FCN Based Deep Learning Architecture for Medical Image Segmentation

By eliminating the laborious and time-consuming task of manually drawing contours around organs and tumors, deep learning algorithms have revolutionized the area of medical image segmentation. Deep learning can significantly improve treatment outcomes for patients with gastrointestinal cancer and many other medical illnesses where precise and effective segmentation is essential for successful treatment by automating the segmentation process. Fully convolutional neural networks (FCNs), a type of convolutional neural network, have been researched for medical image segmentation. In this study, a U-Net model, a form of Fully Convolutional Network (FCN), is evaluated for its use in segmenting brain tumors. The precision and accuracy of medical image processing have been proven to be greatly improved by deep learning-based techniques. Deep learning model interpretability, generalizability across different imaging modalities, and the requirement for larger and more varied datasets are still problems that need to be overcome. This research work intends to provide a performance and generalizability analysis of the U-Net model for medical image segmentation in brain tumor detection.