Engineering Papers⌕ Search

SEARCH · Engineering Papers

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 1,045 records · Page 58

Segment 2/3 Hypertrophy is Greater When Right Portal Vein Embolisation is Extended to Segment 4 in Patients with Colorectal Liver Metastases: A Retrospective Cohort Study

Background: In patients with colorectal cancer liver metastases (CRLM), right portal vein embolisation (RPVE) is used to increase the volume of the future remnant liver (FRL) before major hepatic resection. It is not established whether embolisation of segment 4 in addition RPVE (RPVE + 4) induces greater hypertrophy of the FRL. Limitations of prior studies include heterogenous populations and use of hypertrophy metrics sensitive to baseline variables. Methods: From 2010 to 2015, consecutive patients undergoing RPVE or RPVE + 4 for CRLM, who had not undergone prior major hepatic resection and in whom imaging was available, were included in a retrospective study. Data were extracted from hospital electronic records. Volumetric assessments of segments 2–3 were made on cross-sectional imaging before and after embolisation and corrected for standardised liver volume. Results: Ninety-nine patients underwent PVE, and 60 met the inclusion criteria. Thirty-eight patients underwent RPVE, and 22 underwent RPVE + 4. Forty-five patients had undergone median 6 cycles of prior chemotherapy. Eighteen patients had FRL metastases at PVE, and 16 had undergone subsegmental metastasectomy in the FRL. Assessments of the degree of hypertrophy (DH) of segments 2/3 were made at median 35 (interquartile range 30–49) days after PVE. RPVE + 4 resulted in a significantly greater increase in DH than RPVE (7.7 ± 1.8% vs 11.3 ± 2.6%, p = 0.011). No confounding association between baseline variables and the decision to undertake RPVE or RPVE + 4 was identified. Median survival was 2.4 years and was not influenced by segment 4 embolisation. Conclusion: RPVE + 4 results in greater DH of segments 2/3 than RPVE in people with CLRM.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Poster — Thur Eve — 11: Validation of the orthopedic metallic artifact reduction tool for CT simulations at the Ottawa Hospital Cancer Centre

Metallic implants in patients can produce image artifacts in kilovoltage CT simulation images which can introduce noise and inaccuracies in CT number, affecting anatomical segmentation and dose distributions. The commercial orthopedic metal artifact reduction algorithm (O-MAR) (Philips Healthcare System) was recently made available on CT simulation scanners at our institution. This study validated the clinical use of O-MAR by investigating its effects on CT number and dose distributions. O-MAR corrected and uncorrected images were acquired with a Philips Brilliance Big Bore CT simulator of a cylindrical solid water phantom that contained various plugs (including metal) of known density. CT number accuracy was investigated by determining the mean and standard deviation in regions of interest (ROI) within each plug for uncorrected and O-MAR corrected images and comparing with no-metal image values. Dose distributions were calculated using the Monaco treatment planning system. Seven open fields were equally spaced about the phantom around a ROI near the center of the phantom. These were compared to a “correct” dose distribution calculated by overriding electron densities a no-metal phantom image to produce an image containing metal but no artifacts. An overall improvement in CT number and dose distribution accuracy was achieved by applying the O-MAR correction. Mean CT numbers and standard deviations were found to be generally improved. Exceptions included lung equivalent media, which is consistent with vendor specified contraindications. Dose profiles were found to vary by ±4% between uncorrected or O-MAR corrected images with O-MAR producing doses closer to ground truth.

60 APPLIED LIFE SCIENCES↗

Interobserver variability in identification of breast tumors in MRI and its implications for prognostic biomarkers and radiogenomics

Purpose: To assess the interobserver variability of readers when outlining breast tumors in MRI, study the reasons behind the variability, and quantify the effect of the variability on algorithmic imaging features extracted from breast MRI. Methods: Four readers annotated breast tumors from the MRI examinations of 50 patients from one institution using a bounding box to indicate a tumor. All of the annotated tumors were biopsy proven cancers. The similarity of bounding boxes was analyzed using Dice coefficients. An automatic tumor segmentation algorithm was used to segment tumors from the readers’ annotations. The segmented tumors were then compared between readers using Dice coefficients as the similarity metric. Cases showing high interobserver variability (average Dice coefficient <0.8) after segmentation were analyzed by a panel of radiologists to identify the reasons causing the low level of agreement. Furthermore, an imaging feature, quantifying tumor and breast tissue enhancement dynamics, was extracted from each segmented tumor for a patient. Pearson’s correlation coefficients were computed between the features for each pair of readers to assess the effect of the annotation on the feature values. Finally, the authors quantified the extent of variation in feature values caused by each of the individual reasons for low agreement. Results: The average agreement between readers in terms of the overlap (Dice coefficient) of the bounding box was 0.60. Automatic segmentation of tumor improved the average Dice coefficient for 92% of the cases to the average value of 0.77. The mean agreement between readers expressed by the correlation coefficient for the imaging feature was 0.96. Conclusions: There is a moderate variability between readers when identifying the rectangular outline of breast tumors on MRI. This variability is alleviated by the automatic segmentation of the tumors. Furthermore, the moderate interobserver variability in terms of the bounding box does not translate into a considerable variability in terms of assessment of enhancement dynamics. The authors propose some additional ways to further reduce the interobserver variability.

60 APPLIED LIFE SCIENCES↗

Deblending galaxy superpositions with branched generative adversarial networks

Abstract Near-future large galaxy surveys will encounter blended galaxy images at a fraction of up to 50 per cent in the densest regions of the Universe. Current deblending techniques may segment the foreground galaxy while leaving missing pixel intensities in the background galaxy flux. The problem is compounded by the diffuse nature of galaxies in their outer regions, making segmentation significantly more difficult than in traditional object segmentation applications. We propose a novel branched generative adversarial network to deblend overlapping galaxies, where the two branches produce images of the two deblended galaxies. We show that generative models are a powerful engine for deblending given their innate ability to infill missing pixel values occluded by the superposition. We maintain high peak signal-to-noise ratio and structural similarity scores with respect to ground truth images upon deblending. Our model also predicts near-instantaneously, making it a natural choice for the immense quantities of data soon to be created by large surveys such as Large Synoptic Survey Telescope, Euclid, and Wide-Field Infrared Survey Telescope.

Reiman, David M.↗

A Weakly-Supervised, Multitask Deep Learning Framework for Shadow Mitigation in Remote Sensing Imagery

We propose a weakly-supervised, multitask framework for training a convolutional neural network to solve the problem of cloud shadow mitigation given only cloud and shadow masks as labels. The network minimizes the Wasserstein distance between shadows and their proximal sunlit neighborhoods, generating a supervisory signal directly from within the input image. We extract further utility from the shadow mask through multitask learning by introducing an auxiliary task of shadow segmentation. Our approach is advantageous since it performs mitigation in an end-to-end framework which requires only a shadowed image for inference. We apply this process to the Landsat 8 OLI SPARCS validation data set and demonstrate plausible results.

Couwenhoven, Scott↗

Multipath Curved Planar Reformations of Peripheral CT Angiography: Diagnostic Accuracy and Time Efficiency

ObjectivesTo compare diagnostic performance and time efficiency between 3D multipath curved planar reformations (mpCPRs) and axial images of CT angiography for the pre-interventional assessment of peripheral arterial disease (PAD), with digital subtraction angiography as the standard of reference.MethodsForty patients (10 females, mean age 72 years), referred to CTA prior to endovascular treatment of PAD, were prospectively included and underwent peripheral CT angiography. A semiautomated toolbox was used to render mpCPRs. Twenty-one arterial segments were defined in each leg; for each segment, the presence of stenosis > 70% was assessed on mpCPRs and axial images by two readers, independently, with digital subtraction angiography as gold standard.ResultsBoth readers reached lower sensitivity (Reader 1: 91 vs. 94%, p = 0.08; Reader 2: 89 vs. 93%, p = 0.03) but significantly higher specificity (Reader 1: 94 vs. 89%, p < 0.01; Reader 2: 96 vs. 95%, p = 0.01) with mpCPRs than with axial images. Reader 1 achieved significantly higher accuracy with mpCPRs (93 vs. 91%, p = 0.02), and Reader 2 had similar overall accuracy in both evaluations (94 vs. 94%, p = 0.96). Both readers read mpCPRs significantly faster than axial images (Reader 1: 5′45″ based on mpCPRs vs. 7′40″ based on axial images; Reader 2: 4′41″ based on mpCPRs vs. 6′57″ based on axial images; p < 0.01).ConclusionsmpCPRs are a promising 3D reformation technique that facilitates a fast assessment of PAD with high diagnostic accuracy.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Benchmarking image processing techniques for porosity measurement in polymer additive manufacturing: Review and experimental analysis

An image processing workflow is proposed for porosity measurement in polymer additive manufacturing. Various techniques, including global and local thresholding, region growing, and K-means clustering, were applied to microscopic images of carbon fiber reinforced acrylonitrile butadiene styrene (CF-ABS) and benchmarked for their ability to accurately measure porosity. Global methods included Otsu, minimum error, iterative, and entropy-based thresholding, while local methods included Niblack, Bernsen, Sauvola, and Bradley-Roth algorithms. Artificial uneven illumination was introduced to test local adaptive thresholds. Results showed significant differences in porosity values across methods. Otsu, region growing, and K-means clustering excelled under uniform illumination, while Sauvola and Bradley-Roth performed better with uneven illumination. Comparison with X-ray computed tomography (XCT) revealed slightly lower porosity values (2.55 %) than optimized methods (2.73–2.79 %) due to XCT's lower resolution excluding smaller pores. While XCT offers finer pore detection, it limits sample volume and underestimates porosity due to spatial variation. Validation using artificial grayscale images with 5 % porosity confirmed that Otsu, Bradley-Roth, region growing, and Sauvola algorithms produced accurate results. Although tested on a single material system, these methods can be adapted to others with optimization. In conclusion, given XCT's high computational and time costs, this study highlights suitable image processing techniques as cost-effective alternatives for porosity analysis in polymer composites.

Additive manufacturing↗

Deep learning for Alzheimer's disease: Mapping large-scale histological tau protein for neuroimaging biomarker validation

Abnormal tau inclusions are hallmarks of Alzheimer's disease and predictors of clinical decline. Several tau PET tracers are available for neurodegenerative disease research, opening avenues for molecular diagnosis in vivo. However, few have been approved for clinical use. Understanding the neurobiological basis of PET signal validation remains problematic because it requires a large-scale, voxel-to-voxel correlation between PET and (immuno) histological signals. Large dimensionality of whole human brains, tissue deformation impacting co-registration, and computing requirements to process terabytes of information preclude proper validation. We developed a computational pipeline to identify and segment particles of interest in billion-pixel digital pathology images to generate quantitative, 3D density maps. The proposed convolutional neural network for immunohistochemistry samples, IHCNet, is at the pipeline's core. We have successfully processed and immunostained over 500 slides from two whole human brains with three phospho-tau antibodies (AT100, AT8, and MC1), spanning several terabytes of images. Our artificial neural network estimated tau inclusion from brain images, which performs with ROC AUC of 0.87, 0.85, and 0.91 for AT100, AT8, and MC1, respectively. Introspection studies further assessed the ability of our trained model to learn tau-related features. We present an end-to-end pipeline to create terabytes-large 3D tau inclusion density maps co-registered to MRI as a means to facilitate validation of PET tracers.

60 APPLIED LIFE SCIENCES↗

Random lasing in organo-lead halide perovskite microcrystal networks

We report optically pumped random lasing in planar methylammonium lead iodide perovskite microcrystal networks that form spontaneously from spin coating. Low thresholds (<200 μJ/cm{sup 2}) and narrow linewidths (Δλ < 0.5 nm) reflect lasing from closed quasi-modes that result from ballistic waveguiding in linear network segments linked by scattering at the junctions. Spatio-spectral imaging indicates that these quasi-modes extend over lateral length scales >100 μm and spatially overlap with one another, resulting in chaotic pulse-to-pulse intensity fluctuations due to gain competition. These results demonstrate this class of hybrid organic-inorganic perovskite as a platform to study random lasing with well-defined, low-level disorder, and support the potential of these materials for use in semiconductor laser applications.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗

Design and Characterization of an Optically Segmented Single Volume Scatter Camera Module

Here, the optically segmented single volume scatter camera (OS-SVSC) aims to image neutron sources for nuclear nonproliferation applications using the kinematic reconstruction of elastic double-scatter events. We report on the design, construction, and calibration of one module of a new prototype. The module includes 16 EJ-204 organic plastic scintillating bars individually wrapped in Teflon tape, each measuring 0.5cx0.5cmx20cm . The scintillator array is coupled to two custom silicon photomultiplier (SiPM) boards consisting of a 2x8 array of SensL J-Series-60035 SiPMs, which are read out by a custom 16 channel DRS4 based digitizer board. The electrical crosstalk between SiPMs within the electronics chain is measured as 0.76%±0.11% among all 16 channels. We report the detector response of one module including interaction position, time, and energy, using two different optical coupling materials: EJ-560 silicone rubber optical coupling pads and EJ-550 optical coupling grease. We present results in terms of the overall mean and standard deviation of the z -position reconstruction and interaction time resolutions for all 16 bars in the module. We observed the 1σz -position resolution for gamma interactions in the 0.3–0.4 MeVee range to be 2.24 cm ± 1.10 cm and 1.45 cm ± 0.19 cm for silicone optical coupling pad and optical grease, respectively. The observed 1σ interaction time resolution is 265 ps ± 29 ps and 235 ps ± 10 ps for silicone optical coupling pad and optical grease, respectively.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

Skeletal dosimetry in a voxel-based rat phantom for internal exposures to photons and electrons

Purpose: The skeleton makes a significant contribution to the whole body absorbed dose evaluation of rats, since the bone marrow and bone surface in the skeleton express high radiosensitivity and are considered to be important dose-limiting tissues. The bone marrow can be categorized as red bone marrow (RBM) and yellow bone marrow (YBM). It is important to investigate the bone marrow in skeletal dosimetry. Methods: Cryosectional color images of the skeleton of a 156 g rat were segmented into mineral bone (including cortical bone and trabecular bone), RBM, and YBM. These three tissue types were identified at 40 different bone sites and integrated into a previously developed voxel-based rat computational phantom. Photon and electron skeletal absorbed fractions were then calculated using the MCNPX Monte Carlo code. Results: Absorbed fraction (AF) and specific absorbed fraction (SAF) for mineral bone, RBM, and YBM at the 40 different bone sites were established for monoenergetic photon and electron sources placed in 18 organs and seven bone sites. Discrete photon energy was varied from 0.01 to 5.0 MeV in 21 discrete steps, while 21 discrete electron energies were studied, from 0.1 to 10.0 MeV. The trends and values found were consistent with the results of other researchers [M. G. Stabin, T. E. Peterson, G. E. Holburn, and M. A. Emmons, ''Voxel-based mouse and rat models for internal dose calculations,'' J. Nucl. Med. 47, 655-659 (2006)]. S-factors for the radionuclides {sup 169}Er, {sup 143}Pr, {sup 89}Sr, {sup 32}P, and {sup 90}Y, located in 18 organs and seven bone sites for the skeleton, were calculated and are provided in detail. Conclusions: For internal dose calculations, the AF data reveal that the mineral bone in the rat skeletal system is responsible for significant attenuation of gamma rays, especially at low energies. The photon SAF curves of RBM show that, for photon energies greater than 0.6 MeV, there is an increase in secondary photons emitted from the mineral bone as photon energy increases. The SAF values calculated in this study can also be used to determine the absorbed dose to the skeletal system of rats. The S-factors generated here will be useful in preclinical targeted radiotherapy experiments.

61 RADIATION PROTECTION AND DOSIMETRY↗

SU-E-T-545: A MLC-Equipped Robotic Radiosurgery-Radiotherapy Combined System in Treating Hepatic Lesions: Delivery Efficiency as Compared to a Standard Linac for Treating Hepatic Lesions

Purpose: The CyberKnife (CK) M6 Series introduced a mulitleaf collimator (MLC) beam for extending its capability to the conventional radiotherapy. This work is to investigate delivery efficiency of this system as compared to a standard Varian linac when treating hepatic lesions. Methods: Nine previously treated patients were divided into three groups with three patients in each. Group one: fractionated radiotherapy; Group two: SBRT-like treatments and Group three: fractionated radiotherapy targeting two PTVs. The clinically used plans were generated with the Eclipse treatment planning system (TPS). We re-planned these cases using a Mulitplan (MP) TPS for the CK M6 and normalized to the same PTV dose coverage. CK factors (CF) (defined as modulation scaling factor in this work), number of nodes (NN), number of MLC segments (NS) and beam delivery time (BT) with an estimated image interval of 60 seconds, were used for evaluation of delivery efficiency. Results: Generated plans from the MP and Eclipse TPS demonstrated the similar quality in terms of PTV confomality index, minimum and maximum PTV doses, and doses received by critical structures. Group one: CF ranged from 8.1 to 8.7, NN from 30 to 40, NS from 120 to 155 and BT from 20 to 23 minutes; group two: CF from 4.7 to 8.5, NN from 15 to 19, NS from 82 to 141 and BT from 18 to 24 minutes; and group three: CF from 7.9 to 10, NN from 47 to 49, NS from 110 to 113 and BT from 20 to 22 minutes. Conclusions: Delivery time is longer for the CK M6 than for the Varian linac (7.8 to 13.7 minutes). Further investigation will be necessary to determine if a PTV reduction from the tracking feature will shorten the delivery time without decreasing plan quality.

60 APPLIED LIFE SCIENCES↗

SU-F-BRD-05: Robustness of Dose Painting by Numbers in Proton Therapy

Purpose: Proton range uncertainties may cause important dose perturbations within the target volume, especially when steep dose gradients are present as in dose painting. The aim of this study is to assess the robustness against setup and range errors for high heterogeneous dose prescriptions (i.e., dose painting by numbers), delivered by proton pencil beam scanning. Methods: An automatic workflow, based on MATLAB functions, was implemented through scripting in RayStation (RaySearch Laboratories). It performs a gradient-based segmentation of the dose painting volume from 18FDG-PET images (GTVPET), and calculates the dose prescription as a linear function of the FDG-uptake value on each voxel. The workflow was applied to two patients with head and neck cancer. Robustness against setup and range errors of the conventional PTV margin strategy (prescription dilated by 2.5 mm) versus CTV-based (minimax) robust optimization (2.5 mm setup, 3% range error) was assessed by comparing the prescription with the planned dose for a set of error scenarios. Results: In order to ensure dose coverage above 95% of the prescribed dose in more than 95% of the GTVPET voxels while compensating for the uncertainties, the plans with a PTV generated a high overdose. For the nominal case, up to 35% of the GTVPET received doses 5% beyond prescription. For the worst of the evaluated error scenarios, the volume with 5% overdose increased to 50%. In contrast, for CTV-based plans this 5% overdose was present only in a small fraction of the GTVPET, which ranged from 7% in the nominal case to 15% in the worst of the evaluated scenarios. Conclusion: The use of a PTV leads to non-robust dose distributions with excessive overdose in the painted volume. In contrast, robust optimization yields robust dose distributions with limited overdose. RaySearch Laboratories is sincerely acknowledged for providing us with RayStation treatment planning system and for the support provided.

60 APPLIED LIFE SCIENCES↗

Tracing Actin Filament Bundles in Three-Dimensional Electron Tomography Density Maps of Hair Cell Stereocilia

Cryo-electron tomography (cryo-ET) is a powerful method of visualizing the three-dimensional organization of supramolecular complexes, such as the cytoskeleton, in their native cell and tissue contexts. Due to its minimal electron dose and reconstruction artifacts arising from the missing wedge during data collection, cryo-ET typically results in noisy density maps that display anisotropic XY versus Z resolution. Molecular crowding further exacerbates the challenge of automatically detecting supramolecular complexes, such as the actin bundle in hair cell stereocilia. Stereocilia are pivotal to the mechanoelectrical transduction process in inner ear sensory epithelial hair cells. Given the complexity and dense arrangement of actin bundles, traditional approaches to filament detection and tracing have failed in these cases. In this study, we introduce BundleTrac, an effective method to trace hundreds of filaments in a bundle. A comparison between BundleTrac and manually tracing the actin filaments in a stereocilium showed that BundleTrac accurately built 326 of 330 filaments (98.8%), with an overall cross-distance of 1.3 voxels for the 330 filaments. BundleTrac is an effective semi-automatic modeling approach in which a seed point is provided for each filament and the rest of the filament is computationally identified. We also demonstrate the potential of a denoising method that uses a polynomial regression to address the resolution and high-noise anisotropic environment of the density map.

stereocilia↗

Quantitative Insight to Fission Gas Pores Distribution in Irradiated Annular U-10Zr Metallic Fuel Using Machine Learning

Metallic fuels, particularly U-10Zr and its performance in reactor irradiation conditions, have been thoroughly investigated and are a promising candidate for next-generation sodium-cooled fast spectrum nuclear reactors. Irradiation in reactors can lead to the formation of fission gas and increased pore formation which can significantly impact fuel performance. Due to the large number of pores and various phases formed in metallic fuel during irradiation, a quantitative description of fission gas pores as a function of irradiation conditions is not yet available, undermining the fidelity of fuel performance modeling to support fuel qualification. It has been difficult to clearly detect pore boundaries and distinguish matrix phases from fission gas pores using optical microscopy by using simple threshold methods working with low magnification images. The pre-trained deep learning model for fission gas pore detection was applied to ~10,260 high magnification scanning electron microscopy images. The model increased the accuracy of fission gas pore segmentation to obtain statistical features, which cannot be processed manually. A pre-trained decision tree model was used to classify pores as isolated or connected pores, providing new insight into the correlation between the movement of lanthanides, solid fission products, and the radial temperature gradient developed in fuel irradiation conditions. This paper emphasizes the potential that artificial intelligence-based machine learning models have to accelerate qualification and support nuclear fuel development.

36 MATERIALS SCIENCE↗

Ultrasensitive measurement of microcantilever displacement below the shot-noise limit

The displacement of micro-electro-mechanical-systems (MEMs) cantilevers is used to measure a variety of phe- nomena in devices ranging from force microscopes for single spin detection[1] to biochemical sensors[2] to un- cooled thermal imaging systems[3]. The displacement readout is often performed optically with segmented de- tectors or interference measurements. Until recently, var- ious noise sources have limited the minimum detectable displacement in MEMs systems, but it is now possible to minimize all other sources[4] so that the noise level of the coherent light eld, called the shot noise limit (SNL), becomes the dominant source. Light sources dis- playing quantum-enhanced statistics below this limit are available[5, 6], with applications in gravitational wave astronomy[7] and bioimaging[8], but direct displacement measurements of MEMS cantilevers below the SNL have been impossible until now. Here, we demonstrate the rst direct measurement of a MEMs cantilever displace- ment with sub-SNL sensitivity, thus enabling ultratrace sensing, imaging, and microscopy applications. By com- bining multi-spatial-mode quantum light sources with a simple dierential measurement, we show that sub-SNL MEMs displacement sensitivity is highly accessible com- pared to previous eorts that measured the displacement of macroscopic mirrors with very distinct spatial struc- tures crafted with multiple optical parametric ampliers and locking loops[9]. We apply this technique to a com- mercially available microcantilever in order to detect dis- placements 60% below the SNL at frequencies where the microcantilever is shot-noise-limited. These results sup- port a new class of quantum MEMS sensor whose ulti- mate signal to noise ratio is determined by the correla- tions possible in quantum optics systems.

47 OTHER INSTRUMENTATION↗

SU-F-J-95: Impact of Shape Complexity On the Accuracy of Gradient-Based PET Volume Delineation

Purpose: Explore correlation of tumor complexity shape with PET target volume accuracy when delineated with gradient-based segmentation tool. Methods: A total of 24 clinically realistic digital PET Monte Carlo (MC) phantoms of NSCLC were used in the study. The phantom simulated 29 thoracic lesions (lung primary and mediastinal lymph nodes) of varying size, shape, location, and {sup 18}F-FDG activity. A program was developed to calculate a curvature vector along the outline and the standard deviation of this vector was used as a metric to quantify a shape’s “complexity score”. This complexity score was calculated for standard geometric shapes and MC-generated target volumes in PET phantom images. All lesions were contoured using a commercially available gradient-based segmentation tool and the differences in volume from the MC-generated volumes were calculated as the measure of the accuracy of segmentation. Results: The average absolute percent difference in volumes between the MC-volumes and gradient-based volumes was 11% (0.4%–48.4%). The complexity score showed strong correlation with standard geometric shapes. However, no relationship was found between the complexity score and the accuracy of segmentation by gradient-based tool on MC simulated tumors (R{sup 2} = 0.156). When the lesions were grouped into primary lung lesions and mediastinal/mediastinal adjacent lesions, the average absolute percent difference in volumes were 6% and 29%, respectively. The former group is more isolated and the latter is more surround by tissues with relatively high SUV background. Conclusion: The complexity shape of NSCLC lesions has little effect on the accuracy of the gradient-based segmentation method and thus is not a good predictor of uncertainty in target volume delineation. Location of lesion within a relatively high SUV background may play a more significant role in the accuracy of gradient-based segmentation.

60 APPLIED LIFE SCIENCES↗