Employing deep learning and transfer learning for accurate brain tumor detection

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Machine Learning

Background:

• Brain tumor diagnosis is challenging due to complex brain structures.
• Magnetic resonance imaging (MRI) is the preferred imaging modality.
• Artificial intelligence (AI) and deep learning enhance diagnostic accuracy.

Purpose of the Study:

• To evaluate deep transfer learning architectures for improved brain tumor diagnosis accuracy.
• To assess the performance of ResNet152, VGG19, DenseNet169, and MobileNetv3.
• To explore the utility of transfer learning in medical imaging with limited labeled data.

Main Methods:

• Four deep transfer learning models (ResNet152, VGG19, DenseNet169, MobileNetv3) were trained and validated.
• A Kaggle dataset comprising pituitary, normal, meningioma, and glioma images was used.
• Image enhancement techniques and five-fold cross-validation were employed.

Main Results:

• MobileNetv3 achieved the highest diagnostic accuracy at 99.75%.
• All evaluated transfer learning models demonstrated high performance.
• The study confirmed the effectiveness of transfer learning for brain tumor classification.

Conclusions:

• Deep transfer learning architectures, particularly MobileNetv3, offer a highly accurate approach to brain tumor diagnosis.
• This methodology has the potential to significantly advance the field of neuro-oncology and medical imaging analysis.
• Transfer learning provides a powerful solution for medical image analysis tasks with data scarcity.