Multimodal Attention Cascaded U_NET for 3d Brain Tumour Segmentation

Abstract

Medical imaging is crucial to cancer treatment, diagnosis, and response assessment. magnetic resonance imaging (MRI) is a popular acquisition modality for brain tumour investigation (e.g., conventional, and advanced). Tumours vary in shape, location, and appearance, making MRI segmentation is more typical. Despite several research, automatic tumour segmentation is still challenging in performance index. newline newlineThis research intends to automate the segmentation of brain tumour-associated low- grade gliomas using multimodal MRI images to aid radiologists in diagnosis and treatment. A novel method is proposed to identify unwanted growth in brain using SVM-PUK on convoluted textural features with reduced Gabor wavelet features. After pre-processing, GLCM features of image are extracted and further, convoluted with reduced Gabor features using PCA of the slices. Then, the convoluted GLCM features, and reduced Gabor features classified with the SVM using PUK kernel. The proposed method performance is evaluated on BRATS 18 database and achieved an accuracy of newline91.31 % in recognizing the effected tissues and shown better performance over baseline models. newline newlineMultimodal Attention-gated Cascaded U-Net (MAC U-Net) model is proposed to address the performance issues observed in the detection and segmentation of low-grade tumors. The effectiveness of group normalization with attention gate is also explored with skip connections to segment small-scale brain tumors using several highlighted salient features. The model is evaluated on the brain tumor benchmark dataset BraTS2018 over various performance metrics such as Dice, IoU, Sensitivity, Specificity, and Accuracy. Experimental results illustrate that the proposed MAC U-net on BraTS 2018 dataset outperforms baseline U-nets with 94.47, 84.12, and 82.72 dice similarity coefficient values on HGG and 85.71, 78.85 and 74.16 on LGG subjects with Ground Truth values of Complete Tumor, Tumor Core, and Enhancing tumor, respectively. newline