Liver Tumour Semantic Segmentation Using Segan

Abstract

Liver tumor detection and segmentation are critical tasks in medical imaging, aiding in early diagnosis and treatment planning. Convolutional neural networks (CNNs) have demonstrated significant success in the examination of medical images, particularly in semantic segmentation. In this study, we proposed a system called Semantic Segmentation with Adversarial Network (SegAN) for the accurate and efficient semantic segmentation of liver tumors from medical images. SegAN combines the power of generative adversarial networks (GANs) with CNN to address the challenges of limited annotated data and class imbalance in liver tumor segmentation (LiTS). The SegAN architecture have a generator network that produces high-quality tumor segmentation masks and a discriminator network that enforces segmentation mask realism. Our method leverages a collection of medical images to train the network and adapts GAN-based adversarial training to enhance segmentation accuracy. We evaluate SegAN on a diverse data set of liver tumor images and contrast it to other state-of-the-art segmentation methods. The output demonstrate that SegAN achieves superior accuracy and robustness, outperforming existing techniques in LiTS tasks.