A Sequential Machine Learning-cum-Attention Mechanism for Effective Segmentation of Brain Tumor

• DOI: 10.3389/fonc.2022.873268
• Publication Date: 2022-06-01T13:47:07Z
• Authors: Tahir Mohammad Ali; Ali Nawaz; Attique Ur Rehman; Rana Zeeshan Ahmad; Abdul Rehman Javed; Thippa Reddy Gade...; Chin-Ling Chen; Chih-Ming Wu
• Keywords: Artificial intelligence; Brain Tumors; Classification of Brain Tumor Type and Grade; 02 engineering and technology; Image Segmentation; Quantum mechanics; Segmentation; Artificial Intelligence; Object Detection; Machine learning; 0202 electrical engineering, electronic engineering, information engineering; Psychology; Theory and Applications of Extreme Learning Machines; RC254-282; Physics; UNET; Neoplasms. Tumors. Oncology. Including cancer and carcinogens; Life Sciences; Mechanism (biology); Transfer Learning; brain tumor segmentation; Computer science; FOS: Psychology; BRATS; Oncology; Neurology; Computer Science; Physical Sciences; Deep Learning in Computer Vision and Image Recognition; Semantic Segmentation; Computer Vision and Pattern Recognition; attention mechanism; VGG19; MRI; Neuroscience

Magnetic resonance imaging is the most generally utilized methodology that permits radiologists to look inside cerebrum using radio waves and magnets for tumor identification. However, it tedious complex identify tumorous nontumorous regions due complexity in region. Therefore, reliable automatic segmentation prediction are necessary of brain tumors. This paper proposes a efficient neural network variant, i.e., an attention-based convolutional segmentation. Specifically, encoder part UNET pre-trained VGG19 followed by adjacent decoder parts with attention gate noise induction denoising mechanism avoiding overfitting. The dataset we BRATS’20, which comprises four different MRI modalities one target mask file. abovementioned algorithm resulted dice similarity coefficient 0.83, 0.86, 0.90 enhancing, core, whole tumors, respectively.