Segmentation Artifacts Research Articles

Nanocrystal segmentation in scanning precession electron diffraction data.

Scanning precession electron diffraction (SPED) enables the local crystallography of materials to be probed on the nanoscale by recording a two-dimensional precession electron diffraction (PED) pattern at every probe position as a dynamically rocking electron beam is scanned across the specimen. SPED data from nanocrystalline materials commonly contain some PED patterns in which diffraction is measured from multiple crystals. To analyse such data, it is important to perform nanocrystal segmentation to isolate both the location of each crystal and a corresponding representative diffraction signal. This also reduces data dimensionality significantly. Here, two approaches to nanocrystal segmentation are presented, the first based on virtual dark-field imaging and the second on non-negative matrix factorization. Relative merits and limitations are compared in application to SPED data obtained from partly overlapping nanoparticles, and particular challenges are highlighted associated with crystals exciting the same diffraction conditions. It is demonstrated that both strategies can be used for nanocrystal segmentation without prior knowledge of the crystal structures present, but also that segmentation artefacts can arise and must be considered carefully. The analysis workflows associated with this work are provided open-source. LAY DESCRIPTION: Scanning precession electron diffraction is an electron microscopy technique that enables studies of the local crystallography of a broad selection of materials on the nanoscale. The technique involves the acquisition of a two-dimensional diffraction pattern for every probe position in an area of the sample. The four-dimensional dataset collected by this technique can typically comprise up to 500 000 diffraction patterns. For nanocrystalline materials, it is common that single diffraction patterns contain signals from overlapping crystals. To process such data, we use nanocrystal segmentation, where a representative diffraction pattern is constructed for each individual crystal, together with a real space image showing its morphology and location in the data. This reduces the dimensionality of the data and allows unmixing of signals from overlapping crystals. In this work, we demonstrate two methods for nanocrystal segmentation, one based on creating virtual dark-field images, and one based on unsupervised machine learning. A model system of partly overlapping nanoparticles is used to demonstrate the segmentation, and a demanding case for segmentation is highlighted, where some crystals are not discernible based on their diffraction patterns. To obtain a more complete nanocrystal segmentation, we add an image segmentation routine to both methods, and we discuss benefits and limitations of the two methods. The demonstration data and the used code are provided open-source, so that it can be used by everyone for analysis of nanocrystalline materials or as a starting point for further development of nanocrystal segmentation in scanning precession electron diffractiondata.

Comparison of Automated versus Manually Modified OCT Angiography En Face Slabs for Detection of Choroidal Neovascularization

To evaluate the sensitivity and specificity for the detection of choroidal neovascularization (CNV) using automatically generated versus manually modified swept-source OCT angiography (SS-OCTA) en face images. Prospective cohort study. Consecutive patients with various chorioretinal diseases and subretinal hyperreflective material (SHRM) or pigment epithelial detachments (PEDs) on OCT possibly corresponding to CNV in at least 1 eye. A total of 102 eyes of 63 patients were included in this study. Fluorescein angiography (FA) and SS-OCTA imaging (PLEX Elite 9000, Carl Zeiss Meditec, Dublin, CA) were performed at the same day. OCTA en face images were generated using the "retina," "avascular," "choriocapillaris," and "outer retina to choriocapillaris (ORCC)" slabs automatically provided by the software. In addition, a custom slab was created by manual modification of the automatically provided boundary "retinal pigment epithelium fit" positioned at the level of Bruch's membrane and anterior to any SHRM or PED to ensure that a possible CNV was captured in its entirety. Two graders independently evaluated OCTA en face images for the presence of CNV masked to all other images of the patient. Sensitivity and specificity for detection of CNV using FA as the reference. In 40% of cases (41/102), a CNV was detected on FA. Automatically provided OCTA en face slabs revealed the highest sensitivity for the "ORCC" slab (46.3%), followed by the "retina" slab (22.0%), "avascular" slab (17.1%), and "choriocapillaris" slab (14.6%). Specificity ranged between 93.4% for "ORCC" and 100% for the "retina" and "avascular" slabs. Sensitivity could be increased to 78.1% using the custom slab with a specificity of 88.5%. Concordance between FA and OCTA en face slabs was moderate for the "ORCC" slab (κ = 0.43; range, 0.41-0.60) and substantial for the custom slab (κ = 0.67; range, 0.61-0.80). Segmentation artifacts and incomplete coverage of CNV on SS-OCTA en face images may impede identification of CNV. Manual modification of the position of automatically generated segmentation lines anterior and posterior to any suspected CNV (SHRM or PED) increases the sensitivity of CNV detection compared with automatically generated slabs.

Controlling for Artifacts in Widefield Optical Coherence Tomography Angiography Measurements of Non-Perfusion Area

The recent clinical adoption of optical coherence tomography (OCT) angiography (OCTA) has enabled non-invasive, volumetric visualization of ocular vasculature at micron-scale resolutions. Initially limited to 3 mm × 3 mm and 6 mm × 6 mm fields-of-view (FOV), commercial OCTA systems now offer 12 mm × 12 mm, or larger, imaging fields. While larger FOVs promise a more complete visualization of retinal disease, they also introduce new challenges to the accurate and reliable interpretation of OCTA data. In particular, because of vignetting, wide-field imaging increases occurrence of low-OCT-signal artifacts, which leads to thresholding and/or segmentation artifacts, complicating OCTA analysis. This study presents theoretical and case-based descriptions of the causes and effects of low-OCT-signal artifacts. Through these descriptions, we demonstrate that OCTA data interpretation can be ambiguous if performed without consulting corresponding OCT data. Furthermore, using wide-field non-perfusion analysis in diabetic retinopathy as a model widefield OCTA usage-case, we show how qualitative and quantitative analysis can be confounded by low-OCT-signal artifacts. Based on these results, we suggest methods and best-practices for preventing and managing low-OCT-signal artifacts, thereby reducing errors in OCTA quantitative analysis of non-perfusion and improving reproducibility. These methods promise to be especially important for longitudinal studies detecting progression and response to therapy.

Method of Threshold CT Image Segmentation of Skeletal Bones

Background . Segmentation of images in the existing application software does not adequately separate the background area qualitatively, the allocation of anatomical structures, in particular skeletal bones, is partial with a significant number of artifacts, which are complicating further 3D modeling. Objective. The aim of the paper is development of the technique for automatized CT images segmentation of skeletal bones. Methods. The CT images of bone were segmented based on the developed algorithm, which included: threshold segmentation; morphological transformations of the unbound domains connections, the distance between them does not exceed the set value; the filling of areas with zero values, which are separated by pixels with values 1; comparison of segmentation results for neighboring sections. Testing techniques for segmentation multi-cut CT image of the patient with heterotopic ossification of the hip joints were analyzed. Segmentation results were compared with the images processed by specialists. The criteria for quality of segmentation were errors of the first and second kind: true-positive, true-negative, false-negative, false-positive voxels that were marked. Results. The developed algorithm for automatized segmentation of skeletal bones according to CT data shows 22% more qualitative results of research objects selection compared to usual threshold method; segmentation error was less than 8%. Calculated values of specificity were 99.9%, accuracy – 99.8%, sensitivity – 92.5% and for threshold method – 99.9%, 99.3%, and 70.7% respectively. Conclusions. The obtained results significantly reduce the time of CT images processing by a specialist in the area of radiation diagnostics and 3D printing of biological tissues and their models. Future prospects for the proposed methodology development are: its integration into specialized software tools with a user interface with a wide range of tools; improvement of machine code, reducing of computer time calculations; improvement of the segmentation algorithm, reducing of the segmentation artifacts.

Quantity and quality of image artifacts in optical coherence tomography angiography.

ObjectiveTo analyze quality and frequency of OCTA artifacts and to evaluate their impact on the interpretability of OCTA images.Design75 patients with diabetic retinopathy (DR), retinal artery occlusion (RAO), retinal vein occlusion (RVO), or neovascular age-related macular degeneration (nAMD) and healthy controls were enrolled in this cross-sectional study in the outpatient department of a tertiary eye care center.MethodsAll participants underwent an OCTA examination (spectral domain OCT Cirrus 5000 equipped with the AngioPlex module). OCTA scans were analyzed independently by two experienced ophthalmologists. Frequency of various artifacts for the entire OCTA scan and for different segmentation layers and the grading of OCTA interpretability were investigated.ResultsThe analysis of 75 eyes of 38 women and 37 men between 24 and 94 years were included. Six eyes had no retinal disease, 19 eyes had nAMD, 16 had DR, 19 eyes had RVO, and 15 eyes showed RAO. A macular edema (ME) was present in 40 of the diseased eyes. Projection artifacts occurred in all eyes in any structure below the superficial retinal vessel layer, segmentation and motion artifacts were found in 55% (41/75) and 49% (37/75) of eyes, respectively. Other artifacts occurred less frequently. Segmentation artifacts were significantly more frequent in diseased than in healthy eyes (p<0.01). Qualitative assessment of OCTA images was graded as excellent in 65% and sufficient in 25% of cases, adding up to 91% images deemed acceptable for examination. Presence of ME was associated with a significantly poorer interpretability (p<0.01).Conclusion and RelevanceVarious artifacts appear at different frequencies in OCTA images. Nevertheless, a qualitative assessment of the OCTA images is almost always possible. Good knowledge of possible artifacts and critical analysis of the complete OCTA dataset are essential for correct clinical interpretation and determining a precise clinical diagnosis.

Improving Arterial Spin Labeling by Using Deep Learning.

Purpose To develop a deep learning algorithm that generates arterial spin labeling (ASL) perfusion images with higher accuracy and robustness by using a smaller number of subtraction images. Materials and Methods For ASL image generation from pair-wise subtraction, we used a convolutional neural network (CNN) as a deep learning algorithm. The ground truth perfusion images were generated by averaging six or seven pairwise subtraction images acquired with (a) conventional pseudocontinuous arterial spin labeling from seven healthy subjects or (b) Hadamard-encoded pseudocontinuous ASL from 114 patients with various diseases. CNNs were trained to generate perfusion images from a smaller number (two or three) of subtraction images and evaluated by means of cross-validation. CNNs from the patient data sets were also tested on 26 separate stroke data sets. CNNs were compared with the conventional averaging method in terms of mean square error and radiologic score by using a paired t test and/or Wilcoxon signed-rank test. Results Mean square errors were approximately 40% lower than those of the conventional averaging method for the cross-validation with the healthy subjects and patients and the separate test with the patients who had experienced a stroke (P < .001). Region-of-interest analysis in stroke regions showed that cerebral blood flow maps from CNN (mean ± standard deviation, 19.7 mL per 100 g/min ± 9.7) had smaller mean square errors than those determined with the conventional averaging method (43.2 ± 29.8) (P < .001). Radiologic scoring demonstrated that CNNs suppressed noise and motion and/or segmentation artifacts better than the conventional averaging method did (P < .001). Conclusion CNNs provided superior perfusion image quality and more accurate perfusion measurement compared with those of the conventional averaging method for generation of ASL images from pair-wise subtraction images. © RSNA, 2017.

CHOROIDAL BLOOD FLOW VISUALIZATION IN HIGH MYOPIA USING A PROJECTION ARTIFACT METHOD IN OPTICAL COHERENCE TOMOGRAPHY ANGIOGRAPHY.

To visualize choroidal blood flow in larger vessels in highly myopic eyes using a phenomenon of the projection artifact to in the sclera using optical coherence tomography angiography. The retrospective study included 92 eyes (54 patients) with greater than 8 diopters of myopia. All eyes were examined using optical coherence tomography angiography (RTVue XR Avanti; Optovue Inc, Fremont, CA). The blood flow in choroidal vessels was evaluated by attempting to directly segment the choroid and also by placing the segmentation layer behind the choroid, within the sclera. Subfoveal choroidal thickness was also measured at the same time. The authors also evaluated the 54 normal eyes (54 cases) without high myopia as a control group. Segmentation artifacts occurred in 68 cases (73.9%) and precluded direct visualization of the choroidal blood flow in larger vessels. When the segmentation slab was placed posterior to the choroid within the sclera, the choroidal blood flow was visualized in 41 eyes (44.6%). The mean subfoveal choroidal thickness in eyes with visualization of choroidal blood flow was thinner than without visualization (50.3 ± 42.2 μm vs. 100.3 ± 44.4 μm, P < 0.01). Choroidal blood flow in larger vessels was imaged in no control eye. The choroidal vessel anatomy could be imaged by detecting flow using the projection artifact in the sclera with optical coherence tomography angiography. This technique may be useful in estimating the vascularity of the choroid.

A Cloud Computing Strategy for Region-Growing Segmentation

This paper proposes a distributed strategy for region-growing segmentation of arbitrarily large images. The strategy is suited for most distributed architectures, including cloud-computing environments. It focuses on handling very large data and tackles the main problem of tiled-based segmentation: artifacts that occur in the segments adjacent to tile borders. Our proposal is to use a specific indexing mechanism and a hierarchical stitching method to suppress segmentation artifacts. The strategy was validated by a series of experiments carried out on a virtual cluster in a commercial cloud-computing infrastructure, over WorldView-2 images of different sizes. The experiments demonstrated that the proposed strategy is able to eliminate tile-based segmentation artifacts. The experimental analysis further showed that the strategy brings significant gains in terms of processing time, particularly for very large images, as more processing units are added to the distributed processing infrastructure.

Image artefacts in swept-source optical coherence tomography angiography

PurposeTo describe optical coherence tomography angiography (OCTA) image artefacts in eyes with and without ocular pathologies.MethodsThe OCTA images of healthy subjects and patients with age-related macular degeneration, diabetic retinopathy and...

Impact of MRI-based Segmentation Artifacts on Amyloid- and FDG-PET Quantitation.

Magnet resonance image (MRI)-based segmentations are widely used for clinical brain research, especially in conjunction with positron-emission-tomography (PET). Although artifacts due to segmentation errors arise commonly, the impact of these artifacts on PET quantitation has not yet been investigated systematically. Therefore, the aim of this study was to assess the effect of segmentation errors on [(18)F]-AV45 and [(18)F]-FDG PET quantitation, with and without correction for partial volume effects (PVE). 119 subjects with both [(18)F]-AV45, and [(18)F]-FDG PET as well as T1-weighted MRI at baseline and at two-year follow-up were selected from the ADNI cohort, and their MRI brain images were segmented using PMOD 3.5. MRIs with segmentation artifacts were masked with the corresponding [(18)F]-FDG PET standard-uptake-value (SUV) images to elucidate and quantify the impact of artifacts on PET analyses for six defined volumes-of-interest (VOI). Artifact volumes were calculated for each VOI, together with error-[%] and root-mean-square-errors (RMSE) in uncorrected and PVE corrected SUV results for the two PET tracers. We also assessed the bias in longitudinal PET data. Artifacts occurred most frequently in the parietal cortex VOI. For [(18)F]-AV45 and [(18)F]-FDG PET, the percentage-errors were dependent on artifact volumes. PVEC SUVs were consequently more distorted than were their uncorrected counterparts. In static and longitudinal assessment, a small subgroup of subjects with large artifacts (≥1500 voxels; ≙5.06 cm³) accounted for much of the PET quantitation bias. Large segmentation artifacts need to be detected and resolved as they considerably bias PET quantitation, especially when PVEC is applied to PET data.

A Two Stage Approach with ICA and Double Density Wavelet Transform for Artifacts Removal in Multichannel EEG Signals

Presence of artifacts in electroencephalographs (EEG) is major hurdles for the precise analysis of spectral behavior. For suppression of ocular artifact in EEG this paper proposes a component based Independent Component Analysis (ICA) model. It involves the generating a set of individual components of given signal followed by rejection of unwanted artifacts. Further this paper presents a novel method with combination of ICA, information sharing and double density wavelet transform to reject the artifacts from the signal. The Independent Component Analysis (ICA) here is used to segment artifact peaks in the signal. Then the Discrete Wavelet Transform is applied for multi-level transfer of signal data until the reception of significant result. The Wavelet ICA suppression not only removes artifacts but also preserves the spectral and coherence properties of brain signals.

Three-dimensional atrial wall thickness maps to inform catheter ablation procedures for atrial fibrillation.

Transmural lesion formation is critical to success in atrial fibrillation ablation and is dependent on left atrial wall thickness (LAWT). Pre- and peri-procedural planning may benefit from LAWT measurements. To calculate the LAWT, the Laplace equation was solved over a finite element mesh of the left atrium derived from the segmented computed tomographic angiography (CTA) dataset. Local LAWT was then calculated from the length of field lines derived from the Laplace solution that spanned the wall from the endocardium or epicardium. The method was validated on an atrium phantom and retrospectively applied to 10 patients who underwent routine coronary CTA for standard clinical indications at our institute. The Laplace wall thickness algorithm was validated on the left atrium phantom. Wall thickness measurements had errors of <0.2 mm for thicknesses of 0.5-5.0 mm that are attributed to image resolution and segmentation artefacts. Left atrial wall thickness measurements were performed on 10 patients. Successful comprehensive LAWT maps were generated in all patients from the coronary CTA images. Mean LAWT measurements ranged from 0.6 to 1.0 mm and showed significant inter and intra patient variability. Left atrial wall thickness can be measured robustly and efficiently across the whole left atrium using a solution of the Laplace equation over a finite element mesh of the left atrium. Further studies are indicated to determine whether the integration of LAWT maps into pre-existing 3D anatomical mapping systems may provide important anatomical information for guiding radiofrequency ablation.

Improvement of Attenuation Correction in Time-of-Flight PET/MR Imaging with a Positron-Emitting Source

Quantitative PET imaging relies on accurate attenuation correction. Recently, there has been growing interest in combining state-of-the-art PET systems with MR imaging in a sequential or fully integrated setup. As CT becomes unavailable for these systems, an alternative approach to the CT-based reconstruction of attenuation coefficients (μ values) at 511 keV must be found. Deriving μ values directly from MR images is difficult because MR signals are related to the proton density and relaxation properties of tissue. Therefore, most research groups focus on segmentation or atlas registration techniques. Although studies have shown that these methods provide viable solutions in particular applications, some major drawbacks limit their use in whole-body PET/MR. Previously, we used an annulus-shaped PET transmission source inside the field of view of a PET scanner to measure attenuation coefficients at 511 keV. In this work, we describe the use of this method in studies of patients with the sequential time-of-flight (TOF) PET/MR scanner installed at the Icahn School of Medicine at Mount Sinai, New York, NY. Five human PET/MR and CT datasets were acquired. The transmission-based attenuation correction method was compared with conventional CT-based attenuation correction and the 3-segment, MR-based attenuation correction available on the TOF PET/MR imaging scanner. The transmission-based method overcame most problems related to the MR-based technique, such as truncation artifacts of the arms, segmentation artifacts in the lungs, and imaging of cortical bone. Additionally, the TOF capabilities of the PET detectors allowed the simultaneous acquisition of transmission and emission data. Compared with the MR-based approach, the transmission-based method provided average improvements in PET quantification of 6.4%, 2.4%, and 18.7% in volumes of interest inside the lung, soft tissue, and bone tissue, respectively. In conclusion, a transmission-based technique with an annulus-shaped transmission source will be more accurate than a conventional MR-based technique for measuring attenuation coefficients at 511 keV in future whole-body PET/MR studies.

Quantitative accuracy of attenuation correction in the Philips Ingenuity TF whole-body PET/MR system: a direct comparison with transmission-based attenuation correction

ObjectiveEvaluation of the quantitative accuracy of MR-based attenuation correction (MRAC) in the Philips Ingenuity TF whole-body PET/MR.Materials and methodsIn 13 patients, PET emission data from the PET/MR were reconstructed using two different methods for attenuation correction. In the first reconstruction, the vendor-provided standard MRAC was used. In the second reconstruction, a coregistered transmission-based attenuation map from a second immediately preceding investigation with a stand-alone Siemens ECAT EXACT HR+ PET scanner was used (TRAC). The two attenuation maps were compared regarding occurrence of segmentation artifacts in the MRAC procedure. Standard uptake values (SUVs) of multiple VOIs (liver, cerebellum, hot focal structures at various locations in the trunk) were compared between both reconstructed data sets. Furthermore, a voxel-wise intensity correlation analysis of both data sets in the lung and trunk was performed.ResultsVOI averaged SUV differences between MRAC and TRAC were as follows (relative differences, mean ± standard deviation): (+12 ± 6) % cerebellum, (−4 ± 9) % liver, (−2 ± 11) % hot focal structures. The fitted slopes of the voxel-wise correlations in the lung and trunk were 0.87 ± 0.17 and 0.95 ± 0.10 with averaged adjusted R 2 values of 0.96 and 0.98, respectively. These figures include two instances with partially erroneous lung segmentation due to artifacts in the underlying MR images.ConclusionThe MR-based attenuation correction implemented on the Philips Ingenuity PET/MR provides reasonable quantitative accuracy. On average, deviations from TRAC-based results are small (on the order of 10 % or below) across the trunk, but due to interindividual variability of the segmentation quality, deviations of more than 20 % can occur. Future improvement of the segmentation quality would help to increase the quantitation accuracy further and to reduce the inter-subject variability.

An Analysis Approach for High-Field fMRI Data from Awake Non-Human Primates

fMRI experiments with awake non-human primates (NHP) have seen a surge of applications in recent years. However, the standard fMRI analysis tools designed for human experiments are not optimal for analysis of NHP fMRI data collected at high fields. There are several reasons for this, including the trial-based nature of NHP experiments, with inter-trial periods being of no interest, and segmentation artefacts and distortions that may result from field changes due to movement. We demonstrate an approach that allows us to address some of these issues consisting of the following steps: 1) Trial-based experimental design. 2) Careful control of subject movement. 3) Computer-assisted selection of trials devoid of artefacts and animal motion. 4) Nonrigid between-trial and rigid within-trial realignment of concatenated data from temporally separated trials and sessions. 5) Linear interpolation of inter-trial intervals and high-pass filtering of temporally continuous data 6) Removal of interpolated data and reconcatenation of datasets before statistical analysis with SPM. We have implemented a software toolbox, fMRI Sandbox (http://code.google.com/p/fmri-sandbox/), for semi-automated application of these processing steps that interfaces with SPM software. Here, we demonstrate that our methodology provides significant improvements for the analysis of awake monkey fMRI data acquired at high-field. The method may also be useful for clinical applications with subjects that are unwilling or unable to remain motionless for the whole duration of a functional scan.

Automated extraction of symmetric microstructure features in serial sectioning images

Serial sectioning methods continue to produce a wealth of image data for quantifying the three-dimensional nature of material microstructures. In this work, we discuss a computational methodology for automated detection and 3D characterization of dendrite cores from images taken from slices of a production turbine blade made of a heat-treated single crystal Ni-based superalloy. The dendrite core locations are detected using an automated segmentation technique that incorporates information over multiple length scales and exploits the four-fold symmetry of the dendrites when viewed down the 〈100〉 growth direction. Additional rules that take advantage of the continuity of the dendrites from slice to slice help to exclude segmentation artifacts and improve dendrite core segmentation. The significance of this technique is that it may be extended to include any symmetric features.

Robust Surface Reconstruction via Laplace-Beltrami Eigen-Projection and Boundary Deformation

In medical shape analysis, a critical problem is reconstructing a smooth surface of correct topology from a binary mask that typically has spurious features due to segmentation artifacts. The challenge is the robust removal of these outliers without affecting the accuracy of other parts of the boundary. In this paper, we propose a novel approach for this problem based on the Laplace-Beltrami (LB) eigen-projection and properly designed boundary deformations. Using the metric distortion during the LB eigen-projection, our method automatically detects the location of outliers and feeds this information to a well-composed and topology-preserving deformation. By iterating between these two steps of outlier detection and boundary deformation, we can robustly filter out the outliers without moving the smooth part of the boundary. The final surface is the eigen-projection of the filtered mask boundary that has the correct topology, desired accuracy and smoothness. In our experiments, we illustrate the robustness of our method on different input masks of the same structure, and compare with the popular SPHARM tool and the topology preserving level set method to show that our method can reconstruct accurate surface representations without introducing artificial oscillations. We also successfully validate our method on a large data set of more than 900 hippocampal masks and demonstrate that the reconstructed surfaces retain volume information accurately.

SU-GG-I-51: Hybrid Computational Phantoms of the Developing Human Fetus

Purpose: Anthropomorphic computational phantoms are a useful tool for assessing ionizing radiationdosesdelivered through medical imaging,radiation therapy, and nuclear medicine procedures. The majority of computational phantom construction efforts have been directed towards pediatric and adult populations, with comparatively little treatment given to the developing fetus. Fetal radiosensitivity combined with growing concerns regarding medical radiationdoses suggests a need for computational phantoms that provide adequate representation of fetal anatomy. This work introduces a new set of fetal computational phantoms to be used for use radiationdosimetry purposes. Method and Materials: High‐resolution MR and CTimages of two preserved fetal specimens (13.5 and 23 weeks) were utilized to construct computational phantoms of the developing fetus. 4.7 T MR images of the younger fetus and clinical 1.5 T MR and CTimages of the older fetus were acquired. Images were segmented using 3D‐Doctor™ and exported to Rhinoceros™ where non‐uniform rational B‐spline (NURBS) surfaces were used to offset minor imaging and segmentation artifacts in the skeleton. Published fetal biometry data were used for age verification and as design criteria for volumetrically adjusting the phantoms to different gestational ages. Results: The thorough representation of the internal fetal anatomy comprising these phantoms is on par with currently available pediatric and adult computational phantoms. Conclusion: High‐resolution MR and CTimages of two preserved fetuses, in conjunction with established segmentation and modeling techniques, have produced two computational phantoms that provide improved representation of the fetal internal anatomy compared to those currently available. In the future, completed fetal phantoms may be combined with an adult female phantom to provide a powerful tool with which to simulate in utero internal and external medical radiation exposures.

SU‐FF‐T‐408: Tissue Inhomogeneities in Monte Carlo Treatment Planning for Proton Therapy

Purpose: To investigate the effect of tissue segmentation and metal streaking artifacts for Monte Carlo (MC) dose calculations in proton therapy. Method and Materials: CT images of a phantom with 9 tissue equivalent inserts were segmented into material and mass density maps using the conventional single‐energy CT and a more accurate dual‐energy (DECT) material extraction. MC dose calculations for a broad 200 MeV proton beam were performed in the exact geometry and in the single‐energy and dual‐energy CT geometries in the MCNPX code. The dose calculation errors for the two tissue segmentation approaches were quantified. MC dose calculations were performed for a 147 MeV proton beam treatment plan of a patient with metal bilateral hip prostheses based on water‐only geometry, on original CT images with severe streaking artifacts and based on artifact corrected images. The effect of the artifacts and their correction on MC dose distribution was evaluated. Results: The materials of three inserts were incorrectly assigned using the conventional approach. The conventional tissue segmentation yielded dose calculation errors below 2%. In both the single‐energy CT and DECT geometry, there was a 0.7 cm shift in the position of the Bragg peak suggesting that density assignment is more important than correct tissue segmentation in proton beam MC dose calculations. The patient dose calculations using CT images with streaking artifacts showed large statistical errors in the artifact corrupted voxels and differences up to 1.5 cm in the 20% and 30% isodose lines due to the artifacts. Conclusions: The shift in the Bragg peak demonstrates the need for careful mass density assignment in proton beam MC dose calculations. The use of DECT tissue segmentation might therefore have only a small added benefit. The patient study shows that a metal artifact correction is necessary for patients with bilateral hip prostheses.

Voting-based simultaneous tracking of multiple video objects

This paper proposes an automatic object tracking method based on both object segmentation and motion estimation for real-time content-oriented video applications. The method focuses on the issues of speed of execution and reliability in the presence of noise, coding artifacts, shadows, occlusion, and object split. Objects are tracked based on the similarity of their features in successive frames. This is done in three steps: feature extraction, object matching, and feature monitoring. In the first step, objects are segmented and their spatial and temporal features are computed. In the second step, using a nonlinear two-stage voting strategy, each object of the previous frame is matched with an object of the current frame creating a unique correspondence. In the third step, object changes, such objects occlusion or split, are monitored and object features are corrected. These new features are then used to update results of previous steps creating module interaction. The contributions in this paper are the real-time two-stage voting strategy, the monitoring of object changes to handle occlusion and object split, and the spatiotemporal adaptation of the tracking parameters. Experiments on indoor and outdoor video shots containing over 6000 frames, including deformable objects, multi-object occlusion, noise, and coding and object segmentation artifacts have demonstrated the reliability and real-time response of the proposed method.