BTS-VNET: BRAIN TUMOUR SEGMENTATION VIA DEEP LEARNING BASED DUAL ATTENTION INTEGRATED V-NETWORK

Abstract

Abstract

Brain tumor occurs when abnormal cells form within the brain. There are two main types of tumors: malignant (cancerous) tumors and benign (non-cancerous). Brain Tumor Segmentation (BTS) is a medical image analysis task that involves the separation of brain tumors from normal brain tissue in magnetic resonance imaging (MRI) scans. Segmenting tumor automatically in human brain MRI images is challenging because of uneven, irregular and unstructured size and shape of the tumor. In this paper, a novel BTS-VNET has been proposed for brain tumor detection and segmentation using a deep learning approach. The process begins with input MRI scans that undergo pre-processing steps, including skull stripping and scalable range based adaptive bilateral filter (SCRAB) for removing the noisy artifacts. The pre-processed images are then passed through a Dilated Regularized Network for initial tumor detection for classifying the MRI images as either normal or abnormal. Afterwards, the abnormal images are fed into the tumor segmentation phase using a Dual Attention V-Network for resulting in segmented output that contains tumor regions. The proposed method is evaluated in terms of FScore, sensitivity, time taken, Segmentation Accuracy (SA), Average Misclassification Ratio (AMR), and Eye Perception-based Quality Index (EPQI).