Convoluted neural network and transfer learning algorithm for improved brain tumor classifications in MRI

Abstract

Listen

Translate article

Artificial intelligence (AI) has made significant use cases to improve patient care, particularly in medical image analysis. This study aims to develop a deep-learning model for disease classification in medical images and compare its performance in four-class MRI and two-class X-ray classification tasks. We utilize Convolutional Neural Networks (CNNs) for diagnosing pneumonia from chest X-rays and various tumors from brain MRIs, leveraging transfer learning to improve performance. Transfer learning, which reuses pre-trained models like VGG-16, is more efficient than building models from scratch. The VGG-16 model, pre-trained on over a million ImageNet images, achieved 92.7% accuracy. By fine-tuning, we reached 93.6% accuracy. Data augmentation techniques, such as flipping, rotation, and brightness adjustments, further enhance classification accuracy and performance.