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

Haobo Chen,Jianwei Jiang,Jie Shi,Man Chen,Yuqun Wang,Wanying Chang,Qi Zhang,Jingyu Xiong

doi:10.1007/s40846-021-00670-8

Abstract

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

Highlights

Lymph nodes (LNs) assist the immune system in building an immune response, and lymph nodes (LNs) swell and develop lymphadenopathy in cases of invasion by cancer and immune disorders
Inspired by the above observations, we propose a U-Net based framework integrated with the Gabor-based anisotropic diffusion (GAD) to reduce speckle noise and morphological operations to fill echogenic hila, which allows automatic segmentation of entire LNs in ultrasound images
The image filtered with the GAD. (d) The segmentation result of the probability map. (e) The region of interest depicted as a binary mask. (f) Images after the morphological operations of the opening, closing and convex hull computing. (g) The result of morphological operations is marked in the original ultrasound image with a green contour

Summary

Introduction

Lymph nodes (LNs) assist the immune system in building an immune response, and LNs swell and develop lymphadenopathy in cases of invasion by cancer and immune disorders. Adequate assessment of LN status is crucial to diagnose diseases and make treatment decisions. Ultrasound is generally the preferred method for the diagnosis of lymphadenopathy due to its real-time imaging, non-invasiveness, vast availability, and flexibility. In order to quantitatively assess lymphadenopathy using ultrasonography, it requires image segmentation for localizing areas of LNs and finding their borders. Segmentation of LNs in ultrasound images is generally performed manually by professional experts such as experienced radiologists or ultrasonologists, which is very time-consuming, tedious and subjective. Due to the slow process and tedious nature of the manual segmentation approaches, there is a critical demand for computer algorithms that segment images automatically, accurately and quickly without human interactions

Methods

Results

Discussion

Conclusion