Brain tumor detection using inpainting and deep ensemble model

Abstract

In this work, the authors have applied image inpainting on MRI images of the brain to highlight the tumors present in the image. These highlighted images are used for the training of the ensemble model. Three convolutional neural networks (CNNs) are used as the base classifier and their outputs are fed to a Multilayer Perceptron (MLP) for further training and final classification. Classification is done to check whether the brain is having a tumor or it is healthy using a data set that is available at Kaggle for open access. The proposed method provided 100% and 98.33% training and testing accuracies that show the effectiveness of applying inpainting on the data set images.

Subject Classification:

• 68Q85
• 68T45
• 68U10

Keywords:

• Brain tumor detection
• Image inpainting
• Convolutional neural network
• Multilayer perceptron
• Ensemble learning