Automatic detection of lumen and media in the IVUS images using U-Net with VGG16 Encoder

Abstract

Coronary heart disease is one of the top rank leading cause of mortality in the world which can be because of plaque burden inside the arteries. Intravascular Ultrasound (IVUS) has been recognized as powerful imaging technology which captures the real time and high resolution images of the coronary arteries and can be used for the analysis of these plaques. The IVUS segmentation involves the extraction of two arterial walls components namely, lumen and media. In this paper, we investigate the effectiveness of Convolutional Neural Networks including U-Net to segment ultrasound scans of arteries. In particular, the proposed segmentation network was built based on the the U-Net with the VGG16 encoder. Experiments were done for evaluating the proposed segmentation architecture which show promising quantitative and qualitative results.

Optimal surface segmentation with convex priors in irregularly sampled space

“…For the UNET method (Balakrishna et al, 2018), the authors published the performance of their method with respect to Jaccard Metric. It can be seen from the results that based on the Jaccard metric, the proposed method outperforms the UNET method.…”

“…Additionally, we compare the measures obtained from our method to a deep learning method with a UNET architecture (UNET) which was applied on the same dataset and was reported in Ref. (Balakrishna et al, 2018). Overview of each method's feature, including whether the algorithm was applied to lumen and/or media, whether the segmentation was done in 2-D or 3-D and whether the method was semi-automated or fully automated is shown in Table 2.…”

VGG16Unet: effective passive model for image splicing forgery localization

“…After upsampling operations on the decoder side, U-net uses long skip connections to maintain the low-level features and concatenate them to the high-level features. [45][46][47] In the VGG16UNet model, the final three fully connected layers of the VGG16 neural network are not included in the extraction process. There are five steps in the expanding journey.…”

Precise Repositioning of Robotic Ultrasound: Improving Registration-based Motion Compensation using Ultrasound Confidence Optimization

“…To provide good imaging quality and guarantee patient safety during the scanning, a compliant controller is employed to maintain a constant force between the probe and the contact surface [20], [21]. In addition, to monitor the object movement, a neural network (UNet-VGG16 [22]) was trained to segment the arm from RGB images. The segmented results of images acquired at t 1 and t 2 are further used to compute the dice coefficient.…”

“…To effectively detect the potential movement (expected or unexpected) during the scanning, the UNet-VGG16 architecture [22], [27] is employed to segment the arm from RGB images. The U-Net was proposed by Ronneberger [28] for segmentation tasks based on Fully convolutional networks (FCNs).…”