Final Region Of Interest Research Articles

CViTF-Net: A Convolutional and Visual Transformer Fusion Network for Small Ship Target Detection in Synthetic Aperture Radar Images

Detecting small ship targets in large-scale synthetic aperture radar (SAR) images with complex backgrounds is challenging. This difficulty arises due to indistinct visual features and noise interference. To address these issues, we propose a novel two-stage detector, namely a convolutional and visual transformer fusion network (CViTF-Net), and enhance its detection performance through three innovative modules. Firstly, we designed a pyramid structured CViT backbone. This design leverages convolutional blocks to extract low-level and local features, while utilizing transformer blocks to capture inter-object dependencies over larger image regions. As a result, the CViT backbone adeptly integrates local and global information to bolster the feature representation capacity of targets. Subsequently, we proposed the Gaussian prior discrepancy (GPD) assigner. This assigner employs the discrepancy of Gaussian distributions in two dimensions to assess the degree of matching between priors and ground truth values, thus refining the discriminative criteria for positive and negative samples. Lastly, we designed the level synchronized attention mechanism (LSAM). This mechanism simultaneously considers information from multiple layers in region of interest (RoI) feature maps, and adaptively adjusts the weights of diverse regions within the final RoI. As a result, it enhances the capability to capture both target details and contextual information. We achieved the highest comprehensive evaluation results for the public LS-SSDD-v1.0 dataset, with an mAP of 79.7% and an F1 of 80.8%. In addition, the robustness of the CViTF-Net was validated using the public SSDD dataset. Visualization of the experimental results indicated that CViTF-Net can effectively enhance the detection performance for small ship targets in complex scenes.

Automated region of interest placement on cervical ultrasound images for assessing preterm birth risk

Spontaneous preterm birth (sPTB) is one of the leading causes of infant morbidity. Medical interventions can prevent death caused by preterm birth if the risk is predicted at early stages. Quantitative ultrasound (QUS) is found valuable for predicting sPTB risk with a limitation of requiring a medically trained image analyst to manually draw a region of interest (ROI) on the cervix of a B-mode ultrasound image. An automated ROI placement algorithm was designed and trained to reduce the reliance on human annotations. The algorithm utilized a deep neural network with an optimized U-Net architecture to locate cervical tissues. A total of 8670 ultrasound images with sonographer-drawn ROIs were used for algorithm training and testing, followed by several postprocessing steps to yield the final ROIs. Quantitative comparison between algorithm-generated and sonographer-drawn ROIs yielded an average pixel accuracy of 96% and a dice coefficient of 88%. In addition, the QUS’s attenuation coefficient (AC) and backscatter coefficient (BSC) obtained from the algorithm-generated ROIs were highly correlated to those obtained from the sonographer-drawn ROIs with a Pearson correlation coefficient of 0.93 and 0.85, respectively. The results support the feasibility of automating QUS imaging for sPTB risk assessment. [Work supported by NIHR01HD089935.]

Enhanced Region Growing for Brain Tumor MR Image Segmentation.

A brain tumor is one of the foremost reasons for the rise in mortality among children and adults. A brain tumor is a mass of tissue that propagates out of control of the normal forces that regulate growth inside the brain. A brain tumor appears when one type of cell changes from its normal characteristics and grows and multiplies abnormally. The unusual growth of cells within the brain or inside the skull, which can be cancerous or non-cancerous has been the reason for the death of adults in developed countries and children in under developing countries like Ethiopia. The studies have shown that the region growing algorithm initializes the seed point either manually or semi-manually which as a result affects the segmentation result. However, in this paper, we proposed an enhanced region-growing algorithm for the automatic seed point initialization. The proposed approach’s performance was compared with the state-of-the-art deep learning algorithms using the common dataset, BRATS2015. In the proposed approach, we applied a thresholding technique to strip the skull from each input brain image. After the skull is stripped the brain image is divided into 8 blocks. Then, for each block, we computed the mean intensities and from which the five blocks with maximum mean intensities were selected out of the eight blocks. Next, the five maximum mean intensities were used as a seed point for the region growing algorithm separately and obtained five different regions of interest (ROIs) for each skull stripped input brain image. The five ROIs generated using the proposed approach were evaluated using dice similarity score (DSS), intersection over union (IoU), and accuracy (Acc) against the ground truth (GT), and the best region of interest is selected as a final ROI. Finally, the final ROI was compared with different state-of-the-art deep learning algorithms and region-based segmentation algorithms in terms of DSS. Our proposed approach was validated in three different experimental setups. In the first experimental setup where 15 randomly selected brain images were used for testing and achieved a DSS value of 0.89. In the second and third experimental setups, the proposed approach scored a DSS value of 0.90 and 0.80 for 12 randomly selected and 800 brain images respectively. The average DSS value for the three experimental setups was 0.86.

Measuring panoramic radiomorphometric indices for mandible bone using active shape model and Bayesian information criterion‐support vector machine

AbstractThis article proposes an automatic method based on a combination of active shape model (ASM) and nonlinear support vector machine (SVM) accompanied with Bayesian information criterion which is utilized in order to measure radiomorphometric indices for digital X‐ray panoramic system. After omitting Poisson noise of input images, the image is divided. It attempts to choose suitable region of interest (ROI) in order to extract indices. This region becomes small and smaller during some steps until the final ROI is found and the mandible's cortex and ramus are detected. Finally, because of measuring the required parameters, ASM and SVM are used in order to evaluate exact setting of the detected region. Experimental results show no significant differences between right and left sides in terms of mental index (MI), panoramic mandibular index (PMI), gonial angle (GA), gonial index (GI), antegonial index (AGI), ramus height (RH), and ramus width (RW); however, in terms of GI and AGI for both sexes. There are significant differences between men and women for all indices in the mandible (P‐value <.05). There was a mild negative significant correlation among GA and MI, GI, AGI and a moderate negative significant correlation among GA and PMI, RH, RW for both genders (P‐value <.05). Besides, it has been shown measuring the panoramic indices provide useful information about trauma and prediction of asymptomatic subjects with osteoporosis at an earlier step.

Real-Time Vehicle Detection Algorithm Based on Vision and Lidar Point Cloud Fusion

Vehicle detection is one of the most important environment perception tasks for autonomous vehicles. The traditional vision-based vehicle detection methods are not accurate enough especially for small and occluded targets, while the light detection and ranging- (lidar-) based methods are good in detecting obstacles but they are time-consuming and have a low classification rate for different target types. Focusing on these shortcomings to make the full use of the advantages of the depth information of lidar and the obstacle classification ability of vision, this work proposes a real-time vehicle detection algorithm which fuses vision and lidar point cloud information. Firstly, the obstacles are detected by the grid projection method using the lidar point cloud information. Then, the obstacles are mapped to the image to get several separated regions of interest (ROIs). After that, the ROIs are expanded based on the dynamic threshold and merged to generate the final ROI. Finally, a deep learning method named You Only Look Once (YOLO) is applied on the ROI to detect vehicles. The experimental results on the KITTI dataset demonstrate that the proposed algorithm has high detection accuracy and good real-time performance. Compared with the detection method based only on the YOLO deep learning, the mean average precision (mAP) is increased by 17%.

Extraction technique of region of interest from stereoscopic video

A feature fusion approach is presented to extract the region of interest (ROI) from the stereoscopic video. [0]Based on human vision system (HVS), the depth feature, the color feature and the motion feature are chosen as vision features. [0]The algorithm is shown as follows. Firstly, color saliency is calculated on superpixel scale. Color space distribution of the superpixel and the color difference between the superpixel and background pixel are used to describe color saliency and color salient region is detected. Then, the classic visual background extractor (Vibe) algorithm is improved from the update interval and update region of background model. The update interval is adjusted according to the image content. The update region is determined through non-obvious movement region and background point detection. So the motion region of stereoscopic video is extracted using improved Vibe algorithm. The depth salient region is detected by selecting the region with the highest gray value. Finally, three regions are fused into final ROI. Experiment results show that the proposed method can extract ROI from stereoscopic video effectively. In order to further verify the proposed method, stereoscopic video coding application is also carried out on the joint model (JM) encoder with different bit allocation in ROI and the background region.

Multi-Level Segmentation of Infrared Images with Region of Interest Extraction

Infrared (IR) imaging has been researched for various applications such as surveillance. IR radiation has the capability to detect thermal characteristics of objects under low-light conditions. However, automatic segmentation for finding the object of interest would be challenging since the IR detector often provides the low spatial and contrast resolution image without color and texture information. Another hindrance is that the image can be degraded by noise and clutters. This paper proposes multi-level segmentation for extracting regions of interest (ROIs) and objects of interest (OOIs) in the IR scene. Each level of the multi-level segmentation is composed of a k-means clustering algorithm, an expectation-maximization (EM) algorithm, and a decision process. The k-means clustering initializes the parameters of the Gaussian mixture model (GMM), and the EM algorithm estimates those parameters iteratively. During the multi-level segmentation, the area extracted at one level becomes the input to the next level segmentation. Thus, the segmentation is consecutively performed narrowing the area to be processed. The foreground objects are individually extracted from the final ROI windows. In the experiments, the effectiveness of the proposed method is demonstrated using several IR images, in which human subjects are captured at a long distance. The average probability of error is shown to be lower than that obtained from other conventional methods such as Gonzalez, Otsu, k-means, and EM methods.

Automatic detection of regions of interest in breast ultrasound images based on local phase information.

Due to the inherent speckling and low contrast of ultrasonic images, the accurate and efficient location of regions of interest (ROIs) is still a challenging task for breast ultrasound (BUS) computer-aided diagnosis (CAD) systems. In this paper, a fully automatic and efficient ROI generation approach is proposed. First, a BUS image is preprocessed to improve image quality. Second, a phase of max-energy orientation (PMO) image of the preprocessed image is calculated. Otsu's threshold selection method is then used to binarize the preprocessed image, the phase image and the composed image obtained by adding and normalizing the set of two images. Finally, a region selection algorithm is developed to select the true tumor region from these three binary images before generating a final ROI. The method was validated on a BUS database with 168 cases (81 benign and 87 malignant); the accuracy, average precision rate and average recall rate are calculated and compared with conventional method. The results indicate that the proposed method is more accurate and efficient in locating ROIs.

Development of Evidence-Based Region of Interest Matching Guidelines to Standardize Volumetric Image-Guided Radiation Therapy Practice for Sarcoma

Development of Evidence-Based Region of Interest Matching Guidelines to Standardize Volumetric Image-Guided Radiation Therapy Practice for Sarcoma

Development of an automated region of interest selection method for 3D surface monitoring of head motion

To simplify the often complex and user-dependent manual region of interest (ROI) selection process for head motion monitoring, an automatic ROI selection method was developed. The automatic ROI selection algorithm calculated the displacements and velocities of 3D surface points between a temporally correlated 3D image series and a reference image. Only facial surfaces satisfying certain spatial and temporal criteria were selected. The algorithm was tested on five healthy volunteers instructed to perform different types of facial movements for a total of 27 real-time image sets (40-120 images for each image set). The algorithm detected and excluded surface areas affected by different types of local facial movements that were independent of actual net head motion. Eye, eyebrow, and mandible motion were most commonly detected as being independent of head motion and were excluded from the final ROI. For 3D images taken with substantial facial or whole head motion, either most of the facial area was excluded or only small areas with random patterns were included in the final ROI. Surface image registration using iterative closest point (ICP) methods showed more stable real-time head tracking using the automatically selected ROI than manual user defined ROIs. The automatic selection method successfully found ROIs stable over time for tracking head motion by excluding locally varying facial motions. By automating the ROI selection process, it is feasible that the time and complexity of current ROI definition can be reduced, together with user-dependent registration errors.

TH-A-220-01: An Automated Optimal ROI Discovery Method for Real-Time 3D Surface Image Monitoring of Head Motion

Purpose: To increase image registration accuracy, and to remove the time‐ consuming manual region‐of‐interest (ROI) selection process for patient monitoring 3D surface imaging systems (AlignRT,VisionRT,UK), a method was developed to adaptively define an optimal ROI based on discovering topographically stable surfaces over time. Methods: Our method calculates the displacement and velocity of 3D surface points between a series of images captured in real‐time and the reference image. Using a set of selection criteria, the algorithm adaptively constructs a ROI by including only facial surfaces that move rigidly with the target and excludes surfaces that fluctuate locally such as the mandible regions and eyes. Data for five volunteers were acquired over various facial surface and motion types. A total of 27 data sets were analyzed, with 2–7 data sets for each volunteer and each data set contained consecutively captured 40–120 real‐time images. Results: The adaptive ROI selection algorithm correspondingly detected surfaces affected by different types of facial movement independent of target motions. It interpreted the eyes, eyebrows and mandible motions shown in the real‐time 3D images as independent of target motion even when AlignRT misinterpreted them, and excluded them from the final ROI. For 3D images taken with substantial facial motions, either most of the facial areas were excluded or only small areas with random patterns were included in the final ROI. The algorithm excluded the upper side and nose where the view of the cameras was blocked, and thus surface was not well‐ reconstructed in many of real‐time 3D images when couch was rotated greater than 40°. Conclusion: The method successfully located suitable ROI surfaces undergoing minimal displacement and temporal fluctuation. By automating the ROI selection process, the time and complexity of current ROI definition can be reduced, together with user‐dependent registration errors.

Region of interest extraction algorithm of color image based on visual perception information

A Region of Interest(ROI) extraction algorithm of color image is proposed,which is based on the classical seeded region growing method guided by eye-tracking perception information.Based on the analysis of subjects' eye-tracking data,the proposed algorithm can generate the interesting points automatically,whose neighborhood information is then used to obtain a series of initial ROIs by color homogeneity criterion in specific color space.The final ROI is picked up from the initial ROIs according to the viewer's attentive degree.The proposed algorithm is tested in RGB,HSV and L*a*b color space.Experimental results show that the proposed algorithm using in RGB and HSV color space can accord human's subjective perception,and have certain application value.

Optimizing the sampling and staging for simulations of rare events via forward flux sampling schemes

In this work, we present an adaptive algorithm to optimize the phase space sampling for simulations of rare events in complex systems via forward flux sampling (FFS) schemes. In FFS, interfaces are used to partition the phase space along an order parameter lambda connecting the initial and final regions of interest. Since the kinetic "bottleneck" regions along the order parameter are not usually known beforehand, an adaptive procedure is used that first finds these regions by estimating the rate constants associated with reaching subsequent interfaces; thereafter, the FFS simulation is reset to concentrate the sampling on those bottlenecks. The approach can optimize for either the number and position of the interfaces (i.e., optimized lambda phase staging) or the number M of fired trial runs per interface (i.e., the {M(i)} set) to minimize the statistical error in the rate constant estimation per simulation period. For example, the optimization of the lambda staging leads to a net constant flux of partial trajectories between interfaces and hence a constant flux of connected paths throughout the region between the two end states. The method is demonstrated for several test systems, including the folding of a lattice protein. It is shown that the proposed approach leads to an optimized lambda staging and {M(i)} set which increase the computational efficiency of the sampling algorithm.