Brain tumor detection in the Spark system

Abstract

Machine learning (ML) and computer vision systems revolutionized the world, especially deep learning (DL) for convolutional neural networks, which has proven breakthroughs in brain tumor (BT) diagnosis. This study investigates a convolutional neural network (CNN) approach for image classification for BT detection using the EfficientNetB1 architecture with global average pooling (GAP) layers in a big data setting. A classification layer is done with a softMax layer. The system is created in the Apache Spark environment. Spark system is a unified and ultra-fast analysis engine for large-scale data processing. It is mainly dedicated to big data and deep learning (DL). Experiments are carried out using the brain magnetic resonance imaging (MRI) dataset containing 3,264 MRI scans to predict the performance of the model. The dataset is decomposed into two datasets. The model's performance was assessed and compared to existing models, it yielded a high precision, precision, and f1-score. In our work, we have achieved an accuracy of 97% and a performance of 98% on a dataset of 3,064 brain MRI images.