Search NASA⌕ Search

SEARCH · Search NASA

Results for “Image segmentation”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 451 records · Page 25

Stereology of backscatter electron images of etched surfaces for characterization of particle size distributions and volume fractions: Estimation of imaging bias via Monte Carlo simulations

On metallic specimens in which a secondary phase has been selectively removed by a chemical etchant, the use of backscatter electron (BSE) imaging yields images that are more readily segmented with image processing algorithms than other modes of imaging in the scanning electron microscope. The contrast mechanisms in this imaging mode, however, produce a bias in the observation of particle sizes and volume fractions due to the effects of the electron interaction volume in the specimen. This stereological bias is quantified using Monte Carlo (MC) simulation of backscatter images. It is observed that the overprojection of features with centroids residing beneath the plane of polish is largely canceled out by the reduced segmentation size of features with centroids residing above the plane of polish. - Research Highlights: {yields} Backscatter imaging of selectively-etched surfaces can facilitate segmentation. {yields} Backscatter imaging of voids is simulated to estimate imaging/observation biases. {yields} The biases are quantified and incorporated into the stereological calculation. {yields} Systematic errors and imaging biases are observed to counteract one another. {yields} Results are illustrated using a bimodal gamma prime distribution in a Ni superalloy.

36 MATERIALS SCIENCE↗

A Comparison of Amplitude-Based and Phase-Based Positron Emission Tomography Gating Algorithms for Segmentation of Internal Target Volumes of Tumors Subject to Respiratory Motion

Purpose: To quantitatively compare the accuracy of tumor volume segmentation in amplitude-based and phase-based respiratory gating algorithms in respiratory-correlated positron emission tomography (PET). Methods and Materials: List-mode fluorodeoxyglucose-PET data was acquired for 10 patients with a total of 12 fluorodeoxyglucose-avid tumors and 9 lymph nodes. Additionally, a phantom experiment was performed in which 4 plastic butyrate spheres with inner diameters ranging from 1 to 4 cm were imaged as they underwent 1-dimensional motion based on 2 measured patient breathing trajectories. PET list-mode data were gated into 8 bins using 2 amplitude-based (equal amplitude bins [A1] and equal counts per bin [A2]) and 2 temporal phase-based gating algorithms. Gated images were segmented using a commercially available gradient-based technique and a fixed 40% threshold of maximum uptake. Internal target volumes (ITVs) were generated by taking the union of all 8 contours per gated image. Segmented phantom ITVs were compared with their respective ground-truth ITVs, defined as the volume subtended by the tumor model positions covering 99% of breathing amplitude. Superior-inferior distances between sphere centroids in the end-inhale and end-exhale phases were also calculated. Results: Tumor ITVs from amplitude-based methods were significantly larger than those from temporal-based techniques (P=.002). For lymph nodes, A2 resulted in ITVs that were significantly larger than either of the temporal-based techniques (P<.0323). A1 produced the largest and most accurate ITVs for spheres with diameters of ≥2 cm (P=.002). No significant difference was shown between algorithms in the 1-cm sphere data set. For phantom spheres, amplitude-based methods recovered an average of 9.5% more motion displacement than temporal-based methods under regular breathing conditions and an average of 45.7% more in the presence of baseline drift (P<.001). Conclusions: Target volumes in images generated from amplitude-based gating are larger and more accurate, at levels that are potentially clinically significant, compared with those from temporal phase-based gating.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Inter-slice bidirectional registration-based segmentation of the prostate gland in MR and CT image sequences

Purpose: Accurate segmentation and volume estimation of the prostate gland in magnetic resonance (MR) and computed tomography (CT) images are necessary steps in diagnosis, treatment, and monitoring of prostate cancer. This paper presents an algorithm for the prostate gland volume estimation based on the semiautomated segmentation of individual slices in T2-weighted MR and CT image sequences. Methods: The proposedInter-Slice Bidirectional Registration-based Segmentation (iBRS) algorithm relies on interslice image registration of volume data to segment the prostate gland without the use of an anatomical atlas. It requires the user to mark only three slices in a given volume dataset, i.e., the first, middle, and last slices. Next, the proposed algorithm uses a registration algorithm to autosegment the remaining slices. We conducted comprehensive experiments to measure the performance of the proposed algorithm using three registration methods (i.e., rigid, affine, and nonrigid techniques). Results: The results with the proposed technique were compared with manual marking using prostate MR and CT images from 117 patients. Manual marking was performed by an expert user for all 117 patients. The median accuracies for individual slices measured using the Dice similarity coefficient (DSC) were 92% and 91% for MR and CT images, respectively. The iBRS algorithm was also evaluated regarding user variability, which confirmed that the algorithm was robust to interuser variability when marking the prostate gland. Conclusions: The proposed algorithm exploits the interslice data redundancy of the images in a volume dataset of MR and CT images and eliminates the need for an atlas, minimizing the computational cost while producing highly accurate results which are robust to interuser variability.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Universal image representation based on a multimodal graph

A system for classifying a target image with segments having attributes is provided. The system generates a graph for the target image that includes vertices representing segments of the image and edges representing relationships between the connected vertices. For each vertex, the system generates a subgraph that includes the vertex as a home vertex and neighboring vertices representing segments of the target image within a neighborhood of the segment represented by the home vertex. The system applies an autoencoder to each subgraph to generate latent variables to represent the subgraph. The system applies a machine learning algorithm to a feature vector comprising a universal image representation of the target image that is derived from the generated latent variables of the subgraphs to generate a classification for the target image.

Bremer, Peer-Timo↗

Universal image representation based on a multimodal graph

A system for classifying a target image with segments having attributes is provided. The system generates a graph for the target image that includes vertices representing segments of the image and edges representing relationships between the connected vertices. For each vertex, the system generates a subgraph that includes the vertex as a home vertex and neighboring vertices representing segments of the target image within a neighborhood of the segment represented by the home vertex. The system applies an autoencoder to each subgraph to generate latent variables to represent the subgraph. The system applies a machine learning algorithm to a feature vector comprising a universal image representation of the target image that is derived from the generated latent variables of the subgraphs to generate a classification for the target image.

Bremer, Peer-Timo↗

Rapid macrovoid characterization in membranes prepared via nonsolvent-induced phase separation: A comparison between 2D and 3D techniques

Optimizing the performance of asymmetric membranes prepared via nonsolvent-induced phase separation (NIPS) requires a quantitative understanding of how processing variables influence membrane morphology. Presently, the most useful structural quantification techniques require 3D visualization of the membrane structure and are best suited for studies seeking detailed information on small datasets. This study proposes and validates a rapid and accurate technique for quantifying macroporosity (i.e., D m ), a simple descriptor of sublayer macrovoid content in asymmetric membranes D m . values measured from segmented cross-sectional imaging performed via X-ray computed tomography (XCT) and scanning electron microscopy (SEM) are presented and compared for three asymmetric membranes prepared from commercial polymers. Importantly, analyses of 3D XCT membrane reconstructions reveal that D m is described by a single, centralized mean, which demonstrates that macrovoid content is spatially homogenous. Thus, D m can be approximated from limited sampling of the 2D cross-sectional membrane structure via SEM. A proposed 2D SEM sampling method provides D m estimates within ±6% of corresponding 3D XCT values with 30 independent measurements for the three membranes. Further sensitivity is achieved using complementary descriptors such as macrovoid count density (i.e., C m ). This technique is thus a useful tool for characterizing macroporosity from a broad selection of membrane samples.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Reduced-bias multi-resolution coherent change detection

Described herein are various technologies relating to generating a reduced bias multi-resolution (RBMR) coherent change detection (CCD) image product. A low bias, low resolution (LBLR) CCD image product is generated based upon a first SAR image of a scene and a second SAR image of the scene, where each pixel in the LBLR CCD image has a respective value that is computed using a large box-filter. The LBLR CCD image is segmented into a plurality of segments, and a box-filter size is determined for each segment based upon values of pixels in the segment. A RBMR CCD image product is then computed using box-filters of the determined sizes.

West, Roger Derek↗

A Refined Global Segmentation of X-Ray CT Images for Multi-phase Geomaterials

X-ray computed tomography images of three-phase silica sand and glass bead specimens are analyzed and used to evaluate the segmentation performances of Otsu-, and recursion-based multilevel algorithms. A global image segmentation technique that combines iterative and recursive algorithms, namely a refined statistics-based global segmentation is proposed for segmenting multi-phase granular geomaterials. The performance of the proposed algorithm is tested by segmenting partially saturated silica sand and glass bead specimens. For the tested silica sand specimen, the refined statistics method estimated void ratio and degree of saturation were 0.67 and 39.35%. The estimates for the glass bead specimen yielded 0.64 and 43.49%, respectively. The true void ratio (0.66) and degree of saturation (37.71%) were determined with a user-controlled Image processing software package—Image-Pro. It was found that the proposed method estimated the void ratio and the degree of saturation with 1.52 and 4.35 percent errors for the silica sand and with 15.63 and 0.34 percent errors for the glass bead, respectively. The computational time of the proposed method was found to be shorter than other methods considered. Overall, it is concluded that the proposed technique performed better in segmenting three-phase granular geomaterials.

42 ENGINEERING↗

Semi-Automated Neuron Boundary Detection and Nonbranching Process Segmentation in Electron Microscopy Images

Neuroscientists are developing new imaging techniques and generating large volumes of data in an effort to understand the complex structure of the nervous system. The complexity and size of this data makes human interpretation a labor-intensive task. To aid in the analysis, new segmentation techniques for identifying neurons in these feature rich datasets are required. This paper presents a method for neuron boundary detection and nonbranching process segmentation in electron microscopy images and visualizing them in three dimensions. It combines both automated segmentation techniques with a graphical user interface for correction of mistakes in the automated process. The automated process first uses machine learning and image processing techniques to identify neuron membranes that deliniate the cells in each two-dimensional section. To segment nonbranching processes, the cell regions in each two-dimensional section are connected in 3D using correlation of regions between sections. The combination of this method with a graphical user interface specially designed for this purpose, enables users to quickly segment cellular processes in large volumes.

Connectomics↗

A proximity‐based image‐processing algorithm for colloid assignment in segmented multiphase flow datasets

Summary Colloidal transport and deposition are of both environmental and engineering importance. Easier access to x‐ray microtomography (XMT) coupled with improved imaging resolution has made XMT a unique and viable tool for visualizing and quantifying these processes. Currently, there is scant information in the literature addressing colloid segmentation and analysis in saturated and unsaturated porous media, in particular related to spatial partitioning of colloids. To support this need, an approach to assign segmented colloidal particles and aggregates to different partitioning classes based on their proximity to different phases is presented here. The method uses different markers for each attachment site (e.g. wetting‐nonwetting phase interfaces). An example XMT dataset from a drainage experiment is used to demonstrate the efficacy of the image processing algorithms. Flow conditions, and fluid and colloid properties, can thus be compared to the behaviour of colloids within the porous medium. This algorithm can help elucidate colloidal deposition mechanisms and the importance of different attachment sites, explore the importance of fluid properties, as well as the arrangement and shape of the colloids.

BRUECK, C. L.↗

Multi-Task Learning of Scanning Electron Microscopy and Synthetic Thermal Tomography Images for Detection of Defects in Additively Manufactured Metals

One of the key challenges in laser powder bed fusion (LPBF) additive manufacturing of metals is the appearance of microscopic pores in 3D-printed metallic structures. Quality control in LPBF can be accomplished with non-destructive imaging of the actual 3D-printed structures. Thermal tomography (TT) is a promising non-contact, non-destructive imaging method, which allows for the visualization of subsurface defects in arbitrary-sized metallic structures. However, because imaging is based on heat diffusion, TT images suffer from blurring, which increases with depth. We have been investigating the enhancement of TT imaging capability using machine learning. In this work, we introduce a novel multi-task learning (MTL) approach, which simultaneously performs the classification of synthetic TT images, and segmentation of experimental scanning electron microscopy (SEM) images. Synthetic TT images are obtained from computer simulations of metallic structures with subsurface elliptical-shaped defects, while experimental SEM images are obtained from imaging of LPBF-printed stainless-steel coupons. MTL network is implemented as a shared U-net encoder between the classification and the segmentation tasks. Results of this study show that the MTL network performs better in both the classification of synthetic TT images and the segmentation of SEM images tasks, as compared to the conventional approach when the individual tasks are performed independently of each other.

36 MATERIALS SCIENCE↗

Automated method for tracing leading and trailing processes of migrating neurons in confocal image sequences

Segmentation, tracking, and tracing of neurons in video imagery are important steps in many neuronal migration studies and can be in accurate and time-consuming when performed manually. In this paper, we present an automated method for tracing the leading and trailing processes of migrating neurons in time-lapse image stacks acquired with a confocal fluorescence microscope. In our approach, we first locate and track the soma of the cell of interest by smoothing each frame and tracking the local maxima through the sequence. We then trace the leading process in each frame by starting at the center of the soma and stepping repeatedly in the most likely direction of the leading process. This direction is found at each step by examining second derivatives of fluorescent intensity along curves of constant radius around the current point. Tracing terminates after a fixed number of steps or when fluorescent intensity drops below a fixed threshold. We evolve the resulting trace to form an improved trace that more closely follows the approximate centerline of the leading process. We apply a similar algorithm to the trailing process of the cell by starting the trace in the opposite direction. We demonstrate our algorithm on two time-lapse confocal video sequences of migrating cerebellar granule neurons(CGNs). We show that the automated traces closely approximate ground truth traces to within 1 or 2 pixels on average. Additionally, we compute line intensity profiles of fluorescence along the automated traces and quantitatively demonstrate their similarity to manually generated profiles in terms of fluorescence peak locations.

99 GENERAL AND MISCELLANEOUS↗

Machine learning-assisted upscaling analysis of reservoir rock core properties based on micro-computed tomography imagery

Optimum solutions for geologic modeling and reservoir simulation in industries such as oil and gas recovery and carbon capture and storage require accurate characterization of reservoir properties, which are often heterogeneous. In this study, high-quality micro-computed tomography (CT) images (1.475-μm/pixel resolution) of a sandstone core acquired from the Bell Creek oil field, USA, were used to provide nondestructive analysis of pore- and core-scale heterogeneity across measurement scales of 94–566 μm. In addition to characterizing the as-received sample, the core sample was flooded with brine to evaluate the capacity of the core sample to receive injected fluids. The micro-CT images were systematically segmented into pore spaces and grains via machine learning (ML) steps including image preprocessing, label creation using a traditional ML method based on limited manual image annotation, and finally U-Net segmentation. The segmented image stacks were reconstructed into digital cubes of various scales of voxel lengths. The 3D porosity values were calculated for all the digital cubes, and the fractal dimensions of the cubes were estimated using a box-counting method. The results showed that smaller cubes had greater heterogeneity and that the porosity values could be accurately estimated by fractal dimension and voxel lengths using ML models. For the core sample with brine flooding, the ratio of pores filled by brine to the total pore space was related to the porosity and could also be accurately estimated by porosity, fractal dimension, and voxel lengths using ML models. In conclusion, the results of this study demonstrate that the concept of fractal dimension can be a useful vector to perform upscaling analysis of sandstone rock heterogeneity from the pore to core scale and that fractal dimensions can be used to estimate porosity values and pore space-filling capacity across those scales.

58 GEOSCIENCES↗

SU-F-303-12: Implementation of MR-Only Simulation for Brain Cancer: A Virtual Clinical Trial

Purpose: To perform a retrospective virtual clinical trial using an MR-only workflow for a variety of brain cancer cases by incorporating novel imaging sequences, tissue segmentation using phase images, and an innovative synthetic CT (synCT) solution. Methods: Ten patients (16 lesions) were evaluated using a 1.0T MR-SIM including UTE-DIXON imaging (TE = 0.144/3.4/6.9ms). Bone-enhanced images were generated from DIXON-water/fat and inverted UTE. Automated air segmentation was performed using unwrapped UTE phase maps. Segmentation accuracy was assessed by calculating intersection and Dice similarity coefficients (DSC) using CT-SIM as ground truth. SynCTs were generated using voxel-based weighted summation incorporating T2, FLAIR, UTE1, and bone-enhanced images. Mean absolute error (MAE) characterized HU differences between synCT and CT-SIM. Dose was recalculated on synCTs; differences were quantified using planar gamma analysis (2%/2 mm dose difference/distance to agreement) at isocenter. Digitally reconstructed radiographs (DRRs) were compared. Results: On average, air maps intersected 80.8 ±5.5% (range: 71.8–88.8%) between MR-SIM and CT-SIM yielding DSCs of 0.78 ± 0.04 (range: 0.70–0.83). Whole-brain MAE between synCT and CT-SIM was 160.7±8.8 HU, with the largest uncertainty arising from bone (MAE = 423.3±33.2 HU). Gamma analysis revealed pass rates of 99.4 ± 0.04% between synCT and CT-SIM for the cohort. Dose volume histogram analysis revealed that synCT tended to yield slightly higher doses. Organs at risk such as the chiasm and optic nerves were most sensitive due to their proximities to air/bone interfaces. DRRs generated via synCT and CT-SIM were within clinical tolerances. Conclusion: Our approach for MR-only simulation for brain cancer treatment planning yielded clinically acceptable results relative to the CT-based benchmark. While slight dose differences were observed, reoptimization of treatment plans and improved image registration can address this limitation. Future work will incorporate automated registration between setup images (cone-beam CT and kilovoltage images) for synCT and CT-SIM. Submitting institution holds research agreements with Philips HealthCare, Best, Netherlands and Varian Medical Systems, Palo Alto, CA. Research partially sponsored via an Internal Mentored Research Grant.

60 APPLIED LIFE SCIENCES↗

Pore system characterization of organic-rich shales using nanoscale-resolution 3D imaging

The Vaca Muerta Shale in Argentina is the first signifcant commercial shale oil/gas play outside of North America. High-resolution 2D/3D imaging of shale rocks for the purpose of establishing their mineralogy, total, and connected porosities has become more and more sophisticated. In this paper, nanoscale-resolution focused ion beam (FIB)-scanning electron microscopy (SEM) nano-tomography was used to obtain images of pore structures within two organic-rich regions of interest (ROIs), selected based on correlative SEM and automated mineralogy maps. Advanced machine learning classification tools were used to segment the images and assign porosities and other components. Pore size distribution and pore connectivity analyzes revealed that about 95% of all the pores, present within the two ROIs, had a diameter of less than approximately 75 nm, and that most of these pores were poorly connected. In a similar fashion, the flow rate distribution analysis showed that pores with diameters of about 150-330 nm contributed to over 50% of the flow capacity of the connected pore systems. These results suggest that although most of the pores typically found in shales have pore diameter smaller than about 100 nm, most of the hydrocarbon production may be carried by a relatively small number of larger connected pores with pore diameter greater than about 150 nm. This research implies that a large portion of the organic-hosted pores (with diameter typically smaller than about 100 nm) may not provide permeable flow pathways for the oil and/or gas migration, and hence may have very little contribution to the hydrocarbon production.

58 GEOSCIENCES↗

Sci-Thur AM: YIS – 03: Combining sagittally-reconstructed 3D and live-2D ultrasound for high-dose-rate prostate brachytherapy needle segmentation

Ultrasound-guided high-dose-rate prostate brachytherapy (HDR-BT) needle segmentation is performed clinically using live-2D sagittal images. Organ segmentation is then performed using axial images, introducing a source of geometric uncertainty. Sagittally-reconstructed 3D (SR3D) ultrasound enables both needle and organ segmentation, but suffers from shadow artifacts. We present a needle segmentation technique augmenting SR3D with live-2D sagittal images using mechanical probe tracking to mitigate image artifacts and compare it to the clinical standard. Seven prostate cancer patients underwent TRUS-guided HDR-BT during which the clinical and proposed segmentation techniques were completed in parallel using dual ultrasound video outputs. Calibrated needle end-length measurements were used to calculate insertion depth errors (IDEs), and the dosimetric impact of IDEs was evaluated by perturbing clinical treatment plan source positions. The proposed technique provided smaller IDEs than the clinical approach, with mean±SD of −0.3±2.2 mm and −0.5±3.7mm respectively. The proposed and clinical techniques resulted in 84% and 43% of needles with IDEs within ±3mm, and IDE ranges across all needles of [−7.7mm, 5.9mm] and [−9.3mm, 7.7mm] respectively. The proposed and clinical IDEs lead to mean±SD changes in the volume of the prostate receiving the prescription dose of −0.6±0.9% and −2.0±5.3% respectively. The proposed technique provides improved HDR-BT needle segmentation accuracy over the clinical technique leading to decreased dosimetric uncertainty by eliminating the axial-to-sagittal registration, and mitigates the effect of shadow artifacts by incorporating mechanically registered live-2D sagittal images.

60 APPLIED LIFE SCIENCES↗

Simultaneous hierarchical segmentation and vectorization of satellite images through combined data sampling and anisotropic triangulation

The automatic detection, recognition , and segmentation of object classes in remote sensed images is of crucial importance for scene interpretation and understanding. However, it is a difficult task because of the high variability of satellite data. Indeed, the observed scenes usually exhibit a high degree of complexity, where complexity refers to the large variety of pictorial representations of objects with the same semantic meaning and also to the extensive amount of available det.ails. Therefore, there is still a strong demand for robust techniques for automatic information extraction and interpretation of satellite images. In parallel, there is a growing interest in techniques that can extract vector features directly from such imagery. In this paper, we investigate the problem of automatic hierarchical segmentation and vectorization of multispectral satellite images. We propose a new algorithm composed of the following steps: (i) a non-uniform sampling scheme extracting most salient pixels in the image, (ii) an anisotropic triangulation constrained by the sampled pixels taking into account both strength and directionality of local structures present in the image, (iii) a polygonal grouping scheme merging, through techniques based on perceptual information , the obtained segments to a smaller quantity of superior vectorial objects. Besides its computational efficiency, this approach provides a meaningful polygonal representation for subsequent image analysis and/or interpretation.

47 OTHER INSTRUMENTATION↗

Influence of cold walls on PET image quantification and volume segmentation: A phantom study

Purpose: Commercially available fillable plastic inserts used in positron emission tomography phantoms usually have thick plastic walls, separating their content from the background activity. These “cold” walls can modify the intensity values of neighboring active regions due to the partial volume effect, resulting in errors in the estimation of standardized uptake values. Numerous papers suggest that this is an issue for phantom work simulating tumor tissue, quality control, and calibration work. This study aims to investigate the influence of the cold plastic wall thickness on the quantification of 18F-fluorodeoxyglucose on the image activity recovery and on the performance of advanced automatic segmentation algorithms for the delineation of active regions delimited by plastic walls.Methods: A commercial set of six spheres of different diameters was replicated using a manufacturing technique which achieves a reduction in plastic walls thickness of up to 90%, while keeping the same internal volume. Both sets of thin- and thick-wall inserts were imaged simultaneously in a custom phantom for six different tumor-to-background ratios. Intensity values were compared in terms of mean and maximum standardized uptake values (SUVs) in the spheres and mean SUV of the hottest 1 ml region (SUV{sub max}, SUV{sub mean}, and SUV{sub peak}). The recovery coefficient (RC) was also derived for each sphere. The results were compared against the values predicted by a theoretical model of the PET-intensity profiles for the same tumor-to-background ratios (TBRs), sphere sizes, and wall thicknesses. In addition, ten automatic segmentation methods, written in house, were applied to both thin- and thick-wall inserts. The contours obtained were compared to computed tomography derived gold standard (“ground truth”), using five different accuracy metrics.Results: The authors' results showed that thin-wall inserts achieved significantly higher SUV{sub mean}, SUV{sub max}, and RC values (up to 25%, 16%, and 25% higher, respectively) compared to thick-wall inserts, which was in agreement with the theory. This effect decreased with increasing sphere size and TBR, and resulted in substantial (>5%) differences between thin- and thick-wall inserts for spheres up to 30 mm diameter and TBR up to 4. Thinner plastic walls were also shown to significantly improve the delineation accuracy for the majority of the segmentation methods tested, by increasing the proportion of lesion voxels detected, although the errors in image quantification remained non-negligible.Conclusions: This study quantified the significant effect of a 90% reduction in the thickness of insert walls on SUV quantification and PET-based boundary detection. Mean SUVs inside the inserts and recovery coefficients were particularly affected by the presence of thick cold walls, as predicted by a theoretical approach. The accuracy of some delineation algorithms was also significantly improved by the introduction of thin wall inserts instead of thick wall inserts. This study demonstrates the risk of errors deriving from the use of cold wall inserts to assess and compare the performance of PET segmentation methods.

62 RADIOLOGY AND NUCLEAR MEDICINE↗