Segmentation of Lymph Nodes in Ultrasound Images Using U-Net Convolutional Neural Networks and Gabor-Based Anisotropic Diffusion

The automated segmentation of lymph nodes (LNs) in ultrasound images is challenging, largely because speckle noise and echogenic hila. This paper proposes a fully automatic accurate method for LN that overcomes these issues. proposed integrates diffusion-based despeckling, U-Net convolutional neural networks morphological operations. First, the suppressed node edges are enhanced using Gabor-based anisotropic diffusion (GAD). Then, modified model used to segment LNs excluding any Finally, operations undertaken entire by filling regions occupied A total 531 from 526 patients were segmented method. Its performance was evaluated terms its accuracy, sensitivity, specificity, Jaccard similarity Dice coefficient, which it achieved values 0.934, 0.939, 0.937, 0.763 0.865, respectively. automatically accurately segments images, enhancing prospects being able undertake artificial intelligence (AI)-based diagnosis diseases.