Downloads: 3 | Views: 39 | Weekly Hits: ⮙3 | Monthly Hits: ⮙3
Student Project | Computer Science & Engineering | India | Volume 14 Issue 3, March 2025 | Rating: 5.5 / 10
Enhanced Detection of 3D Brain Tumors through a Deep Learning Approach
Lakshmidevi S, Arun R
Abstract: Detection and segmentation of brain tumors in magnetic resonance imaging (MRI) are essential for clinical diagnoses and therapy planning. This research is on using segmentation algorithms to improve the detection of brain cancers in MRI data. The study seeks to delineate tumor locations with high accuracy by using modern techniques, including Convolutional Neural Networks (CNNs) and U-Net architectures. The proposed method incorporates multi-modal MRI data, including T1, T2, and FLAIR sequences, to get a holistic perspective of tumor shape. Automated segmentation alleviates the manual burden on radiologists and diminishes subjectivity, providing more consistent and precise outcomes. The research contrasts conventional segmentation methods, including thresholding and region-growing, with deep learning approaches to emphasize the benefits of contemporary machine learning in attaining superior accuracy and resilience. The approach has enhanced efficacy in identifying tiny and irregularly shaped tumors, as shown by comprehensive testing on publically accessible datasets such as BraTS. Evaluation measures, such as the Dice Similarity Coefficient (DSC) and Intersection over Union (IoU), are used to assess the efficacy of the algorithms. The results highlight the efficacy of deep learning-based segmentation as a dependable instrument for medical imaging, hence enhancing diagnostic accuracy, treatment planning, and patient outcomes. Future research may investigate the integration of these segmentation outcomes with radiomics to improve tumor characterization.
Keywords: Detection and segmentation, Brain Tumor, Magnetic Resonance Imaging, Convolutional Neural Networks, Dice Similarity Coefficient, and Intersection over Union
Edition: Volume 14 Issue 3, March 2025,
Pages: 1156 - 1160