A Novel MRI Diagnosis Method for Brain Tumor Classification Based on CNN and Bayesian Optimization.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Machine Learning

Background:

• Brain tumors are aggressive, necessitating early diagnosis for effective treatment and improved patient quality of life.
• Manual interpretation of Magnetic Resonance Imaging (MRI) for brain tumor diagnosis is time-consuming and error-prone.
• Automated methods are crucial for efficient and accurate brain tumor identification.

Purpose of the Study:

• To develop an efficient hyperparameter optimization technique for Convolutional Neural Networks (CNNs) using Bayesian Optimization for brain tumor classification.
• To automate the identification of common brain tumors from CE-MRI images.

Main Methods:

• Proposed a Bayesian Optimization-based hyperparameter tuning method for CNNs.
• Evaluated the model on 3064 T-1-weighted CE-MRI images, classifying Glioma, Meningioma, and Pituitary tumors.
• Compared the optimized CNN performance against five pre-trained deep learning models using Transfer Learning.

Main Results:

• The optimized CNN achieved a maximum validation accuracy of 98.70% without data augmentation or cropping.
• Pre-trained models achieved lower accuracies: VGG16 (97.08%), VGG19 (96.43%), ResNet50 (89.29%), InceptionV3 (92.86%), and DenseNet201 (94.81%).
• The proposed model demonstrated superior performance compared to state-of-the-art methods on the CE-MRI dataset.

Conclusions:

• Bayesian Optimization is an effective technique for automating CNN hyperparameter tuning in medical image analysis.
• The developed method offers a feasible and highly accurate approach for automated brain tumor classification.
• This AI-driven strategy enhances diagnostic efficiency and accuracy for critical diseases like brain tumors.