Convolutional Neural Network-Machine Learning Model: Hybrid Model for Meningioma Tumour and Healthy Brain...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Machine Learning

Background:

• Brain tumors are complex and can be difficult to detect with the human eye, even with MRI.
• Early detection of brain tumors is crucial for effective treatment and patient outcomes.
• AI tools offer a promising avenue for improving the accuracy and efficiency of brain tumor diagnosis.

Purpose of the Study:

• To investigate the effectiveness of hybrid AI models for brain tumor classification using MRI.
• To compare the performance of various convolutional neural network (CNN) architectures and machine learning (ML) models.
• To identify the optimal combination of CNN and ML for accurate meningioma tumor detection.

Main Methods:

• Utilized original and pre-trained CNNs (DenseNet169, EfficientNetV2B0) for feature extraction from brain MRI.
• Integrated ML models (Random Forest, KNN, SVM) with CNNs, replacing the SoftMax layer for classification.
• Employed the bagging ensemble method to prevent overfitting and generalize results.

Main Results:

• The EfficientNetB0-SVM combination achieved a high accuracy of 99.5% in binary classification of tumors and healthy brains on the test dataset.
• Comparison studies evaluated different CNN architectures and ML model integrations.
• The study demonstrated the potential of hybrid AI approaches in medical image analysis.

Conclusions:

• Hybrid AI models, particularly the EfficientNetB0-SVM combination, show exceptional accuracy for brain tumor classification from MRI.
• Transfer learning and machine learning integration offer a robust solution for detecting subtle abnormalities.
• The findings support the use of advanced AI techniques to aid radiologists in diagnosing brain tumors.