Automated lung tumor segmentation

Lung tumor segmentation plays a critical role in the early detection and treatment of lung cancer. This project focuses on automating the segmentation process of lung tumors using deep learning techniques, specifically leveraging the power of U-Net and U-Net+ architectures. The dataset used for this study is sourced from Kaggle, containing pre-processed CT scan images of lung tumors. The segmentation model aims to accurately identify and localize the tumor regions within the lung, which is essential for diagnosis and treatment planning. U-Net, a widely used convolutional neural network architecture, is designed to handle medical image segmentation tasks by capturing both local and global context through its encoder-decoder structure. To further improve performance, we utilize U-Net+, an advanced version of U-Net, which incorporates additional modifications to enhance segmentation accuracy and deal with challenges like small tumor regions and unclear boundaries. The model training is performed on the pre-processed dataset, and the results are evaluated based on various performance metrics, such as Intersection over Union (IoU), Dice Coefficient, and accuracy. The outcomes of this study aim to provide a robust tool for radiologists to assist in tumor localization and enhance the efficiency of lung cancer diagnosis.

Keywords: Lung tumor segmentation, U-Net, U-Net+, deep learning, medical image processing, CT scan, tumor localization, image dataset, convolutional neural network, early detection, diagnosis.