Year: 2024
Type: article
Abstract: Accurate segmentation of brain tumors from medical imaging are essential for diagnosis and treatment planning. Deep learning has shown promise in automating this task, but challenges remain in achieving robust results. Several conventional models, like ANN, CNN, and fuzzy C-means, are used to segment the regions of the brain. However, most models face over-segmentation due to sensitive region data appearing in the MRI brain tumor image slices. The proposed models present a novel hybrid deep learning framework by integrating U-Net with DeeplabV3 to achieve better-segmented brain tumour segmentation accuracy, advancing the current state of the art. The experiment results showed better improvement than the conventional model in terms of accuracy and dice coefficients.
Authors: Poornac handu, Srin ivas
Language: English
Subfield: Neurology
Field: Neuroscience
Domain: Life Sciences
Open Access status: diamond
OpenAlex ID: W4407003218