Using Convolutional Neural Network for edge detection in musculoskeletal ultrasound images

Abstract

Fast and accurate segmentation of musculoskeletal ultrasound images is an on-going challenge. Two principal factors make this task difficult: firstly, the presence of speckle noise arising from the interference that accompanies all coherent imaging approaches; secondly, the sometimes subtle interaction between musculoskeletal components that leads to non-uniformity of the image intensity. Our work presents an investigation of the potential of Convolutional Neural Networks (CNNs) to address both of these problems. CNNs are an effective tool that has previously been used in image processing of several biomedical imaging modalities. However, there is little published material addressing the processing of musculoskeletal ultrasound images, particularly using a panoramic technique. In our work we explore the effectiveness of CNNs when trained to act as a pre-segmentation pixel classifier that determines whether a pixel is an edge or non-edge pixel. Our CNNs are trained using two different ground truth interpretations. The first one uses an automatic Canny edge detector to provide the ground truth image; in the second interpretation, the ground truth was obtained using the same image marked-up by an expert anatomist. In this initial study the CNNs have been trained using half of the prepared data from one image, using the other half for testing; validation was also carried out using three unseen ultrasound images. CNN performance was assessed using Mathew's Correlation Coefficient, Sensitivity, Specificity and Accuracy. The results show that CNN performance when using expert ground truth image is better than in the case of using Canny ground truth image. Our technique is promising and has the potential to speed-up the image processing pipeline using appropriately trained CNNs.

Analysis of Edge Detection on Compressed Images with Different Complexities

“…F is within the limits of 0 ≤ F ≤ 1, ideally, F is equal to 1. The precision, also known as the positive predictive value is calculated [31]:…”

“…We are also witnessing an increase in the use of smart networks, the use of artificial intelligence to analyze, collect and process data. Such systems are mainly based on image processing and data processing, where the main processes are the extraction of a particular object from the scene, where edge detection and segmentation play an important role [31][32][33][34]. However, all of this gain particular weight and interest with the emergence and implementation of such systems on devices like Raspberry Pi and Arduino, which very often use real-time image processing, object detection and segmentation [35][36][37][38][39].…”

Improved CNN-Segmentation-Based Finger Vein Recognition Using Automatically Generated and Fused Training Labels

“…The resulting segmentation masks were used as ground-truth labels together with the corresponding bright-field image data to train a CNN. In [14], Canny edge detection was applied to ultrasound images to generate the ground-truth labels required to train a CNN for segmentation of musculo-skeletal ultrasound images. In [9], a part of the ground-truth labels required to train a CNN for brain tumour segmentation was generated by a voted average of segmentation results of top performing classical segmentation algorithms in this field.…”

Using Fuzzy Inference system for detection the edges of Musculoskeletal Ultrasound Images

“…Moreover, speckle noise was dealt with by the statistical adaptive method and an anisotropic diffusion filter before going on to perform musculoskeletal US image segmentation. Other approaches have attempted edge detection of panoramic musculoskeletal ultrasound images [6] and X-ray images of bone [7] by using Convolutional Neural Networks (CNNs). The CNN technique can be time-consuming because it needs extensive training and large training datasets to get a robust edge detection output image.…”