Abstract
Segmenting brain tissue can provide valuable insights into its structure and function. Magnetic resonance imaging (MRI)-based tissue segmentation is an essential procedure for improving tractography and quantifying brain microstructure. In this work, a novel BTS-NEUNET framework is proposed for brain tissue segmentation based on multimodal MRI images. The multimodal MRI images, such as T1W, T2W, perfusion-weighted imaging (PWI), and diffusion-weighted imaging (DWI), undergo pre-processing using the wavelet transform-based bilateral (WTBB) filter and the curvelet transform-based adaptive Gaussian notch (CTBAGN) filter to enhance the image quality. A hybrid DenseGoogLe network is used to extract the relevant features from the enhanced multimodal images. The proposed BTS-NEUNET method uses the White Shark Optimization Algorithm to select features from MRI images. The four types of brain tissues such as grey matter, white matter, cerebrospinal fluid, and ischemic lesions are classified using a Deep Belief Network (DBN). Brain tissues are classified using a nested, attention-based U-Net. The proposed BTS-NEUNET method's performance is assessed using Accuracy, Precision, Recall, Specificity, and F1-Score. The proposed DenseGoogLeNet method for feature extraction achieves an overall Accuracy of 1.64 %, 4.53 %, 0.76 %, and 3.94 % higher than ShuffleNet, ResNet, GhostNet, and MobileNet, respectively. The proposed BTS-NEUNET method achieves the highest Accuracy rate of 99.60 %. The proposed BTS-NEUNET method improves overall Accuracy by 1.92 %, 1.34 %, and 1.74 % over existing methods such as DDSeg, optimal support vector machine (SVM), and chaotic based enhanced Firefly Algorithm integrated with Fuzzy C-Means (CEFAFCM), respectively.