Temporal Median Research Articles

Temporal patterns for inguinal hernia recurrence operations after Shouldice Repair.

Shouldice Repair for inguinal hernias results in a low recurrence rate; however, little is known about the risk factors for these relapses. In the present study, we reviewed all patient's undergoing a reoperation for recurrence after a primary Shouldice Repair. Patients undergoing recurrent inguinal hernia repair from 2013 to 2017 were identified. The subgroup of patients with the first recurrence after a Shouldice Repair at this institution was selected and included. Data collection from the index and the reoperation surgery were performed, as well as statistical analysis. A total of 125 patients were included in the analysis. The mean age was 50.8 ± 13.9 (body mass index: 24.6 ± 2.6kg/m2), 97% were male. The most common interval for a recurrence reoperation was in the first 5years following the initial surgery (37.6% of patients; mode: 1year; median: 7years; mean: 13.7 ± 13.8years). A temporal median pattern for recurrence reoperation according to age interval was noted (patients < 41years old: 20years; patients aged 41-60years old: 10years and patients > 61years: 2years)). Patients undergoing reoperations for recurrent inguinal hernias after a primary Shouldice Repair presented a pattern of temporal recurrence according to age. Older patients present with earlier recurrences than younger ones, an important consideration in the assessment of patients postoperatively.

Discovering spatio-temporal dependencies in mangrove forest change: evidence from northwestern Asian communities

Abstract The Iranian southern intertidal ecosystems are colonized by mangrove forests (mostly Avicennia marina) and presented a notable mangrove area expansion over the past decades. To discover its spatio-temporal dependencies, five mangrove forests were studied along a coastline of more than 1,200 km. Cloud-free Landsat images were processed using a temporal median filter to produce one representative Landsat image for each year from 1988 to 2020. The images were categorized into mangrove and non-mangrove classes using an object-based algorithm and visual post-classification. A consistent mangrove area expansion was observed across all regions, which regressed linearly against time with an adjusted R2 range of 0.914–0.955 (p &amp;lt; 0.01). The expansions were also found to be highly significantly similar (0.182 &amp;lt; p &amp;lt; 0.822) using the t-test applied on the slope of region-specific regression models. Three spatial metrics of infilling, edge-expansion, and outlying were developed to investigate the spatial pattern of expansions. Results showed the dominance of the edge-expansion pattern across all regions (53–78% of the expansions), which was attributed to the better establishment of propagules near and under the canopy of parent trees. The outlying pattern (11–22%) was related to mangrove colonization along the inland creeks.

Efficient Deep Semantic Segmentation for Land Cover Classification Using Sentinel Imagery

Nowadays, different machine learning approaches, either conventional or more advanced, use input from different remote sensing imagery for land cover classification and associated decision making. However, most approaches rely heavily on time-consuming tasks to gather accurate annotation data. Furthermore, downloading and pre-processing remote sensing imagery used to be a difficult and time-consuming task that discouraged policy makers to create and use new land cover maps. We argue that by combining recent improvements in deep learning with the use of powerful cloud computing platforms for EO data processing, specifically the Google Earth Engine, we can greatly facilitate the task of land cover classification. For this reason, we modify an efficient semantic segmentation approach (U-TAE) for a satellite image time series to use, as input, a single multiband image composite corresponding to a specific time range. Our motivation is threefold: (a) to improve land cover classification performance and at the same time reduce complexity by using, as input, satellite image composites with reduced noise created using temporal median instead of the original noisy (due to clouds, calibration errors, etc.) images, (b) to assess performance when using as input different combinations of satellite data, including Sentinel-2, Sentinel-1, spectral indices, and ALOS elevation data, and (c) to exploit channel attention instead of the temporal attention used in the original approach. We show that our proposed modification on U-TAE (mIoU: 57.25%) outperforms three other popular approaches, namely random forest (mIoU: 39.69%), U-Net (mIoU: 55.73%), and SegFormer (mIoU: 53.5%), while also using fewer training parameters. In addition, the evaluation reveals that proper selection of the input band combination is necessary for improved performance.

Scene Independency Matters: An Empirical Study of Scene Dependent and Scene Independent Evaluation for CNN-Based Change Detection

Visual change detection in video is one of the essential tasks in computer vision applications. Recently, a number of supervised deep learning methods have achieved top performance over the benchmark datasets for change detection. However, inconsistent training-testing data division schemes adopted by these methods have led to documentation of incomparable results. We address this crucial issue through our own propositions for benchmark comparative analysis. The existing works have evaluated the model in scene dependent evaluation setup which makes it difficult to assess the generalization capability of the model in completely unseen videos. It also leads to inflated results. Therefore, in this paper, we present a completely scene independent evaluation strategy for a comprehensive analysis of the model design for change detection. We propose well-defined scene independent and scene dependent experimental frameworks for training and evaluation over the benchmark CDnet 2014, LASIESTA and SBMI2015 datasets. A cross-data evaluation is performed with PTIS dataset to further measure the robustness of the models. We designed a fast and lightweight online end-to-end convolutional network called ChangeDet (speed-58.8 fps and model size-1.59 MB) in order to achieve robust performance in completely unseen videos. The ChangeDet estimates the background through a sequence of maximum multi-spatial receptive feature (MMSR) blocks using past temporal history. The contrasting features are produced through the assimilation of temporal median and contemporary features from the current frame. Further, these features are processed through an encoder-decoder to detect pixel-wise changes. The proposed ChangeDet outperforms the existing state-of-the-art methods in all four benchmark datasets.

Spatiotemporal Variability of Dune Velocities and Corresponding Uncertainties, Detected from Optical Image Matching in the North Sinai Sand Sea, Egypt

Understanding the relationship between climatic conditions and dune ecosystems requires the large-scale monitoring of spatiotemporal patterns of dune velocities. Due to their large extent and remoteness, dune fields are ideal for remote sensing techniques. Dune velocities in the Sand Sea North Sinai are characterized by large spatial and temporal variability. To this end, a total of 265 pairs from four Landsat-8 images from April 2013 to April 2018 were automatically matched with the COSI Corr engine to determine the dune velocities. These pairs were selected so that differences in the solar angles were small and spanned at least one year. This helps to reduce shadowing in the deformation fields and the error budget in converting displacements to annual velocities. To improve spatial coverage and reduce measurement uncertainty, the fusion of individual offset maps is considered feasible. We compared the performance of two methods (i.e., inversion and temporal median fusion) in performing the fusion of individual velocities, and the two methods showed good agreement. The fusion of individual velocities allowed us to estimate the final velocities for about 98.8% of the dune areas. Our results suggest that the magnitudes and directions of dune migration at Sand Sea are spatially and temporally variable. The geometric mean of the active features associated with 12 regions in the Sand Sea ranged from 1.65 m/y to 3.52 m/y, with median directions from 56.19° to 173.11°. The stable regions allowed us to estimate the 95% confidence intervals of the final velocities and extend these calculations to the dune targets. The median uncertainties were 0.10 m/y and 0.25 m/y for the stable and moving targets, respectively. We estimated the coherence of the final velocity vector, which can be considered as an indicator of the homogeneity of migration directions between the offset maps. We compared the final Landsat-8 velocities with those from Sentinel-2 to validate the results and found a good agreement in the magnitudes and directions. The process of selecting high-quality pairs and then fusing the individual maps showed a high performance in terms of spatial coverage and reliability of the extracted velocities.

Deriving exclusion maps from C-band SAR time-series in support of floodwater mapping

Synthetic Aperture Radar (SAR) intensity is used as an input to many flood-mapping algorithms. The appearance of floodwater tends to cause a substantial decrease of backscatter intensity over scarcely vegetated terrain. However, limitations exist in areas where the SAR backscatter is not sufficiently sensitive to surface changes, e.g. shadow areas due to topography or obstacles on the ground, densely forested areas, sand, etc. Thus, we argue that it is of paramount importance to complement any SAR-based flood extent map with an exclusion map (EX-map) indicating all areas where the presence of water cannot be derived from SAR intensity observations. In this study, we introduce a methodology for generating an EX-map based on the analysis of time-series of SAR backscatter data. In particular, the identification of the EX-map is based on the combined use of three temporal indicators based on backscatter statistics, i.e. temporal median, minimum and standard deviation. As a test case, EX-maps were derived from Sentinel-1 data acquired during the 2014–2019 time period from six representative study sites. Reference maps were generated using a global land cover map, Digital Elevation Model (DEM)-derived shadow/layover masks, global urban footprint (GUF) data and a Sand Exclusion Layer (SEL). The cross-comparison revealed that the EX-map was consistent with reference maps obtained from other data sources.

Exploring COVID-19 Daily Records of Diagnosed Cases and Fatalities Based on Simple Nonparametric Methods.

Containment strategies to combat epidemics such as SARS-CoV-2/COVID-19 require the availability of epidemiological parameters, e.g., the effective reproduction number. Parametric models such as the commonly used susceptible-infected-removed (SIR) compartment models fitted to observed incidence time series have limitations due to the time-dependency of the parameters. Furthermore, fatalities are delayed with respect to the counts of new cases, and the reproduction cycle leads to periodic patterns in incidence time series. Therefore, based on comprehensible nonparametric methods including time-delay correlation analyses, estimates of crucial parameters that characterise the COVID-19 pandemic with a focus on the German epidemic are presented using publicly available time-series data on prevalence and fatalities. The estimates for Germany are compared with the results for seven other countries (France, Italy, the United States of America, the United Kingdom, Spain, Switzerland, and Brazil). The duration from diagnosis to death resulting from delay-time correlations turns out to be 13 days with high accuracy for Germany and Switzerland. For the other countries, the time-to-death durations have wider confidence intervals. With respect to the German data, the two time series of new cases and fatalities exhibit a strong coherence. Based on the time lag between diagnoses and deaths, properly delayed asymptotic as well as instantaneous fatality–case ratios are calculated. The temporal median of the instantaneous fatality–case ratio with time lag of 13 days between cases and deaths for Germany turns out to be . Time courses of asymptotic fatality–case ratios are presented for other countries, which substantially differ during the first half of the pandemic but converge to a narrow range with standard deviation and mean . Similar results are obtained from comparing time courses of instantaneous fatality–case ratios with optimal delay for the 8 exemplarily chosen countries. The basic reproduction number, , for Germany is estimated to be between and depending on the generation time, which is estimated based on a delay autocorrelation analysis. Resonances at about 4 days and 7 days are observed, partially attributable to weekly periodicity of sampling. The instantaneous (time-dependent) reproduction number is estimated from the incident (counts of new) cases, thus allowing us to infer the temporal behaviour of the reproduction number during the epidemic course. The time course of the reproduction number turns out to be consistent with the time-dependent per capita growth.

Depression Level Prediction Using Deep Spatiotemporal Features and Multilayer Bi-LTSM

Depression is a serious psychiatric disorder that restricts an individuals ability to work properly in both their daily and professional lives. Usually, the diagnosis of depression often needs a thorough assessment by an expert. Recently, significant consideration has been given to automatic depression prediction for more reliable and efficient depression investigation. In this article, we propose a novel framework to estimate the depression level from video data by employing a two-stream deep spatiotemporal network. Our approach extracts spatial information using the Inception-ResNet-v2 network. In contrast, we introduce a volume local directional number (VLDN) based dynamic feature descriptor to capture facial motions. Then, the feature map obtained from the VLDN is fed into a convolutional neural network (CNN) to obtain more discriminative features. Additionally, we designed a multilayer bidirectional long short-term memory (Bi-LSTM) model to obtain temporal information by integrating the temporal median pooling (TMP) approach into the model. The TMP approach is employed on the temporal fragments of spatial and temporal features. Finally, extensive experimental analysis of two challenging datasets, AVEC2013 and AVEC2014, demonstrates that the proposed approach shows promising performance compared to the existing approaches for depression level prediction.

Assessment of Coastal Aquaculture for India from Sentinel-1 SAR Time Series

Aquaculture is one of the fastest growing primary food production sectors in India and ranks second behind China. Due to its growing economic value and global demand, India’s aquaculture industry experienced exponential growth for more than one decade. In this study, we extract land-based aquaculture at the pond level for the entire coastal zone of India using large-volume time series Sentinel-1 synthetic-aperture radar (SAR) data at 10-m spatial resolution. Elevation and slope from Shuttle Radar Topographic Mission digital elevation model (SRTM DEM) data were used for masking inappropriate areas, whereas a coastline dataset was used to create a land/ocean mask. The pixel-wise temporal median was calculated from all available Sentinel-1 data to significantly reduce the amount of noise in the SAR data and to reduce confusions with temporary inundated rice fields. More than 3000 aquaculture pond vector samples were collected from high-resolution Google Earth imagery and used in an object-based image classification approach to exploit the characteristic shape information of aquaculture ponds. An open-source connected component segmentation algorithm was used for the extraction of the ponds based on the difference in backscatter intensity of inundated surfaces and shape metrics calculated from aquaculture samples as input parameters. This study, for the first time, provides spatial explicit information on aquaculture distribution at the pond level for the entire coastal zone of India. Quantitative spatial analyses were performed to identify the provincial dominance in aquaculture production, such as that revealed in Andhra Pradesh and Gujarat provinces. For accuracy assessment, 2000 random samples were generated based on a stratified random sampling method. The study demonstrates, with an overall accuracy of 0.89, the spatio-temporal transferability of the methodological framework and the high potential for a global-scale application.

An effective approach for video condensation by ribbon carving

The proposed system aims in developing a method to stream video based on time intervals. Regular or irregular sub-sampling is not sufficient for scaling of video in time. Seam carving has been used where an Image can be resized either by inserting or removing pixels. Ribbon carving is an extension of seam carving which resizes the video in temporal direction. The non-parametric kernel model used for background subtraction has been replaced with temporal median filter. It enhances the processing speed and memory allocation has been more efficient, thus reducing the time of video without loss of data.

Technical Note: Kinect V2 surface filtering during gantry motion for radiotherapy applications.

In radiotherapy, the Kinect V2 camera, has recently received a lot of attention concerning many clinical applications including patient positioning, respiratory motion tracking, and collision detection during the radiotherapy delivery phase. However, issues associated with such applications are related to some materials and surfaces reflections generating an offset in depth measurements especially during gantry motion. This phenomenon appears in particular when the collimator surface is observed by the camera; resulting in erroneous depth measurements, not only in Kinect surfaces itself, but also as a large peak when extracting a 1D respiratory signal from these data. In this paper, we proposed filtering techniques to reduce the noise effect in the Kinect-based 1D respiratory signal, using a trend removal filter, and in associated 2D surfaces, using a temporal median filter. Filtering process was validated using a phantom, in order to simulate a patient undergoing radiotherapy treatment while having the ground truth. Our results indicate a better correlation between the reference respiratory signal and its corresponding filtered signal (Correlation coefficient of 0.76) than that of the nonfiltered signal (Correlation coefficient of 0.13). Furthermore, surface filtering results show a decrease in the mean square distance error (85%) between the reference and the measured point clouds. This work shows a significant noise compensation and surface restitution after surface filtering and therefore a potential use of the Kinect V2 camera for different radiotherapy-based applications, such as respiratory tracking and collision detection.

LaBGen: A method based on motion detection for generating the background of a scene

Given a video sequence acquired with a fixed camera, the generation of the stationary background of the scene is a challenging problem which aims at computing a reference image for a motionless background. For that purpose, we developed our method named LaBGen, which emerged as the best one during the Scene Background Modeling and Initialization (SBMI) workshop organized in 2015, and the IEEE Scene Background Modeling Contest (SBMC) organized in 2016. LaBGen combines a pixel-wise temporal median filter and a patch selection mechanism based on motion detection. To detect motion, a background subtraction algorithm decides, for each frame, which pixels belong to the background. In this paper, we describe the LaBGen method extensively, evaluate it on the SBI 2016 dataset and compare its performance with other background generation methods. We also study its computational complexity, the performance sensitivity with respect to its parameters, and the stability of the predicted background image over time with respect to the chosen background subtraction algorithm. We provide an open source C++ implementation at http://www.telecom.ulg.ac.be/labgen.

Finding the Optimal Background Subtraction Algorithm for EuroHockey 2015 Video

Background subtraction is a classic step in a vision-based localization and tracking workflow. Previous studies have compared background subtraction algorithms on publicly available datasets; however comparisons were made only with manually optimized parameters. The aim of this research was to identify the optimal background subtraction algorithm for a set of field hockey videos captured at EuroHockey 2015.Paticle Swarm Optimization was applied to find the optimal background subtraction algorithm. The objective function was the F-score, i.e. the harmonic mean of precision and recall. The precision and recall were calculated using the output of the background subtraction algorithm and gold standard labeled images. The training dataset consisted of 15 x 13 second field hockey video segments. The test data consisted of 5 x 13 second field hockey video segments. The video segments were chosen to be representative of the teams present at the tournament, the times of day the matches were played and the weather conditions experienced. Each segment was 960 pixels x 540 pixels and had 10 ground truth labeled frames. Eight commonly used background subtraction algorithms were considered.Results suggest that a background subtraction algorithm must use optimized parameters for a valid comparison of performance. Particle Swarm Optimization is an appropriate method to undertake this optimization. The optimal algorithm, Temporal Median, achieved an F-score of 0.791 on the test dataset, suggesting it generalizes to the rest of the video footage captured at EuroHockey 2015.

Removing Noises Induced by Gamma Radiation in Cerenkov Luminescence Imaging Using a Temporal Median Filter.

Cerenkov luminescence imaging (CLI) can provide information of medical radionuclides used in nuclear imaging based on Cerenkov radiation, which makes it possible for optical means to image clinical radionuclide labeled probes. However, the exceptionally weak Cerenkov luminescence (CL) from Cerenkov radiation is susceptible to lots of impulse noises introduced by high energy gamma rays generating from the decays of radionuclides. In this work, a temporal median filter is proposed to remove this kind of impulse noises. Unlike traditional CLI collecting a single CL image with long exposure time and smoothing it using median filter, the proposed method captures a temporal sequence of CL images with shorter exposure time and employs a temporal median filter to smooth a temporal sequence of pixels. Results of in vivo experiments demonstrated that the proposed temporal median method can effectively remove random pulse noises induced by gamma radiation and achieve a robust CLI image.

High-speed real-time magnetic resonance imaging of fast tongue movements in elite horn players.

This paper describes the use of high-speed real-time (RT) magnetic resonance imaging (MRI) in quantifying very rapid motor function within the oropharyngeal cavity of six elite horn players. Based on simultaneous sound recordings, the efficacy of RT-MRI films at 30 and 100 frames per second (fps) was assessed for tongue movements associated with double tonguing performance. Serial images with a nominal temporal resolution of 10.0 and 33.3 ms were obtained by highly undersampled radial fast low-angle shot (FLASH) sequences (5 and 17 spokes, respectively) using complementary sets of spokes for successive acquisitions (extending over 9 and 5 frames, respectively). Reconstructions of high-speed images were obtained by temporally regularized nonlinear inversion (NLINV) as previously described. A customized MATLAB toolkit was developed for the extraction of line profiles from MRI films to quantify temporal phenomena associated with task performance. The analyses reveal that for the present setting, which required the use of a temporal median filter to optimize image quality, acquisition rates of 30 fps are inadequate to accurately detect tongue movements during double tonguing, but that rates of 100 fps do allow for a precise quantification of movement. These data for the first time demonstrate the extreme performance of elite horn players. High-speed RT-MRI offers so far unavailable opportunities to study the oropharyngeal movements during brass playing with future potential for teaching and the treatment of patients suffering from dystonia.

2091965 How Sensitive is the Computer Assisted Diagnostic Prototype for Malignancy Assessment of Focal Liver Lesions By Ceus?

2091965 How Sensitive is the Computer Assisted Diagnostic Prototype for Malignancy Assessment of Focal Liver Lesions By Ceus?

Use of the temporal median and trimmed mean mitigates effects of respiratory motion in multiple-acquisition abdominal diffusion imaging

Respiratory motion commonly confounds abdominal diffusion-weighted magnetic resonance imaging, where averaging of successive samples at different parts of the respiratory cycle, performed in the scanner, manifests the motion as blurring of tissue boundaries and structural features and can introduce bias into calculated diffusion metrics. Storing multiple averages separately allows processing using metrics other than the mean; in this prospective volunteer study, median and trimmed mean values of signal intensity for each voxel over repeated averages and diffusion-weighting directions are shown to give images with sharper tissue boundaries and structural features for moving tissues, while not compromising non-moving structures. Expert visual scoring of derived diffusion maps is significantly higher for the median than for the mean, with modest improvement from the trimmed mean. Diffusion metrics derived from mono- and bi-exponential diffusion models are comparable for non-moving structures, demonstrating a lack of introduced bias from using the median. The use of the median is a simple and computationally inexpensive alternative to complex and expensive registration algorithms, requiring only additional data storage (and no additional scanning time) while returning visually superior images that will facilitate the appropriate placement of regions-of-interest when analysing abdominal diffusion-weighted magnetic resonance images, for assessment of disease characteristics and treatment response.

White matter microstructural damage in small vessel disease is associated with Montreal cognitive assessment but not with mini mental state examination performances: vascular mild cognitive impairment Tuscany study.

Montreal Cognitive Assessment (MoCA) has been proposed as a screening tool in vascular cognitive impairment. Diffusion tensor imaging is sensitive to white matter microstructural damage. We investigated if diffusion tensor imaging-derived indices are more strongly associated with performances on MoCA or on the widely used mini mental state examination in patients with mild cognitive impairment and small vessel disease. Mild cognitive impairment patients with moderate/severe degrees of white matter hyperintensities on MRI were enrolled. Lacunar infarcts, cortical atrophy, medial temporal lobe atrophy and median values of mean diffusivity and fractional anisotropy of the cerebral white matter were studied and correlated with cognitive tests performances. Seventy-six patients (mean age 75.1±6.8 years, mean years of education 8.0±4.3) were assessed. In univariate analyses, a significant association of both MoCA and mini mental state examination scores with age, education, cortical atrophy, and medial temporal lobe atrophy was found, whereas mean diffusivity and fractional anisotropy were associated with MoCA. In partial correlation analyses, adjusting for all demographic and neuroimaging variables, both mean diffusivity and fractional anisotropy were associated only with MoCA (mean diffusivity: r= -0.275, P=0.023; fractional anisotropy: r=0.246, P=0.043). In patients with mild cognitive impairment and small vessel disease, diffusion tensor imaging-measured white matter microstructural damage is more related to MoCA than mini mental state examination performances. MoCA is suited for the cognitive screening of patients with small vessel disease.

Statistically tuned Gaussian background subtraction technique for UAV videos

Background subtraction is one of the efficient techniques to segment the targets from non-informative background of a video. The traditional background subtraction technique suits for videos with static background whereas the video obtained from unmanned aerial vehicle has dynamic background. Here, we propose an algorithm with tuning factor and Gaussian update for surveillance videos that suits effectively for aerial videos. The tuning factor is optimized by extracting the statistical features of the input frames. With the optimized tuning factor and Gaussian update an adaptive Gaussian-based background subtraction technique is proposed. The algorithm involves modelling, update and subtraction phases. This running Gaussian average based background subtraction technique uses updation at both model generation phase and subtraction phase. The resultant video extracts the moving objects from the dynamic background. Sample videos of various properties such as cluttered background, small objects, moving background and multiple objects are considered for evaluation. The technique is statistically compared with frame differencing technique, temporal median method and mixture of Gaussian model and performance evaluation is done to check the effectiveness of the proposed technique after optimization for both static and dynamic videos.

Video De-Fencing

This paper describes and provides an initial solution to a novel video editing task, i.e., video de-fencing. It targets automatic restoration of the video clips that are corrupted by fence-like occlusions during capture. Our key observation lies in the visual parallax between fences and background scenes, which is caused by the fact that the former are typically closer to the camera. Unlike in traditional image inpainting, fence-occluded pixels in the videos tend to appear later in the temporal dimension and are therefore recoverable via optimized pixel selection from relevant frames. To eventually produce fence-free videos, major challenges include cross-frame subpixel image alignment under diverse scene depth, and correct pixel selection that is robust to dominating fence pixels. Several novel tools are developed in this paper, including soft fence detection, weighted truncated optical flow method, and robust temporal median filter. The proposed algorithm is validated on several real-world video clips with fences.