An integrated method for atherosclerotic carotid plaque segmentation in ultrasound image

Pixel Classification Based on Local Gray Level Rectangle Window Sampling for Amniotic Fluid Segmentation

“…There are several limitations mentioned in the research, including this method is not fully automatic because it requires manual labels in one IVUS training image and need to pre create training database that is large enough to account for all the different appearances of IVUS images Several studies were carried out to determine the use of window size in sampling models for pixelwise classification. For instance, Qian et al [8] utilized a window size of 9 × 9 to extract gray level value features as a sampling model. Auto-context iterative algorithm were employed to effectively integrate features from ultrasound images and later also the iteratively estimated and refined probability maps together with pixel-wise classification methods.…”

“…The proposed model performance in class specified evaluation had a 78.16% accuracy and 54.6% IoU for the amniotic fluid class. In addition, studies carried out on pixel classification model approaches by Qian et al [8], Pazinato et al [9], and Rosati et al [10] for atherosclerotic carotid plaque, ultrasound tissue images and carotid wall segmentation on US images, showed relatively satisfying results.…”

Semantic segmentation with DenseNets for carotid artery ultrasound plaque segmentation and CIMT estimation

“…Basic techniques for CIM region delineation and plaque segmentation presented in the literature include, among others, Hough transform, edge detection [2,3], active contours [4], snakes [5,6], and other solutions such as integrated approaches that combine several basic Machine Learning (ML) methods [7,8].…”

“…In contrast, fully automatic methods [4,2,7,8,11,12] run without any initial setting, or user interaction. The main advantage of these techniques is that they are able to process large amounts of data.…”

“…Preliminary efforts using ML and Deep Learning (DL) in fully automatic CIMT evaluation have been presented [11,8,12,7]. In [11], a standard multilayer perceptron with an auto-encoder is proposed for CA image interpretation, but it does not outperform the snake-based method in [4].…”