2D Image Slices Research Articles

Detection and quantification of pore, solid and gravel spaces in CT images of a 3D soil sample

In this study, we scanned the core of a cylindrical soil sample (60mm diameter and 100mm height) by X-ray Computed Tomography (CT) producing 300 consecutive 2D digital images with 16-bit gray level depth and a resolution of 32 microns (image size 676 × 676 pixels). The aim of this work was to determine the geometry and spatial distribution of the elements in a sample, related in this case to pore, solid and gravel, inside each 2D image for the latter reconstruction of the corresponding 3D approximation of the elements using the total set of 300 soil images. Therefore, it was possible to determine the relative percentage of each element present in each 2D image and, correspondingly, the structure and total percentage in the 3D reconstruction. The identification of elements in the 2D image slices was very well accomplished using three standard segmentation algorithms: k-Means, Fuzzy c-Means and Otsu multilevel. In order to compare and evaluate the quality of results, a non-uniformity (NU) measure was applied such that low values were indicative of homogeneous regions. Due to the depth of the greyscale of the images, the results were very similar with comparable statistics and homogeneity (NU values) among the detected materials of the three algorithms. That suggests that the pore, solid and gravel spaces were very well identified, and this is reflected through their connectivity in the 3D reconstruction. Additionally, the gray level depth was reduced to 8 bits and the same study was undertaken. In this case, the quality of results was comparable to the previous ones, as the number of elements and NU values were very close. However, this also depends largely on the high resolution of the images. Thereby, the soil sample of this work was very well characterized using the simplest and most common algorithms for image segmentation thanks to the high contrast and resolution, and regardless the depth of the grey-level.

Automated glioma grading on conventional MRI images using deep convolutional neural networks.

Gliomas are the most common primary tumor of the brain and are classified into grades I-IV of the World Health Organization (WHO), based on their invasively histological appearance. Gliomas grading plays an important role to determine the treatment plan and prognosis prediction. In this study we propose two novel methods for automatic, non-invasively distinguishing low-grade (Grades II and III) glioma (LGG) and high-grade (grade IV) glioma (HGG) on conventional MRI images by using deep convolutional neural networks (CNNs). All MRI images have been preprocessed first by rigid image registration and intensity inhomogeneity correction. Both proposed methods consist of two steps: (a) three-dimensional (3D) brain tumor segmentation based on a modification of the popular U-Net model; (b) tumor classification on segmented brain tumor. In the first method, the slice with largest area of tumor is determined and the state-of-the-art mask R-CNN model is employed for tumor grading. To improve the performance of the grading model, a two-dimensional (2D) data augmentation has been implemented to increase both the amount and the diversity of the training images. In the second method, denoted as 3DConvNet, a 3D volumetric CNNs is applied directly on bounding image regions of segmented tumor for classification, which can fully leverage the 3D spatial contextual information of volumetric image data. The proposed schemes were evaluated on The Cancer Imaging Archive (TCIA) low grade glioma (LGG) data, and the Multimodal Brain Tumor Image Segmentation (BraTS) Benchmark 2018 training datasets with fivefold cross validation. All data are divided into training, validation, and test sets. Based on biopsy-proven ground truth, the performance metrics of sensitivity, specificity, and accuracy are measured on the test sets. The results are 0.935 (sensitivity), 0.972 (specificity), and 0.963 (accuracy) for the 2D Mask R-CNN based method, and 0.947 (sensitivity), 0.968 (specificity), and 0.971 (accuracy) for the 3DConvNet method, respectively. In regard to efficiency, for 3D brain tumor segmentation, the program takes around ten and a half hours for training with 300 epochs on BraTS 2018 dataset and takes only around 50s for testing of a typical image with a size of 160×216×176. For 2D Mask R-CNN based tumor grading, the program takes around 4h for training with around 60000 iterations, and around 1s for testing of a 2D slice image with size of 128×128. For 3DConvNet based tumor grading, the program takes around 2h for training with 10000 iterations, and 0.25s for testing of a 3D cropped image with size of 64×64×64, using a DELL PRECISION Tower T7910, with two NVIDIA Titan Xp GPUs. Two effective glioma grading methods on conventional MRI images using deep convolutional neural networks have been developed. Our methods are fully automated without manual specification of region-of-interests and selection of slices for model training, which are common in traditional machine learning based brain tumor grading methods. This methodology may play a crucial role in selecting effective treatment options and survival predictions without the need for surgical biopsy.

3D tomographic magnetofluorescence imaging of nanodiamonds.

We demonstrate lensless imaging of three-dimensional phantoms of fluorescent nanodiamonds in solution. Magnetofluorescence imaging is employed, which relies on a dependence of the fluorescence yield on the magnetic field, and pervading the object with an inhomogeneous magnetic field. This field provides a field-free field line, which is rastered through the object. A 3D image of the object is obtained by imaging a set of 2D slices. Each 2D slice image is computed from a set of 1D projections, obtained under different projection directions, using a backprojection algorithm. Reconstructed images containing up to 36 × 36 × 8 voxels are obtained. A spatial resolution better than 2 mm is achieved in three dimensions. The approach has the potential for scalability.

DMCNN: A Deep Multiscale Convolutional Neural Network Model for Medical Image Segmentation

Medical image segmentation is one of the hot issues in the related area of image processing. Precise segmentation for medical images is a vital guarantee for follow-up treatment. At present, however, low gray contrast and blurred tissue boundaries are common in medical images, and the segmentation accuracy of medical images cannot be effectively improved. Especially, deep learning methods need more training samples, which lead to time-consuming process. Therefore, we propose a novelty model for medical image segmentation based on deep multiscale convolutional neural network (CNN) in this article. First, we extract the region of interest from the raw medical images. Then, data augmentation is operated to acquire more training datasets. Our proposed method contains three models: encoder, U-net, and decoder. Encoder is mainly responsible for feature extraction of 2D image slice. The U-net cascades the features of each block of the encoder with those obtained by deconvolution in the decoder under different scales. The decoding is mainly responsible for the upsampling of the feature graph after feature extraction of each group. Simulation results show that the new method can boost the segmentation accuracy. And, it has strong robustness compared with other segmentation methods.

X-ray computed tomography to investigate industrial cast Al-alloys

This paper describes the use of X-ray computed tomography (XCT) to inspect the internal defects of industrial Aluminum alloys die-casting, to justify the adequateness of super sealant resins in filling tiny holes, and to reduce the casting porosity of metals by the impregnation process. High-resolution 2D slice images were generated by a microfocus X-ray CT system with direct conversion Cadmium Telluride (CdTe) flat panel detector (FPD). Moreover, we were able to visualize the internal defects of the two different shapes of casting objects. The gray-value contrasts of the CT images were excellent to distinguish the resin material in the alloy samples after impregnation treatment. Furthermore, to determine the dimensions of internal defects, virtual sectioning was carried out to view the cross-section of the metal samples and dimensions of internal defects. Dual-energy X-ray computed tomography (DXCT) was used with an energy differentiation type 64-pixel linear array photon-counting line sensor for the material characterization. The impregnated resin part inside the castings defects was verified based on the effective atomic number with a 2.92% low error margin. Hence, this will be a useful qualitative and quantitative advancement to rapid and detailed non-destructive analysis within the die-casting industries in improving the quality of the impregnation process.

Development of modified micro computed tomography system for nondestructive testing

This paper presents a development of micro Computed Tomography (microCT) imaging system based on digital radiography. The system comprises stepper motor, CMOS detector, microcontroller and digital radiography unit. Scanning process was done by rotating the object while exposing its cross-sectional surface to X-ray radiation. The object was rotated by using stepper motor controlled by microcontroller. The normal of cross-sectional surface taken as the rotation axis. Therefore, a number of 2D image slices were obtained by rotating the object from 0° to 360°. The 2D image slices were then reconstructed to form 3D image using computer. By using the proposed microCT, analysis of surface and internal of the object can be done in 3D image and 2D image (slice) for nondestructive testing (NDT) application.

Structural analysis on cutting notch of tea stalk by X-ray micro-computed tomography

Quantitative analysis on cutting notch of tea stalk is one method to better understand shear resistance and mechanical properties of the stalk, which could improve cutting performance and efficiency for tea harvesting. Optimal design of the cutter might be achieved through structural analysis of cutting notch by X-ray micro-computed tomography (micro-CT). The third inter-node of the stalks were sampled and scanned by micro-CT after they were cut at different cutting depths (0.7, 1.5, 2.3 mm) and cutting edge angles of cutter (30°, 35°, 40°). The structural change of the cutting notch and cutting force was observed. Grey-scale histogram analysis and bimodal segmentation method were used to extract the information about cutting notch from 2D slice images. A 3D model of the stalk was established with reconstruction and volume rendering. Besides, maximum cross-sectional area ratio of cutting notch (MCSARCN) and volume ratio of cutting notch (VRCN) were calculated. The results show that MCSARCN and VRCN had a closely relationship with the cutting force, cutting depth and cutting edge angle according to 3D tomographic images and 2D slice images. Both MCSARCN and VRCN increased with cutting depth, despite of cutting edge angle. For cutting edge angle of 30°, 35° and 40° at different cutting depths, MCSARCN increased from 4.89% to 9.47%, from 8.51% to 22.83% and from 4.30% to 16.87%, respectively. Meanwhile, VRCN increased from 1.59% to 2.13%, from 2.98% to 5.76% and from 3.04% to 5.01%, respectively. When cutting depth was 1.5 mm, cutting force increased but VRCN present a decreased trend with cutting edge angle. X-ray micro-CT could be used to analyze the structure on cutting notch of tea stalks which is helpful in optimizing the structural parameters of the cutter.

Slice Localization for Three-Dimensional Breast Ultrasound Volume Using Deep Learning

In routine ultrasonography, slice images of the internal organs of the human body are usually generated through an 1D array probe. The position and orientation of the probe is adjusted manually to obtain slice planes with pathological features. This is quite dependent on the experience and technique of a sonographer. This paper aims to locate 2D slice planes in a 3D breast ultrasound volume, which has significant application value in clinical ultrasound examinations. We propose a deep learning approach mapping all possible 2D image slices to their 3D coordinates parameters using a fully connected neural network implemented on MATLAB. We emphasize that this training must be done separately for each patient since the mammary tissue structure varies greatly from one person to another. The trained network can be interpreted as an image-slice location database for each patient. Our study is validated on GE ABUS (Automated Breast Ultrasound System) volume data. Each 2D image slice has four spatial parameters. The method achieves a prediction error of 0.14mm/0.25mm and 0.5 degree/0.3 degree for translation (x/y) and rotation (yaw/roll) parameters respectively, averaged over all practically scannable slices. It takes less than 0.1ms to predict the location of one 64×64 slice image. Thus, slice locations may be displayed with high accuracy in real-time when scanning with a conventional 1D probe, potentially allowing physicians to manipulate the probe to any scan planes of interest.

Methodology of three-dimensional visualization and quantitative characterization of nanopores in coal by using FIB-SEM and its application with anthracite in Qinshui basin

Analysis of the three-dimensional (3D) visualization and reconstruction methods of nanopores, and its morphology and connectivity are the key to explore the storage and migration mechanism of coalbed methane (CBM). The Bofang (BF) sample collected from the Qinshui basin was scanned by the focused ion beam scanning electron microscopy (FIB-SEM), which can simultaneously image the nanopores through 2D slice image and 3D reconstruction image. Firstly, the 2D morphology and development characteristics of nanopores were analyzed. Secondly, the 3D visualization and reconstruction of nanopores were carried out. Then, the pore connectivity, permeability, and geometric characteristics were discussed. Finally, the effect of pore structure on the storage and migration of CBM was emphatically analyzed. The results show that the organic pores, inorganic pores, shrinkage-induced pores and micro-fractures were developed in coal, and their morphology and connectivity were significant differences. The pores and throats can be extracted from the digital cores with sub-volume of 4.9 × 4.9 × 4.5 μm3, the pore number and throat length (avg.) of BF sample are 21951 and 18.2789 nm, respectively, the porosity and permeability of BF sample is 2.004% and 4.405 × 10−3 mD, respectively, which all indicates the BF reservoir has good storage and connectivity capacity. For pores <50 nm, the number, volume and area indicate that the BF sample has good storage capacity. For pores >200 nm, the pores are all interconnected, which indicates the BF sample can provide effective channels for CBM migration. The distribution of pores and throats and their geometric and topological structures indicate that the BF sample has good migration ability. This study on the methodology of 3D visualization and quantitative characterization of nanopores using FIB-SEM can broaden the research method of pores, and lay good foundations for studying the storage and migration of CBM.

An improved simulated annealing algorithm for reconstructing 3D large-scale porous media

Based on the morphological information obtained from 2D slice images of real porous media, an improved simulated annealing algorithm (SAA) was proposed to reconstruct 3D large-scale porous media, which are intractable to handle for conventional SAA. Three different statistical functions were introduced to characterize the morphological information of real sandstone, including the one-point probability function, the two-point probability function and the lineal-path function. By changing the update method of the two-point probability function and the lineal-path function, i.e., using incremental calculation instead of conventional global calculation, the efficiency of reconstructing 3D large-scale porous media was greatly improved. Besides, in the later stage of reconstruction that the basic structure of porous media had been formed, the pixel selection algorithm was performed to speed up the reconstruction process. To evaluate the accuracy of the improved SAA, the similarity between the 3D reconstructed volume and the reference image of prototype sandstone was examined. The results showed good agreement between the reconstructed model and the references. The efficiency of the improved SAA was verified by comparison with the conventional SAA, the results of which indicated that the improved SAA can significantly shorten the reconstruction time of 3D large-scale porous media.

Variations in the Size and Shape of Human Cochlear Malformation Types.

ABSTRACTThe objective of this study is to determine the variations in size and shape of the most widely recognized cochlear malformation types using three‐dimensional (3D) visualization. Using 3D slicer freeware, the complete inner‐ear structures were segmented from 46 anonymized high‐resolution computed tomography (HRCT) image datasets. Cochlear height, internal auditory canal height, and width were measured from the axial plane. Cochlear basal turn diameter was measured from the oblique coronal plane. Number of cochlear turns was measured from the 3D images and the corresponding cochlear duct length (CDL) was estimated using the CDL equations given in Alexiades et al. [Otol Neurotol 36 (2015) 904–907]. Out of 46 preoperative HRCT image datasets of human temporal bone, cochlear anatomy types including normal anatomy (4), enlarged vestibular aqueduct syndrome (3), cochlear aplasia (2), incomplete partition Types I (8), II (Mondini's deformity) (3), and III (X‐linked) (4), cochlear hypoplasia (CH) (17), and common cavity (CC) (5) were identified. Majority of CH cases had cochlear height shorter than 4 mm whereas the CC cases measured cochlear height above 6 mm. For all the other malformation types, cochlear height was between 4 and 6 mm. In terms of “A” value, majority of CH cases showed shorter “A” value of <7.5 mm, which is in the lower end in comparison to the rest of the malformation types reported in this study. 3D‐visualization shows the size and shape variations of all the structures of inner ear and also improves the clinicians' ability to visualize cochlear anatomy and nearby structures much easier than from the 2D image slices. Anat Rec, 302:1792–1799, 2019. © 2019 The Author. The Anatomical Record published by Wiley Periodicals, Inc. on behalf of American Association for Anatomy

Field Reconstruction of Holograms for Interactive Free Space True Three Dimensional Display

Nowadays, the 3D display technology has attracted great academic and industrial attention due to its rapid development and applications in providing more realistic, natural, and extra depth images far superior than the traditional 2D display. But traditional 3D techniques suffer from many drawbacks and hard to meet the requirement of commercialization. In the present study, the field reconstruction method of holograms is proposed for the first time to realize the display of true 3D image in free space which is more practical and better than traditional holographic technique in image quality. An optical switch array (OSA) and a projection lens are used to generate a series of 2D image slice, which can form a virtual 3D object, in free space. Then the light from the virtual object and reference light interference with each other forming an interferometric fringe on the screen and thus the 3D image of the virtual object is reconstructed in free space, timely. Meanwhile, the OSA can work as an image generator and sensor simultaneously which allows user to interact with the reconstructed 3D scenery without extra equipment.

A Geant4-based Monte Carlo study of a benchtop multi-pinhole X-ray fluorescence computed tomography imaging.

BackgroundX-ray fluorescence (XRF) computed tomography (XFCT) has shown promise for molecular imaging of metal nanoparticles such as gold nanoparticles (GNPs) and benchtop XFCT is under active development due to its easy access, low-cost instrumentation and operation.PurposeTo validate the performance of a Geant4-based Monte Carlo (MC) model of a benchtop multi-pinhole XFCT system for quantitative imaging of GNPs.MethodsThe MC mode consisted of a fan-beam x-ray source (125 kVp), which was used to stimulate the emission of XRF from the GNPs, a phantom (3 cm in diameter) which included six or nine inserts (3 mm in diameter), each of which contained the same (1 wt. %) or various (0.08–1 wt. %) concentrations of GNPs, a multi-pinhole collimator which could acquire multiple projections simultaneously and a one-sided or two-sided two-dimensional (2D) detector. Various pinhole diameters (3.7, 2, 1, 0.5 and 0.25 mm) and various particle numbers (20, 40, 80 and 100 billion) were simulated and the results for single pinhole and multi-pinhole (9 pinholes) imaging were compared.ResultsThe image resolution for a 1 mm multi-pinhole was between 0.88 and 1.38 mm. The detection limit for multi-pinhole operation was about 0.09 wt. %, while that for the single pinhole was about 0.13 wt. %. For a fixed number of pinholes, noise increased with decreasing number of photons.ConclusionThe MC mode could acquire 2D slice images of the object without rotation and demonstrated that a multi-pinhole XFCT imaging system could be a potential bioimaging modality for nanomedical applications.

Mesoscopic modeling method of concrete based on statistical analysis of CT images

Few efforts have been made to study actual coarse aggregate shapes in mesoscopic simulation analysis, whereas it has a significant impact on stress distribution and micro-crack propagation during the failure process of concrete. In this study, the true characteristics of aggregates (i.e. edge number, area, equivalent radius, elongation ratio and size distribution) were statistically analyzed by processing the real 2D slice images of concrete obtained from X-ray computed tomography technology. A polygon random aggregate modeling method, considering above statistical features of aggregate, is proposed and verified by an experimental investigation. The test results show that the established model can give an accurate description of the stress-strain behavior and damage evolution of concrete. Moreover, several representative meso-scale concrete models were established, which provides a reference in building meso-scale models with various sizes and cross-sectional shapes, especially for large-scale concrete members that are unable to undergo CT testing due to limitation of operational space.

Transfer Learning From Convolutional Neural Networks for Computer-Aided Diagnosis: A Comparison of Digital Breast Tomosynthesis and Full-Field Digital Mammography

With the growing adoption of digital breast tomosynthesis (DBT) in breast cancer screening, we compare the performance of deep learning computer-aided diagnosis on DBT images to that of conventional full-field digital mammography (FFDM). In this study, we retrospectively collected FFDM and DBT images of 78 biopsy-proven lesions from 76 patients. A region of interest was selected for each lesion on FFDM, synthesized 2D, and DBT key slice images. Features were extracted from each lesion using a pretrained convolutional neural network (CNN) and served as input to a support vector machine classifier trained in the task of predicting likelihood of malignancy. From receiver operating characteristic (ROC) analysis of all 78 lesions, the synthesized 2D image performed best in both the cradiocaudal view (area under the ROC curve [AUC] = 0.81, SE = 0.05) and mediolateral oblique view (AUC = 0.88, SE = 0.04) in the task of lesion characterization. When cradiocaudal and mediolateral oblique data of each lesion were merged through soft voting, DBT key slice image performed best (AUC = 0.89, SE = 0.04). When only masses and architectural distortions (ARDs) were considered, DBT performed significantly better than FFDM (p = 0.024). DBT performed significantly better than FFDM in the merged view classification of mass and ARD lesions. The increased performance suggests that the information extracted by the CNN from DBT images may be more relevant to lesion malignancy status than the information extracted from FFDM images. Therefore, this study provides supporting evidence for the efficacy of computer-aided diagnosis on DBT in the evaluation of mass and ARD lesions.

A new method for spatial structure detection of complex inner cavities based on 3D γ-photon imaging

This paper presents a new three-dimensional (3D) imaging method for detecting the spatial structure of a complex inner cavity based on positron annihilation and γ-photon detection. This method first marks carrier solution by a certain radionuclide and injects it into the inner cavity where positrons are generated. Subsequently, γ-photons are released from positron annihilation, and the γ-photon detector ring is used for recording the γ-photons. Finally, the two-dimensional (2D) image slices of the inner cavity are constructed by the ordered-subset expectation maximization scheme and the 2D image slices are merged to the 3D image of the inner cavity. To eliminate the artifact in the reconstructed image due to the scattered γ-photons, a novel angle-traversal model is proposed for γ-photon single-scattering correction, in which the path of the single scattered γ-photon is analyzed from a spatial geometry perspective. Two experiments are conducted to verify the effectiveness of the proposed correction model and the advantage of the proposed testing method in detecting the spatial structure of the inner cavity, including the distribution of gas–liquid multi-phase mixture inside the inner cavity. The above two experiments indicate the potential of the proposed method as a new tool for accurately delineating the inner structures of industrial complex parts.

Mass detection in digital breast tomosynthesis data using convolutional neural networks and multiple instance learning

Digital breast tomosynthesis (DBT) was developed in the field of breast cancer screening as a new tomographic technique to minimize the limitations of conventional digital mammography breast screening methods. A computer-aided detection (CAD) framework for mass detection in DBT has been developed and is described in this paper. The proposed framework operates on a set of two-dimensional (2D) slices. With plane-to-plane analysis on corresponding 2D slices from each DBT, it automatically learns complex patterns of 2D slices through a deep convolutional neural network (DCNN). It then applies multiple instance learning (MIL) with a randomized trees approach to classify DBT images based on extracted information from 2D slices. This CAD framework was developed and evaluated using 5040 2D image slices derived from 87 DBT volumes. The empirical results demonstrate that this proposed CAD framework achieves much better performance than CAD systems that use hand-crafted features and deep cardinality-restricted Bolzmann machines to detect masses in DBTs.

An automated workflow for segmenting single adult cardiac cells from large-volume serial block-face scanning electron microscopy data.

This paper presents a new algorithm to automatically segment the myofibrils, mitochondria and nuclei within single adult cardiac cells that are part of a large serial-block-face scanning electron microscopy (SBF-SEM) dataset. The algorithm only requires a set of manually drawn contours that roughly demarcate the cell boundary at routine slice intervals (every 50th, for example). The algorithm correctly classified pixels within the single cell with 97% accuracy when compared to manual segmentations. One entire cell and the partial volumes of two cells were segmented. Analysis of segmentations within these cells showed that myofibrils and mitochondria occupied 47.5% and 51.6% on average respectively, while the nuclei occupy 0.7% of the cell for which the entire volume was captured in the SBF-SEM dataset. Mitochondria clustering increased at the periphery of the nucleus region and branching points of the cardiac cell. The segmentations also showed high area fraction of mitochondria (up to 70% of the 2D image slice) in the sub-sarcolemmal region, whilst it was closer to 50% in the intermyofibrillar space. We finally demonstrate that our segmentations can be turned into 3D finite element meshes for cardiac cell computational physiology studies. We offer our large dataset and MATLAB implementation of the algorithm for research use at www.github.com/CellSMB/sbfsem-cardiac-cell-segmenter/. We anticipate that this timely tool will be of use to cardiac computational and experimental physiologists alike who study cardiac ultrastructure and its role in heart function.

Speckle noise reduction for 3D ultrasound images by optimum threshold parameter estimation of bi-dimensional empirical mode decomposition using Fisher discriminant analysis

This paper presents an approach to speckle noise reduction for 3D ultrasound images using bi-dimensional empirical mode decomposition (BEMD). 3D ultrasound is a popular diagnostic system for assessing the progression of diseases for its diverse benefits and application. Speckle noise often obscures the fine details and degrades the spatial resolution and, contrast quality that makes the interpretation of ultrasound image more difficult. The proposed method estimates an optimum threshold value of intrinsic mode functions (IMFs) using Fisher discriminant analysis (FDA) for reducing the speckles in 3D volume of ultrasound images. FDA has applied on 2D IMFs, then explored and extended to 3D. The 3D volume rendering is performed on the basis of integrating 2D slice images that provide strong speckle reduction and edge preservation. The experiment result has compared with the several other state-of-the-art threshold methods. The proposed method is also good in edge preservation and contrast resolution.

Automated Defect Recognition as a Critical Element of a Three Dimensional X-ray Computed Tomography Imaging-Based Smart Non-Destructive Testing Technique in Additive Manufacturing of Near Net-Shape Parts

In this paper, a state of the art automated defect recognition (ADR) system is presented that was developed specifically for Non-Destructive Testing (NDT) of powder metallurgy (PM) parts using three dimensional X-ray Computed Tomography (CT) imaging, towards enabling online quality assurance and enhanced integrity confidence. PM parts exhibit typical defects such as microscopic cracks, porosity, and voids, internal to components that without an effective detection system, limit the growth of industrial applications. Compared to typical testing methods (e.g., destructive such as metallography that is based on sampling, cutting, and polishing of parts), CT provides full coverage of defect detection. This paper establishes the importance and advantages of an automated NDT system for the PM industry applications with particular emphasis on image processing procedures for defect recognition. Moreover, the article describes how to establish a reference library based on real 3D X-ray CT images of net-shape parts. The paper follows the development of the ADR system from processing 2D image slices of a measured 3D X-ray image to processing the complete 3D X-ray image as a whole. The introduced technique is successfully integrated into an automated in-line quality control system highly sought by major industry sectors in Oil and Gas, Automotive, and Aerospace.