Search NASA⌕ Search

SEARCH · Search NASA

Results for “segmentation averaging”

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 217 records · Page 12

Automated temporal tracking and segmentation of lymphoma on serial CT examinations

Purpose: It is challenging to reproducibly measure and compare cancer lesions on numerous follow-up studies; the process is time-consuming and error-prone. In this paper, we show a method to automatically and reproducibly identify and segment abnormal lymph nodes in serial computed tomography (CT) exams. Methods: Our method leverages initial identification of enlarged (abnormal) lymph nodes in the baseline scan. We then identify an approximate region for the node in the follow-up scans using nonrigid image registration. The baseline scan is also used to locate regions of normal, non-nodal tissue surrounding the lymph node and to map them onto the follow-up scans, in order to reduce the search space to locate the lymph node on the follow-up scans. Adaptive region-growing and clustering algorithms are then used to obtain the final contours for segmentation. We applied our method to 24 distinct enlarged lymph nodes at multiple time points from 14 patients. The scan at the earlier time point was used as the baseline scan to be used in evaluating the follow-up scan, resulting in 70 total test cases (e.g., a series of scans obtained at 4 time points results in 3 test cases). For each of the 70 cases, a ''reference standard'' was obtained by manual segmentation by a radiologist. Assessment according to response evaluation criteria in solid tumors (RECIST) using our method agreed with RECIST assessments made using the reference standard segmentations in all test cases, and by calculating node overlap ratio and Hausdorff distance between the computer and radiologist-generated contours. Results: Compared to the reference standard, our method made the correct RECIST assessment for all 70 cases. The average overlap ratio was 80.7 {+-} 9.7% s.d., and the average Hausdorff distance was 3.2 {+-} 1.8 mm s.d. The concordance correlation between automated and manual segmentations was 0.978 (95% confidence interval 0.962, 0.984). The 100% agreement in our sample between our method and the standard with regard to RECIST classification suggests that the true disagreement rate is no more than 6%. Conclusions: Our automated lymph node segmentation method achieves excellent overall segmentation performance and provides equivalent RECIST assessment. It potentially will be useful to streamline and improve cancer lesion measurement and tracking and to improve assessment of cancer treatment response.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Automated Shift Detection in Sensor-Based PV Power and Irradiance Time Series: Preprint

PV power and irradiance sensor-based measurements are prone to error, resulting in issues such as abrupt time series data shifts. These shifts, which are usually unintentional, may be caused by software or hardware configuration changes on a PV system, and do not reflect an actual change in overall system performance. Locating these shifts and segmenting the associated time series aids in more accurate future PV analysis. In this research, an offline changepoint detection (CPD) algorithm that automatically detects these abrupt data shifts in sensor-based time series is introduced. Data shift periods in 101 daily PV power and irradiance time series were labeled manually by two solar experts. These data streams represent sensor-based measurements, and display a variety of data shift behaviors. A changepoint detection algorithm was tuned using the 101 labeled data streams, with each model configuration's ability to detect labeled changepoints benchmarked using metrics such as F1-score, recall, and Rand Index. Best performing models on seasonality-corrected data streams include the Pruned Exact Linear (PELT) method, the Binary Segmentation method, and the Bottom-Up method, all scoring an average F1-score of 0.76 or greater at detecting labeled changepoints within a 30-day window for the labeled data sets. To promote further research in this space, we are releasing the labeled data shift sets on U.S. Department of Energy's (DOE) DuraMAT Data Hub, and the associated algorithm in the Python PVAnalytics package.

changepoint detection↗

Synthetic aperture radar imaging from an inclined geosynchronous orbit

Images of earth can be produced with an assumed synthetic aperture radar (SAR) on a satellite platform undergoing a nutating relative motion from geosynchronous altitude. From a 50 deg inclined circular orbit, the contiguous United States can be imaged in about 3 h of segmented operation at 100-m resolution with 4-azimuth-look averaging. The 2450-MHz transmitter radiates 1312 W of average power from a steerable 15-m-diam antenna. The SAR can image daily an area bounded longitudinally and latitudinally.

Tomiyasu, K.↗

SAR Interferometry as a Tool for Monitoring Coastal Changes in the Nile River Delta of Egypt

The Nile River Delta is experiencing rapid rates of coastal change. The rate of both coastal retreat and accretion in the Eastern Nile Delta requires regular, accurate detection and measurement. Current techniques used to monitor coastal changes in the delta are point measurements and, thus, they provide a spatially limited view of the ongoing coastal changes. SAR interferometry can provide measurements of subtle coastal change at a significantly improved spatial resolution and over large areas (100 sq km). Using data provided by the ERS-1&2 satellites, monitoring can be accomplished as frequently as every 35 days when needed. Radar interferometry is employed in this study to detect segments of erosion and accretion during the 1993-2000 period. The average rates of erosion and accretion in the Eastern Nile Delta are measured to be -11.64 m/yr and +5.12 m/yr, respectively. The results of this interferometric study can be used effectively for coastal zone management and integrated sustainable development for the Nile River Delta.

Aly, Mohamed H.↗

Evaluation of the Terminal Sequencing and Spacing System for Performance Based Navigation Arrivals

NASA has developed the Terminal Sequencing and Spacing (TSS) system, a suite of advanced arrival management technologies combining timebased scheduling and controller precision spacing tools. TSS is a ground-based controller automation tool that facilitates sequencing and merging arrivals that have both current standard ATC routes and terminal Performance-Based Navigation (PBN) routes, especially during highly congested demand periods. In collaboration with the FAA and MITRE's Center for Advanced Aviation System Development (CAASD), TSS system performance was evaluated in human-in-the-loop (HITL) simulations with currently active controllers as participants. Traffic scenarios had mixed Area Navigation (RNAV) and Required Navigation Performance (RNP) equipage, where the more advanced RNP-equipped aircraft had preferential treatment with a shorter approach option. Simulation results indicate the TSS system achieved benefits by enabling PBN, while maintaining high throughput rates-10% above baseline demand levels. Flight path predictability improved, where path deviation was reduced by 2 NM on average and variance in the downwind leg length was 75% less. Arrivals flew more fuel-efficient descents for longer, spending an average of 39 seconds less in step-down level altitude segments. Self-reported controller workload was reduced, with statistically significant differences at the p less than 0.01 level. The RNP-equipped arrivals were also able to more frequently capitalize on the benefits of being "Best-Equipped, Best- Served" (BEBS), where less vectoring was needed and nearly all RNP approaches were conducted without interruption.

terminal sequencing↗

Automated Shift Detection in Sensor-Based PV Power and Irradiance Time Series

PV power and irradiance sensor-based measurements are prone to error, resulting in issues such as abrupt time series data shifts. These shifts, which are usually unintentional, may be caused by software or hardware configuration changes on a PV system, and do not reflect an actual change in overall system performance. Locating these shifts and segmenting the associated time series aids in more accurate future PV analysis. In this research, an offline changepoint detection (CPD) algorithm that automatically detects these abrupt data shifts in sensor-based time series is introduced. Data shift periods in 101 daily PV power and irradiance time series were labeled manually by two solar experts. These data streams represent sensor-based measurements, and display a variety of data shift behaviors. A changepoint detection algorithm was tuned using the 101 labeled data streams, with each model configuration's ability to detect labeled changepoints benchmarked using metrics such as F1-score, recall, and Rand Index. Best performing models on seasonality-corrected data streams include the Pruned Exact Linear (PELT) method, the Binary Segmentation method, and the Bottom-Up method, all scoring an average F1-score of 0.76 or greater at detecting labeled changepoints within a 30-day window for the labeled data sets. To promote further research in this space, we are releasing the labeled data shift sets on U.S. Department of Energy's (DOE) DuraMAT Data Hub, and the associated algorithm in the Python PVAnalytics package.

changepoint detection↗

Automated Shift Detection in Sensor-Based PV Power and Irradiance Time Series

PV power and irradiance sensor-based measurements are prone to error, resulting in issues such as time series data shifts. In this research, a changepoint detection (CPD) algorithm that automatically detects data shifts in sensor-based time series is introduced. Data shift periods in 101 daily PV power and irradiance time series were labeled manually by two solar experts. These data streams represent sensor-based measurements, and display a variety of data shift behaviors. A changepoint detection algorithm was tuned using the 101 labeled data streams, with each model configuration's ability to detect labeled changepoints benchmarked using metrics such as F1-score, recall, and Rand Index. Best performing models on seasonality-corrected data streams include the Pruned Exact Linear (PELT) method, the Binary Segmentation method, and the Bottom-Up method, all scoring an average F1-score of 0.76 or greater at detecting labeled changepoints within a 30-day window across the labeled data sets. Pending approval, we plan to release the labeled data sets for this research on NREL's DuraMAT Data Hub, and the associated algorithm in the Python PVAnalytics package. By supplying the training sets and algorithm, we hope to encourage further development in this research space.

data shift↗

Coadding Techniques for Image-based Wavefront Sensing for Segmented-mirror Telescopes

Image-based wavefront sensing algorithms are being used to characterize optical performance for a variety of current and planned astronomical telescopes. Phase retrieval recovers the optical wavefront that correlates to a series of diversity-defocused point-spread functions (PSFs), where multiple frames can be acquired at each defocus setting. Multiple frames of data can be coadded in different ways; two extremes are in "image-plane space," to average the frames for each defocused PSF and use phase retrieval once on the averaged images, or in "pupil-plane space," to use phase retrieval on every set of PSFs individually and average the resulting wavefronts. The choice of coadd methodology is particularly noteworthy for segmented-mirror telescopes that are subject to noise that causes uncorrelated motions between groups of segments. Using data collected on and simulations of the James Webb Space Telescope Testbed Telescope (TBT) commissioned at Ball Aerospace, we show how different sources of noise (uncorrelated segment jitter, turbulence, and common-mode noise) and different parts of the optical wavefront, segment and global aberrations, contribute to choosing the coadd method. Of particular interest, segment piston is more accurately recovered in "image-plane space" coadding, while segment tip/tilt is recovered in "pupil-plane space" coadding.

Smith, Scott↗

No evidence of a threshold in traffic volume affecting road-kill mortality at a large spatio-temporal scale

Previous studies have found that the relationship between wildlife road mortality and traffic volume follows a threshold effect on low traffic volume roads. We aimed at evaluating the response of several species to increasing traffic intensity on highways over a large geographic area and temporal period. We used data of four terrestrial vertebrate species with different biological and ecological features known by their high road-kill rates: the barn owl (Tyto alba), hedgehog (Erinaceus europaeus), red fox (Vulpes vulpes) and European rabbit (Oryctolagus cuniculus). Additionally, we checked whether road-kill likelihood varies when traffic patterns depart from the average. We used annual average daily traffic (AADT) and road-kill records observed along 1000 km of highways in Portugal over seven consecutive years (2003–2009). We fitted candidate models using Generalized Linear Models with a binomial distribution through a sample unit of 1 km segments to describe the effect of traffic on the probability of finding at least one victim in each segment during the study. We also assigned for each road-kill record the traffic of that day and the AADT on that year to test for differences using Paired Student's t-test. Mortality risk declined significantly with traffic volume but varied among species: the probability of finding road-killed red foxes and rabbits occurs up to moderate traffic volumes (< 20,000 AADT) whereas barn owls and hedgehogs occurred up to higher traffic volumes (40,000 AADT). Perception of risk may explain differences in responses towards high traffic highway segments. Road-kill rates did not vary significantly when traffic intensity departed from the average. In summary, we did not find evidence of traffic thresholds for the analysed species and traffic intensities. We suggest mitigation measures to reduce mortality be applied in particular on low traffic roads (< 5000 AADT) while additional measures to reduce barrier effects should take into account species-specific behavioural traits. - Highlights: • Traffic and road-kills were analysed along 1000 km of highways over seven years. • Mortality risk declined significantly with traffic volume. • Perception of risk may explain different responses towards high traffic sections. • Reducing barrier effects should take into account species behavioural traits.

54 ENVIRONMENTAL SCIENCES↗

SU-C-BRA-02: Gradient Based Method of Target Delineation On PET/MR Image of Head and Neck Cancer Patients

Purpose: Validate the consistency of a gradient-based segmentation tool to facilitate accurate delineation of PET/CT-based GTVs in head and neck cancers by comparing against hybrid PET/MR-derived GTV contours. Materials and Methods: A total of 18 head and neck target volumes (10 primary and 8 nodal) were retrospectively contoured using a gradient-based segmentation tool by two observers. Each observer independently contoured each target five times. Inter-observer variability was evaluated via absolute percent differences. Intra-observer variability was examined by percentage uncertainty. All target volumes were also contoured using the SUV percent threshold method. The thresholds were explored case by case so its derived volume matched with the gradient-based volume. Dice similarity coefficients (DSC) were calculated to determine overlap of PET/CT GTVs and PET/MR GTVs. Results: The Levene’s test showed there was no statistically significant difference of the variances between the observer’s gradient-derived contours. However, the absolute difference between the observer’s volumes was 10.83%, with a range from 0.39% up to 42.89%. PET-avid regions with qualitatively non-uniform shapes and intensity levels had a higher absolute percent difference near 25%, while regions with uniform shapes and intensity levels had an absolute percent difference of 2% between observers. The average percentage uncertainty between observers was 4.83% and 7%. As the volume of the gradient-derived contours increased, the SUV threshold percent needed to match the volume decreased. Dice coefficients showed good agreement of the PET/CT and PET/MR GTVs with an average DSC value across all volumes at 0.69. Conclusion: Gradient-based segmentation of PET volume showed good consistency in general but can vary considerably for non-uniform target shapes and intensity levels. PET/CT-derived GTV contours stemming from the gradient-based tool show good agreement with the anatomically and metabolically more accurate PET/MR-derived GTV contours, but tumor delineation accuracy can be further improved with the use PET/MR.

60 APPLIED LIFE SCIENCES↗

A framework to measure myocardial extracellular volume fraction using dual-phase low dose CT images

Purpose: Myocardial extracellular volume fraction (ECVF) is a surrogate imaging biomarker of diffuse myocardial fibrosis, a hallmark of pathologic ventricular remodeling. Low dose cardiac CT is emerging as a promising modality to detect diffuse interstitial myocardial fibrosis due to its fast acquisition and low radiation; however, the insufficient contrast in the low dose CT images poses great challenge to measure ECVF from the image. Methods: To deal with this difficulty, the authors present a complete ECVF measurement framework including a point-guided myocardial modeling, a deformable model-based myocardium segmentation, nonrigid registration of pre- and post-CT, and ECVF calculation. Results: The proposed method was evaluated on 20 patients by two observers. Compared to the manually delineated reference segmentations, the accuracy of our segmentation in terms of true positive volume fraction (TPVF), false positive volume fraction (FPVF), and average surface distance (ASD), were 92.18% ± 3.52%, 0.31% ± 0.10%, 0.69 ± 0.14 mm, respectively. The interobserver variability measured by concordance correlation coefficient regarding TPVF, FPVF, and ASD were 0.95, 0.90, 0.94, respectively, demonstrating excellent agreement. Bland-Altman method showed 95% limits of agreement between ECVF at CT and ECVF at MR. Conclusions: The proposed framework demonstrates its efficiency, accuracy, and noninvasiveness in ECVF measurement and dramatically advances the ECVF at cardiac CT toward its clinical use.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

An Efficient Pipeline for Abdomen Segmentation in CT Images

Computed tomography (CT) scans usually include some disadvantages due to the nature of the imaging procedure, and these handicaps prevent accurate abdomen segmentation. Discontinuous abdomen edges, bed section of CT, patient information, closeness between the edges of the abdomen and CT, poor contrast, and a narrow histogram can be regarded as the most important handicaps that occur in abdominal CT scans. Currently, one or more handicaps can arise and prevent technicians obtaining abdomen images through simple segmentation techniques. In other words, CT scans can include the bed section of CT, a patient’s diagnostic information, low-quality abdomen edges, low-level contrast, and narrow histogram, all in one scan. These phenomena constitute a challenge, and an efficient pipeline that is unaffected by handicaps is required. In addition, analysis such as segmentation, feature selection, and classification has meaning for a real-time diagnosis system in cases where the abdomen section is directly used with a specific size. A statistical pipeline is designed in this study that is unaffected by the handicaps mentioned above. Intensity-based approaches, morphological processes, and histogram-based procedures are utilized to design an efficient structure. Performance evaluation is realized in experiments on 58 CT images (16 training, 16 test, and 26 validation) that include the abdomen and one or more disadvantage(s). The first part of the data (16 training images) is used to detect the pipeline’s optimum parameters, while the second and third parts are utilized to evaluate and to confirm the segmentation performance. The segmentation results are presented as the means of six performance metrics. Thus, the proposed method achieves remarkable average rates for training/test/validation of 98.95/99.36/99.57% (jaccard), 99.47/99.67/99.79% (dice), 100/99.91/99.91% (sensitivity), 98.47/99.23/99.85% (specificity), 99.38/99.63/99.87% (classification accuracy), and 98.98/99.45/99.66% (precision). In summary, a statistical pipeline performing the task of abdomen segmentation is achieved that is not affected by the disadvantages, and the most detailed abdomen segmentation study is performed for the use before organ and tumor segmentation, feature extraction, and classification.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Poster — Thur Eve — 09: Evaluation of electrical impedance and computed tomography fusion algorithms using an anthropomorphic phantom

Integration of biological conductivity information provided by Electrical Impedance Tomography (EIT) with anatomical information provided by Computed Tomography (CT) imaging could improve the ability to characterize tissues in clinical applications. In this paper, we report results of our study which compared the fusion of EIT with CT using three different image fusion algorithms, namely: weighted averaging, wavelet fusion, and ROI indexing. The ROI indexing method of fusion involves segmenting the regions of interest from the CT image and replacing the pixels with the pixels of the EIT image. The three algorithms were applied to a CT and EIT image of an anthropomorphic phantom, constructed out of five acrylic contrast targets with varying diameter embedded in a base of gelatin bolus. The imaging performance was assessed using Detectability and Structural Similarity Index Measure (SSIM). Wavelet fusion and ROI-indexing resulted in lower Detectability (by 35% and 47%, respectively) yet higher SSIM (by 66% and 73%, respectively) than weighted averaging. Our results suggest that wavelet fusion and ROI-indexing yielded more consistent and optimal fusion performance than weighted averaging.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Interpolating Low Time-Resolution Forecast Data

Methodology that interpolates low time-resolution data (e.g., hourly) to high time-resolution (e.g., minutely) with variability patterns extracted from historical records. Magnitude of the variability inserted into the low time resolution data can be adjusted according to the installed capacity represented by the low time-resolution data compared to that by historical records. This approach enables detailed analysis of the impacts from wind and solar on power system intra-hour operations and balancing reserve requirements even with only hourly data. It also allows convenient creation of high resolution wind or solar generation data with various degree of variability to investigate their operational impacts. The methodology comprises of the following steps: 1. Smooth the historical data (set A) with an appropriate window length l to get its trend (set B); l can be a fraction of an hour (e.g., 15 minutes) or longer than an hour, of which the length of the variability patterns will be; 2. Extract the variable component (set C) of historical data by subtracting the smooth trend from it, i.e. set C = set A - set B 3. For each window length l of the variable component data set, find the average value x (will call it base component) of the corresponding window of the historical data set; 4. Define a series of segments (set D) that the values of data will be grouped into, e.g. (0, 0.1), (0.1, 0.2), ..., (0.9, 1.0) after normalization; Link each variability pattern to a data segment based on its corresponding base component x; after this step, each data segment should be linked to multiple variability patterns after this step; 5. Use spline function to interpolate the low time-resolution forecast data (set E) to become a high time-resolution smooth curve (set F); 6. Based on the window length l , calculate the average value y in each window length of set F; find the data segment that y belongs to; then randomly select one of the variability patterns linked to this data segment; 7. Add the selected variability pattern in each window to set F to get the high time-resolution data (set G) with variability observed from historical data set A; a coefficient k can be multiplied to the selected variability pattern to reflect the potential difference in variability between forecast set E and historical data set A caused by installation capacity and other factors. 8. After set G is constructed, variability statistics can be checked for set A and set G for comparison. Adjust the coefficient k if necessary to achieve desired variability in the reconstructed data.

Lu, Shuai↗

Technology Infusion of CodeSonar into the Space Network Ground Segment (RII07): Software Assurance Symposium Technical Summary

Presents a source code analysis tool (CodeSonar) for use in the Space Network Ground Segment. The Space Network requires 99.9% proficiency and 97.0% availability of systems. Software has historically accounted for an annual average of 28% of the Space Network loss of availability and proficiency. CSCI A and CSCI B account for 42% of the previous eight months of software data loss. The technology infusion of CodeSonar into the Space Network Ground segment is meant to aid in determining the impact of the technology on the project both in the expenditure of effort and the technical results of the technology. Running a CodeSonar analysis and performing a preliminary review of the results averaged 3.5 minutes per finding (approximately 20 hours total). An additional 40 hours is estimated to analyze the 37 findings deemed too complex for the initial review. Using CodeSonar's tools to suppress known non-problems, delta tool runs will not repeat findings that have been marked as non-problems, further reducing the time needed for review. The 'non-interesting' finding rate of 70% is a large number, but filtering, search, and detailed contextual features of CodeSonar reduce the time per finding. Integration of the tool into the build process may also provide further savings by preventing developers from having to configure and operate the tool separately. These preliminary results show the tool to be easy to use and incorporate into the engineering process. These findings also provide significant potential improvements in proficiency and availability on the part of the software. As time-to-fix data become available a better cost trade can be made on person hours saved versus tool cost. Selective factors may be necessary to determine where best to apply CodeSonar to balance cost and benefits.

Benson, Markland J.↗

Formation of edge misfit dislocations in Ge{sub x}Si{sub 1-x}(x{approx}0.4-0.8) films grown on misoriented (001){yields}(111) Si substrates: Features before and after film annealing

The dislocation structure at the initial stage of relaxation of Ge{sub x}Si{sub 1-x} films (x{approx}0.4-0.8) grown on Si(001) substrates tilted at 6 deg. to the nearest (111) plane is studied. One of the directions along which edge dislocations are formed is no longer in the plane of the interface but crosses the latter at the angle of miscut. Therefore, long edge misfit dislocations (MDs) cannot exist in the direction of substrate misorientation, because their ends move away from the interface during dislocation propagation. Two different mechanisms of formation of short segments of edge (90 deg. ) MDs in the direction of substrate misorientation are found. The first mechanism is the correlated nucleation of complementary 60 deg. dislocation half-loops manifested in the form of the so-called Y-center consisting of a short segment of the 90 deg. MD and 60 deg. MDs diverging from this segment in form of two rays in the miscut direction. In the second mechanism, the 90 deg. MD segment is formed owing to intersection of already existing complementary 60 deg. MDs slipping in mirrorlike inclined planes {l_brace}111{r_brace}. Samples annealed at higher temperatures contain MD segments, which lose their contrast on one of the TEM images taken to detect edge dislocation segments in the two-wave diffraction mode but do not coincide with the direction [110] (they pass along the intersections of the mirrorlike inclined planes {l_brace}111{r_brace} with the surface of the misoriented Si substrate). To explain this phenomenon, a model is proposed, where one of the pair of complementary 60 deg. MDs propagating at an angle to each other continuously approaches the other by the cross-slip mechanism. Formation of a chain of edge MD segments connected by a pair of 60 deg. MDs becomes possible. The averaged direction of such a dislocation structure coincides with the direction of one of the 60 deg. MDs.

36 MATERIALS SCIENCE↗

DeepAndes: A Self-Supervised Vision Foundation Model for Multispectral Remote Sensing Imagery of the Andes

By mapping sites at large scales usingremotely sensed data, archaeologists can generate unique insights into long-term demographic trends, interregional social networks, and human adaptations in the past. Remote sensing surveys complement field-based approaches, and their reach can be especially great when combined with deep learning and computer vision techniques. However, conventional supervised deep learning methods face challenges in annotating fine-grained archaeological features at scale. In addition, while recent vision foundation models have shown remarkable success in learning large-scale remote sensing data with minimal annotations, most off-the-shelf solutions are designed for RGB images rather than multispectral satellite imagery, such as the eight-band data used in our study. In this article, we introduce DeepAndes, a transformer-based vision foundation model trained on three million multispectral satellite images, specifically tailored for Andean archaeology. DeepAndes incorporates a customized DINOv2 self-supervised learning algorithm optimized for eight-band multispectral imagery, marking the first foundation model designed explicitly for the Andes region. We evaluate its image understanding performance through imbalanced image classification, image instance retrieval, and pixel-level semantic segmentation tasks. Our experiments show that DeepAndes achieves superior F1 scores, mean average precision, and Dice scores in few-shot learning scenarios, significantly outperforming models trained from scratch or pretrained on smaller datasets. This underscores the effectiveness of large-scale self-supervised pretraining in archaeological remote sensing.

Guo, Junlin [Vanderbilt Univ., Nashville, TN (Unit↗

SU-C-207B-03: A Geometrical Constrained Chan-Vese Based Tumor Segmentation Scheme for PET

Purpose: Accurate segmentation of tumor in PET is challenging when part of tumor is connected with normal organs/tissues with no difference in intensity. Conventional segmentation methods, such as thresholding or region growing, cannot generate satisfactory results in this case. We proposed a geometrical constrained Chan-Vese based scheme to segment tumor in PET for this special case by considering the similarity between two adjacent slices. Methods: The proposed scheme performs segmentation in a slice-by-slice fashion where an accurate segmentation of one slice is used as the guidance for segmentation of rest slices. For a slice that the tumor is not directly connected to organs/tissues with similar intensity values, a conventional clustering-based segmentation method under user’s guidance is used to obtain an exact tumor contour. This is set as the initial contour and the Chan-Vese algorithm is applied for segmenting the tumor in the next adjacent slice by adding constraints of tumor size, position and shape information. This procedure is repeated until the last slice of PET containing tumor. The proposed geometrical constrained Chan-Vese based algorithm was implemented in Matlab and its performance was tested on several cervical cancer patients where cervix and bladder are connected with similar activity values. The positive predictive values (PPV) are calculated to characterize the segmentation accuracy of the proposed scheme. Results: Tumors were accurately segmented by the proposed method even when they are connected with bladder in the image with no difference in intensity. The average PPVs were 0.9571±0.0355 and 0.9894±0.0271 for 17 slices and 11 slices of PET from two patients, respectively. Conclusion: We have developed a new scheme to segment tumor in PET images for the special case that the tumor is quite similar to or connected to normal organs/tissues in the image. The proposed scheme can provide a reliable way for segmenting tumors.

60 APPLIED LIFE SCIENCES↗