Year: 2026
Type: article
Abstract: Brain tumors constitute the most dangerous type of neurological disorder, where delayed or incorrect diagnoses may significantly affect patient survival chances. Owing to the fast development of artificial intelligence, Machine learning (ML) and Deep Learning (DL) approaches along with the Magnetic Resonance Imaging (MRI) method proved to be efficient solutions for brain tumor detection and classification. This research paper presents the novel DNN-based method for precise detection of brain tumors based on MRI images of the BraTS 2020 dataset. In total, 2,892 MRI images were used for the training of the model, which included T1-weighted, T2-weighted, and FLAIR modalities. The preprocessing of the images included data cleaning, resizing, augmentation, one-hot encoding, and Z-score normalization of the images to increase their quality and model generalization. The information set was divided into 70:20:10 training, validation, and testing sets. The proposed DNN framework could learn discriminative tumor features and demonstrated better classification results than CNN, FCN, ResNet50, and InceptionNet models. As a result, the proposed DNN provided the following values: Accuracy (ACC) - 99.80%, Precision (PRE) - 99.85%, Recall (REC) - 99.84% and F1-Score (F1) - 99.81%. These findings Demonstrate the efficacy and dependability of the devised method, which holds tremendous potential as a trustworthy computer-aided diagnostic tool for brain tumour diagnosis.
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7218051182