A Novel Deep Learning Approach for Brain Tumors Classification Using MRI Images

Abstract

Early detection of brain tumors (BTs) can save valuable lives. BTs classification is usually accomplished by using magnetic resonance imaging (MRI), which is commonly carried out earlier than definitive talent surgery. Machine learning (ML) strategies can assist radiologists to diagnose tumors barring invasive measures. One of the challenges of traditional classifiers is that they rely on informative hand-crafted features, which can be a time-consuming process to extract. We proposed fully automatic framework for BTs classification with weighted contrast-enhanced MRI images. The proposed framework includes an enhancement preprocessing to improve input images quality and a classification phase for images classification into three classes of tumors (meningioma, glioma and pituitary tumor) and ordinary cases. The model was built used "Lightweight Convolutional Neural Network (LWCNN)" that allows to automatically extract features. We tested the LWCNN model in two experiments. In the first one, the model has been tested with original datasets. We tested our proposed framework on the same dataset after enhancing the features of MRI images in the second experiment. As per the experiment results, it has been observed that the proposed framework achieves the desired outcome which demonstrates the effectiveness of our proposed framework.

Design and assessment of improved Convolutional Neural Network based brain tumor segmentation and classification system

“…The automated use of magnetic resonance imaging (MRI) is advised to ensure accurate diagnosis. 13 The top tumor spots are revealed using agglomerative clustering after pre-processing photographs to improve visual quality, extracting useful features using two preeducated deep mastering fashions, and creating a hybrid feature vector. The suggested method's classification accuracy was 98.95% when compared to the standard approaches.…”