated Brain Tumor Detection using Ideal Shallow Neural Network with Artificial Jellyfish Optimization

Area of Science:

• Medical Imaging Analysis
• Artificial Intelligence in Healthcare
• Computational Neuroscience

Background:

• Automated brain tumor prediction from MRI and CT scans faces challenges with computational complexity and accuracy.
• Existing automated tools often require significant human intervention for precise tumor area identification.
• There is a need for advanced computational models to enhance accuracy and reduce complexity in brain tumor analysis.

Purpose of the Study:

• To develop an improved automated system for brain tumor classification and segmentation using Magnetic Resonance Imaging (MRI).
• To enhance prediction accuracy and reduce computational complexity in brain tumor analysis.
• To evaluate the efficacy of an Ideal Shallow Neural Network (ISNN) optimized with an Artificial Jellyfish Optimization (AJO) algorithm.

Main Methods:

• Utilized MRI images for enhanced informativeness, employing Gabor filtering for noise reduction and histogram equalization for boundary enhancement.
• Implemented an Ideal Shallow Neural Network (ISNN) optimized by the Artificial Jellyfish Optimization (AJO) algorithm for feature dimensionality reduction and classification.
• Employed Centroid Weighted Segmentation (WCS) combined with the Grasshopper Optimization Algorithm (GOA) for improved segmentation of brain tumor boundaries.

Main Results:

• The proposed AJO-ISNN model achieved a classification accuracy of 95.14%, significantly outperforming Convolutional Neural Network (CNN) at 85.41% and VGG 19 at 93.75%.
• The Centroid Weighted Segmentation with Grasshopper Optimization Algorithm (CWS-GOA) demonstrated a Dice Similarity Coefficient of 93.15% on both BRATS and Kaggle datasets.
• The model achieved efficient classification and segmentation, processing approximately 200 images in about 65 seconds.

Conclusions:

• The AJO-ISNN model offers superior accuracy and efficiency for brain tumor classification and segmentation compared to existing methods like multi-cascaded CNN and InceptionV3.
• The integration of AJO for feature optimization and GOA for segmentation significantly enhances the performance of the neural network model.
• This research presents a computationally efficient and highly accurate approach for automated brain tumor analysis from medical imaging data.