@article{Fettah_Goumidi_Daho_2021, title={Deep Learning Model for Magnetic Resonance Imaging Brain Tumor Recognition}, volume={5}, url={http://worldascience.com/journals/index.php/wassn/article/view/33}, abstractNote={<p class="MDPI17abstract">Human interpretation of a large quantity of Magnetic Resonance Imaging (MRI images) is a tiring task and depends on the practitioner’s expertise and experience. Glioma is one of the most common and dangerous types of primary brain tumors, and its early diagnosis could be life-saving. Precise and fully automatic classification of Glioma on MRI images helps physicians diagnose and monitor patients. <br>In this work, we propose an automatic system to aid in diagnosing Glioma by classifying brain tumors into two categories: High-Grade Glioma (HGG) and Low-Grade Glioma (LGG). To perform this task, we trained three deep learning models (VGG-16, ResNet-50, and Inception-V3) on four brain MRI datasets (one for each MRI modality). To further improve tumor classification, non-tumorous slices were removed from the HGG class of the selected dataset and then were separately used to train the three models. Evaluations on BraTS 2019 attest that T1 presents the most discriminative features with 0.9513, 0.907, and 0.9487 for accuracy, sensitivity, and specificity, respectively. The Inception-V3 model outperforms the other models with 0.9975, 0.9894, and 1 for accuracy, sensitivity, and specificity. Experimental results demonstrate that using the Inception-V3 model with T1 modality can achieve good performances.</p>}, number={1}, journal={WAS Science Nature (WASSN) ISSN: 2766-7715}, author={Fettah, Abir and Goumidi , Bouchra and Daho , Mostafa}, year={2021}, month={Dec.}, pages={1–11} }