Abstract:Predictive models in radiology can now be constructed with remarkable accuracy using an amalgamation of radiomics and artificial intelligence. To effectively prepare for surgical procedures and assess the progression of the tumor, accurate segmentation of gliomas is essential. The current study aims to address a segmentation of the whole tumor (WT), tumor core (TC), and enhancing tumor (ET), three partially overlapping regions of interest within the glioma with two variants, high-graded glioma (HGG) and low-graded glioma (LGG), made available through the BraTS 2019, 2020, and 2021 challenges. The traditional approach has been bypassed by focusing only on the network architecture, but rather the proposed research work is also concentrating on data pre-processing, augmentation, training, and testing strategies to improve the performance of the automatic brain tumor segmentation. UNet and its variants have recently been shown to be effective in automatically segmenting brain tumors from volumetric multi-modal magnetic resonance (MR) images. Motivated from the literature, an improved UNet + + framework (ResUNet + +) is proposed to segment multi-modal volumetric MR images of brain tumor. The ResUNet + + is a 3D (three-dimensional) encoder-decoder model where the encoder path is replaced with the pre-trained backbone of the ResNet50 model. Moreover, the standard convolutional blocks of the traditional UNet architecture are substituted with the 3D dense convolutional blocks, and in the decoder phase, convolutional layers are replaced by the convolutional transpose layers (ConvTranspose), outperforming the existing models in terms of segmenting the WT, TC, and ET in both HGG and LGG. The performance of the ResUNet + + framework is evaluated using five different performance parameters, and when compared with the state-of-the-art models, the results demonstrate the effectiveness of the framework.

MR brain segmentation based on DE-ResUnet combining texture features and background knowledge

Improved UNet‐based Magnetic Resonance Imaging Segmentation of Demyelinating Diseases with Small Lesion Regions

DeU-Net 2.0: Enhanced deformable U-Net for 3D cardiac cine MRI segmentation

Brain tumor magnetic resonance image segmentation by a multiscale contextual attention module combined with a deep residual UNet (MCA-ResUNet)

MRI Image Segmentation of Brain Tumor Based on Res-UNet

MI-UNet: Multi-Inputs UNet Incorporating Brain Parcellation for Stroke Lesion Segmentation From T1-Weighted Magnetic Resonance Images

mResU-Net: multi-scale residual U-Net-based brain tumor segmentation from multimodal MRI

Automated Segmentation of Brain Gray Matter Nuclei on Quantitative Susceptibility Mapping Using Deep Convolutional Neural Network

DeepMRSeg: A convolutional deep neural network for anatomy and abnormality segmentation on MR images

Subcortical Brain Segmentation in MR Image Based on Residual Fully Convolutional Networks

DAUnet: A U-shaped network combining deep supervision and attention for brain tumor segmentation

DR-Unet104 for Multimodal MRI brain tumor segmentation

ResUNet + + : a comprehensive improved UNet + + framework for volumetric semantic segmentation of brain tumor MR images

Dimensionality Reduction Hybrid U-Net for Brain Extraction in Magnetic Resonance Imaging

Research on the magnetic resonance imaging brain tumor segmentation algorithm based on DO‐UNet

Segmentation of brain tumor MRI image based on improved attention module Unet network

Matpr-Unet: A Multi Attention Two-Path Residual Unet for Focal Cortical Dysplasia Lesions Segmentation.

A Strategy of MR Brain Tissue Images' Suggestive Annotation Based on Modified U-Net

MBDRes-U-Net: Multi-Scale Lightweight Brain Tumor Segmentation Network

DASN: Data-Aware Skilled Network for Accurate MR Brain Tissue Segmentation.

Residual Block Based Nested U-Type Architecture for Multi-Modal Brain Tumor Image Segmentation