Post-processing Approach Research Articles

Mitigating Regression Faults Induced by Feature Evolution in Deep Learning Systems

Deep learning (DL) systems have been widely utilized across various domains. However, the evolution of DL systems can result in regression faults. In addition to the evolution of DL systems through the incorporation of new data, feature evolution, such as the addition of new features, is also common and can introduce regression faults. In this work, we first investigate the underlying factors that are correlated with regression faults in feature evolution scenarios, i.e., redundancy and contribution shift. Based on our investigation, we propose a novel mitigation approach called FeaProtect, which aims to minimize the impact of these two factors. To evaluate the performance of FeaProtect, we conducted an extensive study comparing it with state-of-the-art approaches. The results show that FeaProtect outperforms the in-processing baseline approaches, with an average improvement of 50.6% \(\sim\) 56.4% in terms of regression fault mitigation. We also show that FeaProtect can further enhance the effectiveness of mitigating regression faults by integrating with state-of-the-art post-processing approaches.

Plasma-electrolytic oxidation: A rapid single step post processing approach for additively manufactured biomedical implants.

Laser-powder bed fusion (PBF-LB) has enabled production of customised skeletal implants that incorporate porous lattices structures to enable bone ingrowth. However, the inherent surface roughness of PBF-LB, characterised by partially adhered particles and undulating sub-topography, remains a barrier to adoption. As such PBF-LB surfaces require several time-consuming post-processing steps, nevertheless, conventional finishing techniques are often limited by geometrical part complexity, making them unsuitable for porous PBF-LB parts. Herein we explore the possibility to utilise plasma-electrolytic oxidation (PEO) as a rapid, single step surface finishing method not constrained by implant design. Specifically, PEO treatment was performed in a phosphate-based electrolyte on as-printed and polished Ti-6Al-4V PBF-LB samples with complete surface coverage and chemical functionalisation, as observed by optical profilometry, SEM-EDX, XRD and XRF, achieved after only 20min. To test the lack of geometric constraints brought by PEO, clinically relevant BCC porous lattices were also successfully PEO treated accomplishing a coating that either masked or removed surface adhered particles throughout the structure. Promisingly for medical application, no cytotoxicity was noted for MC3T3-E1 murine osteoblasts over 7days and significantly more (p<0.05) mineralisation was observed after 21days compared with as-printed and polished PBF-LB controls. Still, an enhanced pro-inflammatory response, iNOS and TNF-α, was observed in murine RAW261 macrophages seeded on PEO surfaces, indicating further optimisation is required to guide the inflammatory process. Overall, these findings showcase the widespread opportunity to robustly ensure PBF-LB implant safety by using PEO to tackle partially adhered particles while also offering new avenues to enhance functionality through variations in coating chemistry.

Assessing the Repeatability of Inner Choroid Flow Deficit Percentage in Intermediate AMD and Controls: A Comparative Study between Different Post-processing Approaches.

To assess repeatability of macular inner choroid flow deficit percentage (ICFD%) in intermediate age-related macular degeneration (iAMD) and controls employing various post-processing approaches. Cross-sectional, observational study. We included (1) 22 iAMD and 24 control eyes, with (2) age >50 years, (3) visual acuity >20/32, and (4) no additional ocular and systemic confounders. underwent four consecutive 6x6-mm OCTA scans (2 acquired at 100-kHz and 2 at 200-kHz speed; PLEX® Elite 9000) for intra-session analysis. Same protocol was repeated after 30 minutes for inter-session analysis. Three slabs of different thicknesses were generated underneath Bruch's membrane (4-14, 4-19, 4-24 µm). All slabs were processed with: (1) binarization-only using Phansalkar method (r=4-15 pixels); (2) compensation+binarization; (3) averaging+binarization; (4) averaging+compensation+binarization. ICFD% was measured within 3- and 5-mm circles, and measurements were repeated after excluding drusen areas.Repeatability was analyzed with generalized linear mixed-effects models, intraclass correlation coefficients (ICC) and Levene's variance test. Most post-processing approaches demonstrated high repeatability (ICC>0.75) with no significant test-retest differences (p>0.05). Compensation+binarization of 15-µm slabs at 200 kHz showed the highest repeatability (ICC: 0.96-0.98). Excluding drusen did not significantly impact ICFD% measurements (p>0.05), showing increased ICCs for 10 µm-thick and binarized-only slabs. Strong repeatability can be achieved with various post-processing methods for assessing ICFD%, especially with compensation+binarization of 15-µm slabs acquired at 200 kHz. Drusen removal does not seem to impact repeatability in iAMD when using a swept-source device, except for specific settings.

Superconvergence analysis and extrapolation of a BDF2 fully discrete scheme for nonlinear reaction–diffusion equations

The main aim of this paper is to propose a 2-step backward differential formula (BDF2) fully discrete scheme with the bilinear Q11 finite element method (FEM) for the nonlinear reaction–diffusion equation. By use of the combination technique of the element’s interpolation and Ritz projection, and through the interpolation post-processing approach, the superclose and global superconvergence estimates with order O(h2+τ2) in H1-norm are deduced rigorously. Furthermore, with the help of the asymptotic error expansion of the Q11 element, a new suitable fully discrete scheme is developed, and the extrapolation result of order O(h3+τ2) in H1-norm is derived, which is one order higher than that of the above traditional superconvergence estimate with respect to h. Here h is the mesh size and τ is the time step. Finally, some numerical results are provided to verify the theoretical analysis. It seems that the extrapolation of the fully discrete finite element scheme has never been seen in the previous studies.

Superconvergence analysis of low order nonconforming finite element method for coupled nonlinear semiconductor device problem

A linearized decoupled fully discrete scheme is developed and investigated for the coupled nonlinear semiconductor device problem with low order nonconforming EQ1rot element. Then, by use of its special property: the consistency error in the broken H1-norm can reach to second order when the exact solutions belong to H3(Ω), just one order higher than its interpolation error, together with some proper approaches such as the discrete derivative transfer trick, difference quotient between two adjacent time levels, mathematics induction method and so on, the difficulty caused by the nonlinearity is ingeniously coped with, and the superclose estimates about the related variables are derived rigorously. In addition, the satisfactory global superconvergence results are obtained through the interpolation postprocessing approach. Finally, a numerical example is presented to validate the theoretical analysis and the good performance of the proposed method.

Efficient algorithm for thermal nondestructive testing and evaluation by considering the heteroscedastic nature of noise sources in infrared thermography

Abstract Thermal Imaging is a promising Non Destructive Testing &amp; Evaluation (NDT &amp; E) approach to monitor the health of composite materials. Among various post processing approaches adopted in thermal imaging for NDT &amp; E, statistical analysis schemes gained importance due to their reliability and data reduction capabilities. This paper provides an insight to a factor analysis-based statistical approach to detect the hidden defects in the Glass Fiber Reinforced Polymer (GFRP) sample. The proposed approach models the observed data covariance into combination of temporal signal covariance and noise covariance matrices. The modeling of the diagonal covariance matrix (with different elements) is motivated by the presence of heterogeneity in the experimental data obtained from GFRP sample.This novel method is based on the coordinate descent technique, which estimates the covariance matrix of the noise variances iteratively by minimizing the negative log likelihood function. The obtained results from the chosen GFRP samples compared with the widely used statistical Principal Component Thermography (PCT) technique illustrate the improved performance in terms of defect detection with the proposed technique.

Unconditional superconvergence analysis of a new energy stable nonconforming BDF2 mixed finite element method for BBM–Burgers equation

The focus of this paper is to establish a new energy stable 2-step backward differentiation formula (BDF2) fully-discrete mixed finite element method (MFEM) for the Benjamin–Bona–Mahony–Burgers (BBM–Burgers) equation with the nonconforming rectangular EQ1rot element and zero-order Raviart–Thomas (R–T) element (EQ1rot/Q10×Q01). Based on the energy stable property, the existence and uniqueness of the numerical solution are proved by the Brouwer fixed point theorem. Subsequently, with the assistance of some typical properties of this element pair and the interpolation post-processing approach, the unconditional superclose and superconvergence results with O(h2+τ2) are obtained directly without any constraints between the spatial partition parameter h and the time step τ. It is worthy to mention that the method and analysis presented herein are very different from the time–space splitting approach utilized in the previous studies and simplify the implement. Finally, two numerical experiments are executed to validate the theoretical analysis.

Unconditionally superconvergence error analysis of an energy-stable finite element method for Schrödinger equation with cubic nonlinearity

In this paper, the unconditionally superconvergence analysis is studied for the cubic Schrödinger equation with an energy-stable finite element method. A different approach is proposed to obtain the unconditionally superclose error estimate in H1-norm firstly without using the time splitting technique required in the previous literature. The key to the analysis is to use a priori boundedness of the numerical solution in energy norm and control the nonlinear terms rigorously by two cases, i.e., τ≤h2 and τ≥h2, where τ denotes the temporal size and h is the spatial size. Subsequently, the global superconvergence error estimate in H1-norm is derived by an effective interpolation post-processing approach. Finally, some numerical experiments are carried out to confirm the theoretical findings.

Event-Based Measurement of Aeroelastic Structure in High-Speed Flow

In high-speed aerodynamics research, point sensors are ideal for embedding in test models but lack spatial resolution, whereas high-speed cameras offer spatiotemporally resolved measurement but involve significant footprint, cost, and data size. To address these tradeoffs, this study explores the application of nascent event-based cameras for high-speed tests. Event-based cameras support continuous, data-sparse kilohertz-equivalent imaging at 720p resolution, on form factors as small as 36 mm and 40 grams in mass, combining the benefits of point sensors and high-speed cameras. However, these attributes come from asynchronous pixels that necessitate unique operating and postprocessing approaches. Here, the authors adapted event-based cameras for two-/three-dimensional photogrammetric tracking of aeroelastic structures, demonstrating an event-based workflow and two tracking algorithms (mean-shift filtering and circle fit). Bench-top validations achieved three-dimensional precision of 0.35 mm/s on 20 mm/s motion across a 259 mm field of view, while two-dimensional measurements of an aeroelastic titanium panel in Mach 0.76 transonic flow successfully identified millimeter-scale vibrations at 43.7, 120, and 270 Hz, validated against a laser displacement and high-speed camera. The transonic test’s raw data were 145.8 MB on the event-based camera, compared to 88.5 GB on the high-speed camera. The presented results demonstrated the viability of event-based techniques in high-speed aerodynamic testing, while highlighting challenges such as polarity switching and pixel latency.

Iron accumulation/overload and Alzheimer's disease risk factors in the precuneus region: A comprehensive narrative review.

Alzheimer's disease (AD) is a neurodegenerative disease that is characterized by amyloid plaques, neurofibrillary tangles, and neuronal loss. Early cerebral and body iron dysregulation and accumulation interact with AD pathology, particularly in the precuneus, a crucial functional hub in cognitive functions. Quantitative susceptibility mapping (QSM), a novel post-processing approach, provides insights into tissue iron levels and cerebral oxygen metabolism and reveals abnormal iron accumulation early in AD. Increased iron deposition in the precuneus can lead to oxidative stress, neuroinflammation, and accelerated neurodegeneration. Metabolic disorders (diabetes, non-alcoholic fatty liver disease (NAFLD), and obesity), genetic factors, and small vessel pathology contribute to abnormal iron accumulation in the precuneus. Therefore, in line with the growing body of literature in the precuneus region of patients with AD, QSM as a neuroimaging method could serve as a non-invasive biomarker to track disease progression, complement other imaging modalities, and aid in early AD diagnosis and monitoring.

Effects of Post-Processing Parameters on 3D-Printed Dental Appliances: A Review.

In recent years, additive manufacturing (AM) has been recognized as a transformative force in the dental industry, with the ability to address escalating demand, expedite production timelines, and reduce labor-intensive processes. Despite the proliferation of three-dimensional printing technologies in dentistry, the absence of well-established post-processing protocols has posed formidable challenges. This comprehensive review paper underscores the critical importance of precision in post-processing techniques for ensuring the acquisition of vital properties, encompassing mechanical strength, biocompatibility, dimensional accuracy, durability, stability, and aesthetic refinement in 3D-printed dental devices. Given that digital light processing (DLP) is the predominant 3D printing technology in dentistry, the main post-processing techniques and effects discussed in this review primarily apply to DLP printing. The four sequential stages of post-processing support removal, washing, secondary polymerization, and surface treatments are systematically navigated, with each phase requiring meticulous evaluation and parameter determination to attain optimal outcomes. From the careful selection of support removal tools to the consideration of solvent choice, washing methodology, and post-curing parameters, this review provides a comprehensive guide for practitioners and researchers. Additionally, the customization of post-processing approaches to suit the distinct characteristics of different resin materials is highlighted. A comprehensive understanding of post-processing techniques is offered, setting the stage for informed decision-making and guiding future research endeavors in the realm of dental additive manufacturing.

Stress-constrained concurrent multiscale topological design of porous composites based on discrete material optimisation

Porous composites have attracted increasing attention in recent decades. This study develops a concurrent multiscale topology optimisation (CMTO) method under a prescribed stress constraint for designing porous composites with multi-domain microstructures. First, to address the difficulty of predicting local stress due to varying of microstructural type throughout the optimisation process, a continuous and differentiable stress measure is proposed to effectively approximate the local stress. Second, an inverse homogenisation method based on isogeometric analysis (IGA) is developed to improve the accuracy of stress prediction, and then it is integrated into a CMTO which is developed based on the discrete material optimisation (DMO) interpolation scheme. Third, a stress constraint which is differentiable with respect to both macro and micro design variables is proposed to enable the stress-constrained concurrent optimisation of the macrostructural configuration, microstructural configuration and distribution. Fourth, a novel post-processing approach is established to achieve smooth while volume preserving contour of unit cells with layouts. Finally, two benchmark design examples, namely l-bracket and Crack problems, are implemented using the presented CMTO under a global stress constraint to demonstrate the effectiveness of the proposed method. The result indicates that the proposed method can effectively decrease the stress concentration via three design manners, i.e., the macrostructural configuration, microstructural configuration and distribution. Also, an “interface-enlarging” phenomenon was interestingly but reasonably found in those cases when subjected to stress-constraint considerations.

A Flow-based Truncated Denoising Diffusion Model for super-resolution Magnetic Resonance Spectroscopic Imaging

Magnetic Resonance Spectroscopic Imaging (MRSI) is a non-invasive imaging technique for studying metabolism and has become a crucial tool for understanding neurological diseases, cancers and diabetes. High spatial resolution MRSI is needed to characterize lesions, but in practice MRSI is acquired at low resolution due to time and sensitivity restrictions caused by the low metabolite concentrations. Therefore, there is an imperative need for a post-processing approach to generate high-resolution MRSI from low-resolution data that can be acquired fast and with high sensitivity. Deep learning-based super-resolution methods provided promising results for improving the spatial resolution of MRSI, but they still have limited capability to generate accurate and high-quality images. Recently, diffusion models have demonstrated superior learning capability than other generative models in various tasks, but sampling from diffusion models requires iterating through a large number of diffusion steps, which is time-consuming. This work introduces a Flow-based Truncated Denoising Diffusion Model (FTDDM) for super-resolution MRSI, which shortens the diffusion process by truncating the diffusion chain, and the truncated steps are estimated using a normalizing flow-based network. The network is conditioned on upscaling factors to enable multi-scale super-resolution. To train and evaluate the deep learning models, we developed a 1H-MRSI dataset acquired from 25 high-grade glioma patients. We demonstrate that FTDDM outperforms existing generative models while speeding up the sampling process by over 9-fold compared to the baseline diffusion model. Neuroradiologists’ evaluations confirmed the clinical advantages of our method, which also supports uncertainty estimation and sharpness adjustment, extending its potential clinical applications.

Deepest Limits on Scattered Light Emission from the Epsilon Eridani Inner Debris Disk with HST/STIS

Epsilon Eridani is one of the first debris disk systems detected by the Infrared Astronomical Satellite. However, the system has thus far eluded detection in scattered light with no components having been directly imaged. Its similarity to a relatively young solar system combined with its proximity makes it an excellent candidate to further our understanding of planetary system evolution. We present a set of coronagraphic images taken using the Space Telescope Imaging Spectrograph coronagraph on the Hubble Space Telescope at a small inner working angle to detect a predicted warm inner debris disk inside 1″. We used three different postprocessing approaches—nonnegative matrix factorization (NMF), Karhunen–Loève Image Processing (KLIP), and classical reference differential imaging, to best optimize reference star subtraction—and find that NMF performed the best overall while KLIP produced the absolute best contrast inside 1″. We present limits on scattered light from warm dust, with constraints on surface brightness at 6 mJy as−2 at our inner working angle of 0.″6. We also place a constraint of 0.5 mJy as−2 outside 1″, which gives us an upper limit on the brightness for outer disks and substellar companions. Finally, we calculated an upper limit on the dust albedo at ω < 0.487.

A Fresh Revisit of the Issues and Improvements in Impulse Invariance Filter Design for Infinite Impulse Response Filters

The objective of this paper is to first present some issues with impulse invariance filter (IIF) design during the design of digital infinite impulse response (IIR) filters. Engineers are often confused about some inconsistent observations. For instance, if the impulse response of a digital filter is designed using the impulse invariance procedure, then the analog and digital filters’ frequency and step responses are very different. Two simple remedies are presented in this paper. One is a post-processing approach that scales the frequency and step responses of the digital filter by the sampling interval T. Another one is a pre-processing approach that scales the impulse response of the analog filter by T. However, even after these remedies, there is still a steady state bias in the step response of the digital filter for certain cases where there is discontinuity in the analog impulse response. A recommendation is to include a correction term in the digital filter. After that, the steady state bias in the digital filter is then suppressed. Moreover, the MATLAB R2021a command “impinvar” needs to also include a correction term so that the frequency and step responses can be more accurate in the digital filter. Two comparative studies were carried out to compare the improved IIF filter with three competing digital IIR filter design methods. Although the above issues and improvements have been proposed by researchers in the past, many researchers, engineers, and students are still not aware of them. This paper provides a fresh revisit of these issues and improvements by using figures, equations, and examples. Proper credits are also given to those researchers who first pointed out those issues and improvements. It is hoped that through an open access journal, future rediscovery of issues and improvements in IIF can be prevented.

Enhancing streamflow predictions with machine learning and Copula-Embedded Bayesian model averaging

This study proposes a two-step probabilistic post-processing approach that combines different machine learning-based postprocessors through the Copula-Embedded Bayesian Model Averaging (COP-BMA) method to improve the performance of a hydrological model for streamflow predictions. The proposed approach serves a two-fold purpose: firstly, it aims to enhance the accuracy of streamflow predictions, and secondly, it provides probabilistic results that implicitly address the structural uncertainty inherent in different postprocessing methods. We validate our approach by applying it to the Conceptual Functional Equivalent, a lumped hydrologic model utilized for simulating extreme floods during Hurricane Harvey. The validation is conducted across twelve distinct watersheds in the Southeast Texas region at both daily and monthly scales. The findings indicate that the proposed framework significantly enhances the performance of the hydrologic model across the studied watershed. Specifically, on a daily time scale, there is a 23% and 53% improvement in the NSE and KGE respectively, while on a monthly time scale, the framework enhances NSE by 21% and KGE by 25%. Additionally, the MAE (cms) was notably reduced from 4.64 to 2.23 on the daily scale, and from 2.8 to 1.65 on the monthly scale.

Versatile Video Coding-Post Processing Feature Fusion: A Post-Processing Convolutional Neural Network with Progressive Feature Fusion for Efficient Video Enhancement

Advanced video codecs such as High Efficiency Video Coding/H.265 (HEVC) and Versatile Video Coding/H.266 (VVC) are vital for streaming high-quality online video content, as they compress and transmit data efficiently. However, these codecs can occasionally degrade video quality by adding undesirable artifacts such as blockiness, blurriness, and ringing, which can detract from the viewer’s experience. To ensure a seamless and engaging video experience, it is essential to remove these artifacts, which improves viewer comfort and engagement. In this paper, we propose a deep feature fusion based convolutional neural network (CNN) architecture (VVC-PPFF) for post-processing approach to further enhance the performance of VVC. The proposed network, VVC-PPFF, harnesses the power of CNNs to enhance decoded frames, significantly improving the coding efficiency of the state-of-the-art VVC video coding standard. By combining deep features from early and later convolution layers, the network learns to extract both low-level and high-level features, resulting in more generalized outputs that adapt to different quantization parameter (QP) values. The proposed VVC-PPFF network achieves outstanding performance, with Bjøntegaard Delta Rate (BD-Rate) improvements of 5.81% and 6.98% for luma components in random access (RA) and low-delay (LD) configurations, respectively, while also boosting peak signal-to-noise ratio (PSNR).

Bayesian hierarchical model for bias-correcting climate models

Abstract. Climate models, derived from process understanding, are essential tools in the study of climate change and its wide-ranging impacts. Hindcast and future simulations provide comprehensive spatiotemporal estimates of climatology that are frequently employed within the environmental sciences community, although the output can be afflicted with bias that impedes direct interpretation. Post-processing bias correction approaches utilise observational data to address this challenge, although they are typically criticised for not being physically justified and not considering uncertainty in the correction. This paper proposes a novel Bayesian bias correction framework that robustly propagates uncertainty and models underlying spatial covariance patterns. Shared latent Gaussian processes are assumed between the in situ observations and climate model output, with the aim of partially preserving the covariance structure from the climate model after bias correction, which is based on well-established physical laws. Results demonstrate added value in modelling shared generating processes under several simulated scenarios, with the most value added for the case of sparse in situ observations and smooth underlying bias. Additionally, the propagation of uncertainty to a simulated final bias-corrected time series is illustrated, which is of key importance to a range of stakeholders, such as climate scientists engaged in impact studies, decision-makers trying to understand the likelihood of particular scenarios and individuals involved in climate change adaption strategies where accurate risk assessment is required for optimal resource allocation. This paper focuses on one-dimensional simulated examples for clarity, although the code implementation is developed to also work on multi-dimensional input data, encouraging follow-on real-world application studies that will further validate performance and remaining limitations. The Bayesian framework supports uncertainty propagation under model adaptations required for specific applications, providing a flexible approach that increases the scope of data assimilation tasks more generally.

Denoising diffusion post-processing for low-light image enhancement

Low-light image enhancement (LLIE) techniques attempt to increase the visibility of images captured in low-light scenarios. However, as a result of enhancement, a variety of image degradations such as noise and color bias are revealed. Furthermore, each particular LLIE approach may introduce a different form of flaw within its enhanced results. To combat these image degradations, post-processing denoisers have widely been used, which often yield oversmoothed results lacking detail. We propose using a diffusion model as a post-processing approach, and we introduce Low-light Post-processing Diffusion Model (LPDM) in order to model the conditional distribution between under-exposed and normally-exposed images. We apply LPDM in a manner which avoids the computationally expensive generative reverse process of typical diffusion models, and post-process images in one pass through LPDM. Extensive experiments demonstrate that our approach outperforms competing post-processing denoisers by increasing the perceptual quality of enhanced low-light images on a variety of challenging low-light datasets. Source code is available at https://github.com/savvaki/LPDM.

Development and Validation of Four Different Methods to Improve MRI-CEST Tumor pH Mapping in Presence of Fat.

CEST-MRI is an emerging imaging technique suitable for various in vivo applications, including the quantification of tumor acidosis. Traditionally, CEST contrast is calculated by asymmetry analysis, but the presence of fat signals leads to wrong contrast quantification and hence to inaccurate pH measurements. In this study, we investigated four post-processing approaches to overcome fat signal influences and enable correct CEST contrast calculations and tumor pH measurements using iopamidol. The proposed methods involve replacing the Z-spectrum region affected by fat peaks by (i) using a linear interpolation of the fat frequencies, (ii) applying water pool Lorentzian fitting, (iii) considering only the positive part of the Z-spectrum, or (iv) calculating a correction factor for the ratiometric value. In vitro and in vivo studies demonstrated the possibility of using these approaches to calculate CEST contrast and then to measure tumor pH, even in the presence of moderate to high fat fraction values. However, only the method based on the water pool Lorentzian fitting produced highly accurate results in terms of pH measurement in tumor-bearing mice with low and high fat contents.