Classification of Brain MRI Tumor Images Based on Deep Learning PGGAN Augmentation.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Accurate brain tumor classification is crucial for effective treatment planning.
• Convolutional Neural Networks (CNNs) aid in classifying brain tumors from MRIs but can overfit with limited data.
• Data augmentation is a key strategy to improve CNN performance in medical image analysis.

Purpose of the Study:

• To evaluate a novel framework for brain tumor MRI classification.
• To assess the efficacy of a VGG19 feature extractor combined with a progressive growing generative adversarial network (PGGAN) for data augmentation.
• To improve the accuracy of classifying common brain tumor types.

Main Methods:

• Utilized a VGG19 feature extractor and three distinct classifiers.
• Employed a progressive growing generative adversarial network (PGGAN) to generate synthetic brain tumor MRIs.
• Trained and validated the model on a dataset of brain tumor MRIs.

Main Results:

• The proposed framework achieved a high classification accuracy of 98.54% for gliomas, meningiomas, and pituitary tumors.
• The PGGAN augmentation effectively addressed the issue of limited training data, enhancing model robustness.
• The system demonstrated superior performance compared to previous brain tumor classification studies.

Conclusions:

• The developed framework offers a highly accurate and efficient method for brain tumor classification from MRIs.
• PGGAN-based data augmentation is effective in improving deep learning model performance for medical imaging tasks.
• This approach holds significant potential for improving diagnostic accuracy and patient outcomes in neuro-oncology.