Brain tumor detection and segmentation using u-net algorithm
Loading...
Files
Date
2025
Journal Title
Journal ISSN
Volume Title
Publisher
NHCE
Abstract
Medical image segmentation plays a crucial role in the healthcare domain, enabling accurate diagnosis and effective treatment planning. This project focuses on the segmentation of brain MRI images using a hybrid approach that combines the deep learning-based U-Net architecture with traditional image processing techniques. The objective is to enhance segmentation accuracy, particularly in dealing with challenges like noise, irregular boundaries, and varying image quality in medical data.
The U-Net model, specifically designed for medical image segmentation, serves as the foundation of this system due to its ability to capture detailed spatial and contextual features. Complementing the U-Net’s capabilities, traditional image processing methods such as thresholding, watershed segmentation, and morphological operations are applied to refine segmentation results. This combination ensures precise delineation of brain structures, overcoming the limitations of individual methods.
To prepare the dataset for optimal performance, preprocessing steps including resizing, normalization, and data augmentation are implemented. These steps improve the robustness of the system, enabling it to generalize well across diverse datasets. The system's performance is evaluated using standard metrics like the Dice coefficient, Jaccard index, and F1-score, which provide a comprehensive understanding of segmentation accuracy. Visualization techniques are also employed to make the results interpretable and suitable for clinical applications.