Segmentation Of Tomography Images Research Articles

Three-Dimensional X-Ray Computed Tomography Image Segmentation and Point Cloud Reconstruction for Internal Defect Identification in Laser Powder Bed Fused Parts

Abstract Defects shape, volume, and orientation all have a direct impact on the mechanical properties of Laser Powder Bed Fused (L-PBF-ed) parts. Therefore, it is necessary to evaluate and analyze the three-dimensional (3D) geometrical characteristics of these defects. X-ray Computed Tomography (XCT) can reveal an object's internal structure by volumetric scanning through its building direction. Point clouds are 3D data that can be extracted from the stack of XCT images taken from a part to perform further analysis. This study presents a novel approach for 3D segmentation and geometrical analysis of L-PBF defect structures from XCT images. The proposed method integrates Voronoi labeling and 3D point cloud reconstruction to reveal individual defect characteristics from the XCT image stack of a part. A case study showed the proposed methodology's effectiveness in identifying and characterizing defect regions in L-PBF-ed Cobalt-Chrome (CoCr) parts.

Segmentation and quantitative analysis of optical coherence tomography (OCT) images of laser burned skin based on deep learning

Evaluation of skin recovery is an important step in the treatment of burns. However, conventional methods only observe the surface of the skin and cannot quantify the injury volume. Optical coherence tomography (OCT) is a non-invasive, non-contact, real-time technique. Swept source OCT uses near infrared light and analyzes the intensity of light echo at different depths to generate images from optical interference signals. To quantify the dynamic recovery of skin burns over time, laser induced skin burns in mice were evaluated using deep learning of Swept source OCT images. A laser-induced mouse skin thermal injury model was established in thirty Kunming mice, and OCT images of normal and burned areas of mouse skin were acquired at day 0, day 1, day 3, day 7, and day 14 after laser irradiation. This resulted in 7000 normal and 1400 burn B-scan images which were divided into training, validation, and test sets at 8:1.5:0.5 ratio for the normal data and 8:1:1 for the burn data. Normal images were manually annotated, and the deep learning U-Net model (verified with PSPNe and HRNet models) was used to segment the skin into three layers: the dermal epidermal layer, subcutaneous fat layer, and muscle layer. For the burn images, the models were trained to segment just the damaged area. Three-dimensional reconstruction technology was then used to reconstruct the damaged tissue and calculate the damaged tissue volume. The average IoU value and f-score of the normal tissue layer U-Net segmentation model were 0.876 and 0.934 respectively. The IoU value of the burn area segmentation model reached 0.907 and f-score value reached 0.951. Compared with manual labeling, the U-Net model was faster with higher accuracy for skin stratification. OCT and U-Net segmentation can provide rapid and accurate analysis of tissue changes and clinical guidance in the treatment of burns.

Accuracy Analysis of 3D Bone Fracture Models: Effects of Computed Tomography (CT) Imaging and Image Segmentation

The introduction of three-dimensional (3D) printed anatomical models has garnered interest in pre-operative planning, especially in orthopedic and trauma surgery. Identifying potential error sources and quantifying their effect on the model dimensional accuracy are crucial for the applicability and reliability of such models. In this study, twenty radii were extracted from anatomic forearm specimens and subjected to osteotomy to simulate a defined fracture of the distal radius (Colles’ fracture). Various factors, including two different computed tomography (CT) technologies (energy-integrating detector (EID) and photon-counting detector (PCD)), four different CT scanners, two scan protocols (i.e., routine and high dosage), two different scan orientations, as well as two segmentation algorithms were considered to determine their effect on 3D model accuracy. Ground truth was established using 3D reconstructions of surface scans of the physical specimens. Results indicated that all investigated variables significantly impacted the 3D model accuracy (p < 0.001). However, the mean absolute deviation fell within the range of 0.03 ± 0.20 to 0.32 ± 0.23 mm, well below the 0.5 mm threshold necessary for pre-operative planning. Intra- and inter-operator variability demonstrated fair to excellent agreement for 3D model accuracy, with an intra-class correlation (ICC) of 0.43 to 0.92. This systematic investigation displayed dimensional deviations in the magnitude of sub-voxel imaging resolution for all variables. Major pitfalls included missed or overestimated bone regions during the segmentation process, necessitating additional manual editing of 3D models. In conclusion, this study demonstrates that 3D bone fracture models can be obtained with clinical routine scanners and scan protocols, utilizing a simple global segmentation threshold, thereby providing an accurate and reliable tool for pre-operative planning.

Rapid measurement of epidermal thickness in OCT images of skin

Epidermal thickness (ET) changes are associated with several skin diseases. To measure ET, segmentation of optical coherence tomography (OCT) images is essential; manual segmentation is very time-consuming and requires training and some understanding of how to interpret OCT images. Fast results are important in order to analyze ET over different regions of skin in rapid succession to complete a clinical examination and enable the physician to discuss results with the patient in real time. The well-known CNN-graph search (CNN-GS) methodology delivers highly accurate results, but at a high computational cost. Our objective was to build a computational core, based on CNN-GS, able to accurately segment OCT skin images in real time. We accomplished this by fine-tuning the hyperparameters, testing a range of speed-up algorithms including pruning and quantization, designing a novel pixel-skipping process, and implementing the final product with efficient use of core and threads on a multicore central processing unit (CPU). We name this product CNN-GS-skin. The method identifies two defined boundaries on OCT skin images in order to measure ET. We applied CNN-GS-skin to OCT skin images, taken from various body sites of 63 healthy individuals. Compared with CNN-GS, our described method reduced computation time by 130 ×\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\ imes$$\\end{document} with minimal reduction in ET determination accuracy (from 96.38 to 94.67%).

Rapid automated lumen segmentation of coronary optical coherence tomography images followed by 3D reconstruction of coronary arteries.

Optical coherence tomography has emerged as an important intracoronary imaging technique for coronary artery disease diagnosis as it produces high-resolution cross-sectional images of luminal and plaque morphology. Precise and fast lumen segmentation is essential for efficient OCT morphometric analysis. However, due to the presence of various image artifacts, including side branches, luminal blood artifacts, and complicated lesions, this remains a challenging task. Our research study proposes a rapid automatic segmentation method that utilizes nonuniform rational B-spline to connect limited pixel points and identify the edges of the OCT lumen. The proposed method suppresses image noise and accurately extracts the lumen border with a high correlation to ground truth images based on the area, minimal diameter, and maximal diameter. We evaluated the method using 3300 OCT frames from 10 patients and found that it achieved favorable results. The average time taken for automatic segmentation by the proposed method is 0.17s per frame. Additionally, the proposed method includes seamless vessel reconstruction following the lumen segmentation. The developed automated system provides an accurate, efficient, robust, and user-friendly platform for coronary lumen segmentation and reconstruction, which can pave the way for improved assessment of the coronary artery lumen morphology.

Data-driven approaches for electrical impedance tomography image segmentation from partial boundary data

Data-driven approaches for electrical impedance tomography image segmentation from partial boundary data

Improved heuristics for solving large-scale Scanning Transmission Electron Microscopy image segmentation using the ordered median problem

The discrete ordered median problem can be applied in a wide variety of areas. The application of this problem in electron tomography image segmentation is currently being considered since high-quality images are obtained when this model is applied. However, its application means that a high computing time is required to obtain solutions for large-scale instances. The size of the images is of great importance in electron tomography experiments, since the larger the image size, the higher the quality of the image. With the goal of reducing the computation times, this paper introduces different heuristic procedures to obtain feasible solutions for the ordered median problem that provide high-quality images in low computing times. Moreover, some noticeable improvements for the heuristic techniques are developed, taking advantage of the particular versions of the ordered median function that have been proven to be especially suitable for electron tomography image segmentation.

BOA: A CT-Based Body and Organ Analysis for Radiologists at the Point of Care.

The study aimed to develop the open-source body and organ analysis (BOA), a comprehensive computed tomography (CT) image segmentation algorithm with a focus on workflow integration. The BOA combines 2 segmentation algorithms: body composition analysis (BCA) and TotalSegmentator. The BCA was trained with the nnU-Net framework using a dataset including 300 CT examinations. The CTs were manually annotated with 11 semantic body regions: subcutaneous tissue, muscle, bone, abdominal cavity, thoracic cavity, glands, mediastinum, pericardium, breast implant, brain, and spinal cord. The models were trained using 5-fold cross-validation, and at inference time, an ensemble was used. Afterward, the segmentation efficiency was evaluated on a separate test set comprising 60 CT scans. In a postprocessing step, a tissue segmentation (muscle, subcutaneous adipose tissue, visceral adipose tissue, intermuscular adipose tissue, epicardial adipose tissue, and paracardial adipose tissue) is created by subclassifying the body regions. The BOA combines this algorithm and the open-source segmentation software TotalSegmentator to have an all-in-one comprehensive selection of segmentations. In addition, it integrates into clinical workflows as a DICOM node-triggered service using the open-source Orthanc research PACS (Picture Archiving and Communication System) server to make the automated segmentation algorithms available to clinicians. The BCA model's performance was evaluated using the Sørensen-Dice score. Finally, the segmentations from the 3 different tools (BCA, TotalSegmentator, and BOA) were compared by assessing the overall percentage of the segmented human body on a separate cohort of 150 whole-body CT scans. The results showed that the BCA outperformed the previous publication, achieving a higher Sørensen-Dice score for the previously existing classes, including subcutaneous tissue (0.971 vs 0.962), muscle (0.959 vs 0.933), abdominal cavity (0.983 vs 0.973), thoracic cavity (0.982 vs 0.965), bone (0.961 vs 0.942), and an overall good segmentation efficiency for newly introduced classes: brain (0.985), breast implant (0.943), glands (0.766), mediastinum (0.880), pericardium (0.964), and spinal cord (0.896). All in all, it achieved a 0.935 average Sørensen-Dice score, which is comparable to the one of the TotalSegmentator (0.94). The TotalSegmentator had a mean voxel body coverage of 31% ± 6%, whereas BCA had a coverage of 75% ± 6% and BOA achieved 93% ± 2%. The open-source BOA merges different segmentation algorithms with a focus on workflow integration through DICOM node integration, offering a comprehensive body segmentation in CT images with a high coverage of the body volume.

Segmentation of two-phase flow X-ray tomography images to determine contact angle using deep autoencoders

This study improves the characterization of in situ contact angles in porous media by employing deep learning techniques (SegNet, UNet, ResNet, and UResNet) for multiphase segmentation of micro-CT images. The algorithms were tested on high-resolution X-ray images of a steady-state flow experiment where two fluid phases were simultaneously injected at different fractional flows. The models were trained to segment the images into solid, aqueous phase, and non-aqueous phase liquid (NAPL). The UResNet demonstrated the best performance with an f1-score of 0.966 for the test dataset. More importantly, the UResNet offered higher reliability than the watershed algorithm for various fractional flows based on visual inspection and phase distribution analysis. The porosity calculation error of the watershed method (7.8 %) was reduced to 5.1 % by UResNet. Furthermore, UResNet accurately depicted the consistently mixed-wet condition of the rock sample throughout the experiment, in contrast to the watershed segmentation that yielded inconsistencies in contact angle calculations at an aqueous phase fractional flow of 0.01.

Human selection bias drives the linear nature of the more ground truth effect in explainable deep learning optical coherence tomography image segmentation.

Supervised deep learning (DL) algorithms are highly dependent on training data for which human graders are assigned, for example, for optical coherence tomography (OCT) image annotation. Despite the tremendous success of DL, due to human judgment, these ground truth labels can be inaccurate and/or ambiguous and cause a human selection bias. We therefore investigated the impact of the size of the ground truth and variable numbers of graders on the predictive performance of the same DL architecture and repeated each experiment three times. The largest training dataset delivered a prediction performance close to that of human experts. All DL systems utilized were highly consistent. Nevertheless, the DL under-performers could not achieve any further autonomous improvement even after repeated training. Furthermore, a quantifiable linear relationship between ground truth ambiguity and the beneficial effect of having a larger amount of ground truth data was detected and marked as the more-ground-truth effect.

Radiographic, tomography and three-dimensional description of the clinical anatomy of the long bones of Rupornis Magnirostris

Birds of prey play an important role in the ecological balance of several biomes, but their proximity to urban areas often causes harm to the animal. The investigation of clinical surgical approaches requires a precise understanding of the anatomical system, especially when applied to imaging exams of wild animals. In this article we present the clinical anatomy of the long bones of Rupornis magnirostris, roadside hawk, focusing on its express form in the imaging tests adopted. Cadavers were submitted to digital radiography and computed tomography, providing models of both two-dimensional images and three-dimensional images of long bones through the segmentation of tomographic images, both 2D and 3D data were contrasted by anatomical dissection. In this article we obtain: (1) A summary of the nomenclature used in the literature of the long bones of the birds perceived in the imaging examinations. The approach presented here provides considerable advantages for the interpretation and communication of anatomical information and its visualization; (2) the 3D model produced from the segmentation process as an interactive model in the form of an embedded 3D image, which can be viewed at any angle and size, allowing the structure to be evidenced as necessary to provide an understanding of the normal anatomy of long bones of this species.

Evaluation of the accuracy of cone-beam computed tomography image segmentation of isolated tooth roots based on the dynamic threshold method

ObjectiveAccurate quantification of the root surface area (RSA) plays a decisive role in the advancement of periodontal, orthodontic, and restorative treatment modalities. In this study, we aimed to develop a dynamic threshold-based computer-aided system for segmentation and calculation of the RSA of isolated teeth on cone-beam computed tomography (CBCT) and to assess the accuracy of the measured data.MethodWe selected 24 teeth to be extracted, including single-rooted and multi-rooted teeth, from 22 patients who required tooth extraction. In the experimental group, we scanned 24 isolated teeth using CBCT with a voxel size of 0.3 mm. We designed a computer-aided system based on a personalized dynamic threshold algorithm to automatically segment the roots of 24 isolated teeth in CBCT images and calculate the RSA. In the control group, we employed digital intraoral scanner devices to perform optical scanning on 24 isolated teeth and subsequently manually segmented the roots using 3-matic software to calculate the RSA. We used the paired t-test (P < 0.05) and Bland-Altman plots to analyze the consistency of the two measurement methods.ResultsThe results of the paired t-test showed that there was no significant difference in the RSAs obtained using the dynamic threshold method and the optical scanning image reconstruction (t = 1.005, P = 0.325 > 0.05). As per the Bland-Altman plot, the results were evenly distributed within the region of ± 1.96 standard deviations of the mean, with no increasing or decreasing trends and good consistency.ConclusionIn this study, we designed a computer-aided root segmentation system based on a personalized dynamic threshold algorithm to automatically segment the roots of isolated teeth in CBCT images with a voxel size of 0.3 mm. We found that the RSA calculated using this approach was highly accurate, and a voxel of 0.3 mm in size could accurately display the surface area data in CBCT images. Overall, our findings in this study provide a foundation for future work on accurate automatic segmentation of tooth roots in full-mouth CBCT images and the computation of RSA.

Liver Tumor Computed Tomography Image Segmentation Based on an Improved U-Net Model

An automated segmentation method for computed tomography (CT) images of liver tumors is an urgent clinical need. Tumor areas within liver cancer images are easily missed as they are small and have unclear borders. To address these issues, an improved liver tumor segmentation method based on U-Net is proposed. This involves incorporating attention mechanisms into the U-Net’s skip connections, giving higher weights to important regions. Through dynamically adjusting the attention recognition characteristics, the method achieves accurate localization that is focused on and discriminates target regions. Testing using the LiTS (liver tumor segmentation) public dataset resulted in a Dice similarity coefficient of 0.69. The experiments demonstrated that this method can accurately segment liver tumors.

Computed Tomography Image Segmentation of Lung Corona Virus Infection Region Based on Combination of Grayscale Morphological Reconstruction and Fast Marching Method

Recently, X-ray computed tomography-imaging modality is considered as golden standard for diagnosis of coronavirus lungs infection. In worldwide, infectious patients increase rapidly that lead to weariness in health services staff, as well as instant treatment required to avoid patients’ health deterioration due to infection development. Image processing would be reinforcing health services by considering computer-based segmentation. However, a ground glass computed tomography image fashion of coronavirus lungs infection characterized by disappearance of edge region of interest and lack of object structure. In this study, these challenges addressed by introducing a new algorithm that combined both morphological reconstruction and fast marching method. The proposed algorithm applied on archived computed tomography dataset for coronavirus infected patients, results showed consistent determination of ground glass infection region compared to manual delineation of senior physician. The proposed algorithm restricted to empirical adjustment of FMM’s threshold that would be addressed in upcoming study.

Automated Segmentation of Optical Coherence Tomography Images of the Human Tympanic Membrane Using Deep Learning.

Optical Coherence Tomography (OCT) is a light-based imaging modality that is used widely in the diagnosis and management of eye disease, and it is starting to become used to evaluate for ear disease. However, manual image analysis to interpret the anatomical and pathological findings in the images it provides is complicated and time-consuming. To streamline data analysis and image processing, we applied a machine learning algorithm to identify and segment the key anatomical structure of interest for medical diagnostics, the tympanic membrane. Using 3D volumes of the human tympanic membrane, we used thresholding and contour finding to locate a series of objects. We then applied TensorFlow deep learning algorithms to identify the tympanic membrane within the objects using a convolutional neural network. Finally, we reconstructed the 3D volume to selectively display the tympanic membrane. The algorithm was able to correctly identify the tympanic membrane properly with an accuracy of ~98% while removing most of the artifacts within the images, caused by reflections and signal saturations. Thus, the algorithm significantly improved visualization of the tympanic membrane, which was our primary objective. Machine learning approaches, such as this one, will be critical to allowing OCT medical imaging to become a convenient and viable diagnostic tool within the field of otolaryngology.

Porosity evolution in proton irradiated microfine-grained POCO graphite

This work investigated the porosity evolution of POCO ZXF-5Q graphite that has been irradiated by 340 kW, 120 GeV protons inside NT02 target system in Fermilab's NuMI beamline. This POCO graphite has undergone direct bulk dimensional swelling at low dose irradiation and its local microstructural change is still not well-understood during this process. In this work, the (sub-) micrometre scale porosity from six locations across proton beam fluence and temperature gradients have been studied using focused ion beam-scanning electron microscopy (FIB-SEM) tomography. A deep learning-based tomographic image segmentation technique has been established and implemented for porosity segmentation and quantification. It has been found that there is a decrease in the total volumetric percentage of the porosity at proton beam centre (∼ 8 – 8.4 vol.%), by comparing to un-irradiated POCO (∼ 12 – 13 vol.%) and to beam 2σ and 5σ radii (∼ 12 vol.%). This decrease in porosity volume percentage was found to be caused by the reduction in pores with volumes > 0.1 μm3 induced by material bulk dimensional swelling at proton beam centre area. The porosity reduction in relation to dimensional change and irradiation creep was discussed among with other contributing factors, and further investigations through well-controlled irradiation experiment are still needed.

Unraveling the complexity of Optical Coherence Tomography image segmentation using machine and deep learning techniques: A review.

Optical Coherence Tomography (OCT) is an emerging technology that provides three-dimensional images of the microanatomy of biological tissue in-vivo and at micrometer-scale resolution. OCT imaging has been widely used to diagnose and manage various medical diseases, such as macular degeneration, glaucoma, and coronary artery disease. Despite its wide range of applications, the segmentation of OCT images remains difficult due to the complexity of tissue structures and the presence of artifacts. In recent years, different approaches have been used for OCT image segmentation, such as intensity-based, region-based, and deep learning-based methods. This paper reviews the major advances in state-of-the-art OCT image segmentation techniques. It provides an overview of the advantages and limitations of each method and presents the most relevant research works related to OCT image segmentation. It also provides an overview of existing datasets and discusses potential clinical applications. Additionally, this review gives an in-depth analysis of machine learning and deep learning approaches for OCT image segmentation. It outlines challenges and opportunities for further research in this field.

FDTR: Weakening feature disparity transformer for accurate multicategory computed tomography image segmentation

Accurately segmenting object structures in computed tomography (CT) images is crucial for computer-aided surgery, diagnosis, and other interdisciplinary applications. Most state-of-the-art CT segmentation methods employ a classical jump structure to integrate shallow and deep feature information. However, the effectiveness of the current CT image segmentation and related subtasks must still be improved due to feature disparity between different layers and information transfer loss. To alleviate the problem, we propose a novel three-dimensional multicategory segmentation model, the weakening feature disparity transformer (FDTR), based on the transformer structure for CT imaging. Firstly, to effectively capture global features, we design a vision transformer-based encoder. Secondly, to enhance the model’s information representation in a lightweight manner, we devise a nested structure of dense compression connections. Lastly, to mitigate the disparity in semantic features at shallow layers, we propose supervised semantic signals. By employing the vision transformer-based encoder and the dense connection structure, we enhance the detailed information of deep features. Additionally, the semantic supervisory signal introduces deep semantic information to the shallow features, enriching the overall feature representation. We conducted extensive experiments on the KIPA2022 dataset for multiorgan segmentation and the YC2022 dataset for volumetric trait segmentation in broiler breeding. The proposed method, FDTR, effectively enhances feature transfer between shallow and deep features, achieving optimal results compared to other advanced models and demonstrating broad prospects for application.

An active learning method based on result quality evaluation for printed circuit board computed tomography image segmentation

AbstractElement detection is a key step in non‐destructive testing of printed circuit board (PCB) based on computed tomography (CT). In recent years, some image segmentation methods based on deep learning have shown great potential in the element segmentation task of PCB CT images, and greatly improved the efficiency and accuracy. However, since image segmentation is based on pixel‐level classification, the annotation of training data is difficult and costly. Aiming at this problem, the authors proposed a new active learning method based on the integration of relevant information about segmentation tasks and data variance. In this method, the Result Quality Evaluation Module (RQEM) proposed by us is used to generate task‐related information, and an adversarial network is used to generate the difference information between samples and the initial labelled data. Finally, the two parts of information are fused and used as the standard of data selecting. In the PCB CT image element segmentation task, the authors only need to select 12.7% of the whole training set with their proposed method to make Mean Intersection Over Union (MIOU) reach 79.7, which has reached 95% of the optimal performance of 83.7. The in‐depth analysis also verifies the effectiveness and stability of the authors’ method.

Computed tomography image segmentation of irregular cerebral hemorrhage lesions based on improved U-Net

ObjectiveThis paper aims to improve U-Net for more accurate segmentation of irregular intracranial hemorrhage lesions in CT images. MethodsThe residual octave convolution (ResOctConv) module was introduced to overcome the semantic gap issue in U-Net, and a hybrid attention mechanism called mixed attention mechanism (MAM) was proposed to further enhance the performance of U-Net. 40 patients with irregular cerebral hemorrhage images were selected from head CT scans performed between August and December 2022. Two radiologists independently traced the edge of each selected image three times, and the final segmentation boundary was determined by consensus. The effectiveness of the lesion segmentation was measured using the Dice coefficient, Jaccard Index, and Relative volume difference. ResultsBased on the box plot of the Dice coefficient for all 40 patients, the improved U-Net demonstrated higher accuracy in segmenting irregular intracranial hemorrhage lesions in CT images compared to the original U-Net. Moreover, the comparison of Dice coefficient, Jaccard coefficient, and RVD indicates that the improved U-Net outperforms both the original U-Net and the region growing algorithm in segmenting irregular cerebral hemorrhage lesions. ConclusionsThe proposed improved U-Net outperforms both the original U-Net and the region growing algorithm in segmenting irregular cerebral hemorrhage lesions, providing an advanced toolset for radiologists to accurately identify and diagnose irregular cerebral hemorrhage lesions.