Post-processing Algorithm Research Articles

Techniques and Strategies to Minimize Radiation Exposure in Pediatric Computed Tomography (CT) Abdominal Examinations: A Review.

As children are more vulnerable to radiation-induced cancers and have longer life expectancies, it is essential to implement strict radiation protection measures in pediatric imaging. This study aimed to review radiation dose-minimizing measures in pediatric abdominal computed tomography (CT) examinations. A systematic search across various databases, including Web of Science, PubMed, SpringerLink, ScienceDirect, and Google Scholar, yielded a total of 7,314 articles. The search used keywords that aligned with the objectives of the study. This study included 77 publications after applying the criteria for inclusion and exclusion. We carefully reviewed these selected articles for compliance with the inclusion criteria and excluded them if they did not meet the specified criteria. Only 12 articles fulfilled the strict criteria. An in-depth review of 12 selected articles demonstrated the radiation dose reduction techniques and strategies, which include prefiltering and post-processing algorithms, careful adjustment of exposure parameters such as tube voltage (kVp) and current (mAs), and the establishment of diagnostic reference levels (DRL). Reduction of radiation exposure in pediatric CT imaging demands multifaceted approaches. To reduce the ionizing radiation dose while still obtaining high-quality diagnostic images, healthcare practitioners should adhere to DRL, adjust exposure factors, implement prefiltration, employ AI, and use post-processing algorithms.

Three-dimensional forward modeling and quantitative assessment of electrode offset effects in ERT

Resistivity data has important applications in geophysical exploration, but the impact of electrode offsets on resistivity response characteristics remains unclear. This study aims to explore the influence of horizontal electrode offset angles and vertical offsets caused by topographical variations on the forward modeling of resistivity data. By analyzing experimental models with different measurement arrays, the paper revealed their influence laws on the buried depth of the target body and resistivity resolution. Utilizing tools like ZondRes3D, we conducted 3D resistivity forward modeling and analyzed the results in detail. It is found that horizontal electrode offsets lead to pseudo-anomalies in the apparent resistivity response, which is related to the offset angles and the number of electrodes. Under different conditions, the horizontal electrode offsets exhibit a “gradient variation” pattern. In addition, topographical variations can also cause distortions and offsets in the apparent resistivity curves and the locations of the anomaly response. Specifically, the measuring lines near the edge of the target bodies are more susceptible to these effects. Based on the comprehensive experimental results, we have drawn several conclusions regarding the impact of electrode offsets and topographical variations, including the effects of offset angles on the pseudo-anomalies, the anomalous response laws under different topographic conditions, as well as anomalous situations under specific angles. These findings provide crucial insights for interpreting resistivity data in geophysical exploration and addressing practical engineering problems, and offer guidance for optimizing measuring line layouts and post-processing terrain correction algorithms.

Quantitative MRI post-processing algorithm and visualization research based on moisture status detection of winter jujube

Quantitative Magnetic Resonance Imaging (qMRI) offers precise measurements of the relaxation characteristics of microstructures, representing a cutting-edge method in non-destructive fruit analysis. This study aims to visualize information on changes in moisture status and distribution at the subcellular level of winter jujube. The 0.5 T nuclear magnetic imaging equipment was utilized to rapidly, non-invasively, and accurately capture the internal relaxation status of the sample with multiple-echo-imaging. By examining the signal and noise data, a simulated dataset was developed to tackle the optimization challenge of estimating parameters for the discrete relaxation model from the multiple-echo-imaging data, especially under conditions of low signal-to-noise ratio (SNR) and in the context of heteroscedastic noise. An optimal weighting factor and the T2NR truncation model have been identified to establish an effective experimental inversion strategy. Subsequently, multiple-echo-imaging can rapidly and stably yielded voxel-level maps under conditions of low signal-to-noise ratio. Utilizing this experimental approach, data from winter jujube was collected and analyzed, facilitating an exploration of water activity (T2 mapping) and associated water content (A2 mapping). Through analyzing winter jujube fruits across two maturity stages, this study elucidates the role of precise quantification and voxel-wise visualization in moisture status detection. The methodology presents an innovative approach for assessing internal moisture distribution in fruits.

Hybrid algorithms for SAR matrix compression and the impact of post-processing on SAR calculation complexity.

This study proposes faster virtual observation point (VOP) compression as well as post-processing algorithms for specific absorption rate (SAR) matrix compression. Furthermore, it shows the relation between the number of channels and the computational burden for VOP-based SAR calculation. The proposed new algorithms combine the respective benefits of two different criteria for determining upper boundedness of SAR matrices by the VOPs. Comparisons of the old and new algorithms are performed for head coil arrays with various channel counts. The new post-processing algorithm is used to post-process the VOP sets of nine arrays, and the number of VOPs for a fixed median relative overestimation is compared. The new algorithms are faster than the old algorithms by a factor of two to more than 10. The compression efficiency (number of VOPs relative to initial number of SAR matrices) is identical. For a fixed median relative overestimation, the number of VOPs increases logarithmically with the number of RF coil channels when post-processing is applied. The new algorithms are much faster than previous algorithms. Post-processing is very beneficial for online SAR supervision of MRI systems with high channel counts, since for a given number of VOPs the relative SAR overestimation can be lowered.

Bayesian mixture models (in)consistency for the number of clusters

AbstractBayesian nonparametric mixture models are common for modeling complex data. While these models are well‐suited for density estimation, recent results proved posterior inconsistency of the number of clusters when the true number of components is finite, for the Dirichlet process and Pitman–Yor process mixture models. We extend these results to additional Bayesian nonparametric priors such as Gibbs‐type processes and finite‐dimensional representations thereof. The latter include the Dirichlet multinomial process, the recently proposed Pitman–Yor, and normalized generalized gamma multinomial processes. We show that mixture models based on these processes are also inconsistent in the number of clusters and discuss possible solutions. Notably, we show that a postprocessing algorithm introduced for the Dirichlet process can be extended to more general models and provides a consistent method to estimate the number of components.

Deep Sketch Vectorization via Implicit Surface Extraction

We introduce an algorithm for sketch vectorization with state-of-the-art accuracy and capable of handling complex sketches. We approach sketch vectorization as a surface extraction task from an unsigned distance field, which is implemented using a two-stage neural network and a dual contouring domain post processing algorithm. The first stage consists of extracting unsigned distance fields from an input raster image. The second stage consists of an improved neural dual contouring network more robust to noisy input and more sensitive to line geometry. To address the issue of under-sampling inherent in grid-based surface extraction approaches, we explicitly predict undersampling and keypoint maps. These are used in our post-processing algorithm to resolve sharp features and multi-way junctions. The keypoint and undersampling maps are naturally controllable, which we demonstrate in an interactive topology refinement interface. Our proposed approach produces far more accurate vectorizations on complex input than previous approaches with efficient running time.

An event-triggered background-oriented schlieren technique combined with dynamic projection using dynamic mirror device

Abstract This paper reports a high-frequency event-triggered background-oriented schlieren (BOS) technique using a combination of an event-triggered camera and dynamic projection. To combine the advantages of continuous and pulsed illumination for the event-triggered camera, a novel background pattern is first developed to incorporate static and dynamic textures generated through projection utilizing a dynamic mirror device (DMD). Then, a specific post-processing algorithm is proposed to reconstruct frames with high time accuracy from event data. This technique allows for the continuous observation and capturing of flows at 4000 frames per second (FPS) with a very low cost, breaking through the short operating times of current high-frame-rate BOS. Moreover, the proposed BOS technique can visualize the flow in real-time with high temporal accuracy, a capability that is challenging to achieve with traditional BOS. To examine the proposed technique, BOS experiments were conducted on a sweeping jet actuator with various inlet pressure. The sweeping dynamics and the start-up process of the sweeping jet at various inlet pressure were visualized and investigated. It is found that the proposed event-triggered BOS can continuously visualize and record the jet flow at a resolution of 1280 × 720 pixels with an equivalent frame rate of up to 4000 FPS. The oscillation frequency of the sweeping jet was found to increase linearly with increasing inlet pressure. It reaches 117.2 Hz at an inlet pressure of 0.5 Mpa. Within the first ten milliseconds or so of start-up, the shape of the sweep was found to be symmetrical. Within the next hundred milliseconds, the jet commences to sweep and saturates. The start-up time of the sweeping jet was quantitatively measured and was observed to decrease with increased inlet pressures.

Enhancing 3D object detection by using neural network with self-adaptive thresholding

Robust 3D object detection remains a pivotal concern in the domain of autonomous field robotics. Despite notable enhancements in detection accuracy across standard datasets, real-world urban environments, characterized by their unstructured and dynamic nature, frequently precipitate an elevated incidence of false positives, thereby undermining the reliability of existing detection paradigms. In this context, our study introduces an advanced post-processing algorithm that modulates detection thresholds dynamically relative to the distance from the ego object. Traditional perception systems typically utilize a uniform threshold, which often leads to decreased efficacy in detecting distant objects. In contrast, our proposed methodology employs a Neural Network with a self-adaptive thresholding mechanism that significantly attenuates false negatives while concurrently diminishing false positives, particularly in complex urban settings. Empirical results substantiate that our algorithm not only augments the performance of 3D object detection models in diverse urban and adverse weather scenarios but also establishes a new benchmark for adaptive thresholding techniques in field robotics.

Enhancing 3D object detection by using neural network with self-adaptive thresholding

Robust 3D object detection remains a pivotal concern in the domain of autonomous field robotics. Despite notable enhancements in detection accuracy across standard datasets, real-world urban environments, characterized by their unstructured and dynamic nature, frequently precipitate an elevated incidence of false positives, thereby undermining the reliability of existing detection paradigms. In this context, our study introduces an advanced post-processing algorithm that modulates detection thresholds dynamically relative to the distance from the ego object. Traditional perception systems typically utilize a uniform threshold, which often leads to decreased efficacy in detecting distant objects. In contrast, our proposed methodology employs a Neural Network with a self-adaptive thresholding mechanism that significantly attenuates false negatives while concurrently diminishing false positives, particularly in complex urban settings. Empirical results substantiate that our algorithm not only augments the performance of 3D object detection models in diverse urban and adverse weather scenarios but also establishes a new benchmark for adaptive thresholding techniques in field robotics.

Delineation of ECG waveform components using encoder-decoder architecture with Postprocess algorithm

Delineation of ECG waveform components using encoder-decoder architecture with Postprocess algorithm

Cross‐Species Segmentation of Animal Prostate Using a Human Prostate Dataset and Limited Preoperative Animal Images: A Sampled Experiment on Dog Prostate Tissue

ABSTRACTIn the development of medical devices and surgical robot systems, animal models are often used for evaluation, necessitating accurate organ segmentation. Deep learning‐based image segmentation provides a solution for automatic and precise organ segmentation. However, a significant challenge in this approach arises from the limited availability of training data for animal models. In contrast, human medical image datasets are readily available. To address this imbalance, this study proposes a fine‐tuning approach that combines a limited set of animal model images with a comprehensive human image dataset. Various postprocessing algorithms were applied to ensure that the segmentation results met the positioning requirements for the evaluation of a medical robot under development. As one of the target applications, magnetic resonance images were used to determine the position of the dog's prostate, which was then used to determine the target location of the robot under development. The MSD TASK5 dataset was used as the human dataset for pretraining, which involved a modified U‐Net network. Ninety‐nine pretrained backbone networks were tested as encoders for U‐Net. The cross‐training validation was performed using the selected network backbone. The highest accuracy, with an IoU score of 0.949, was achieved using the independent validation set from the MSD TASK5 human dataset. Subsequently, fine‐tuning was performed using a small set of dog prostate images, resulting in the highest accuracy of an IoU score of 0.961 across different cross‐validation groups. The processed results demonstrate the feasibility of the proposed approach for accurate prostate segmentation.

Rotating machinery weak fault features enhancement via line-defect phononic crystal sensing

The fault features of rotating machinery at the early degradation stage are always weak as a result of interference from strong noise and irrelevant harmonics. Although traditional acoustic diagnosis techniques have attracted much attention with the merits of rich information and non-contact, extracting meaningful features related to faults in extremely low signal-to-noise ratios (SNRs) has always been a challenging problem. Considering the extraordinary physical properties of phononic crystals (PnCs) and acoustic metamaterials, this study proposes a structural enhancement method for rotating machinery fault diagnosis via line-defect PnC sensing. In contrast to conventional acoustic filtering and enhancement methods, this method can directly filter out noise and enhance fault features in the pre-processing stage of acoustic perception without the need for complex post-processing algorithms. Consequently, the original information of the fault features is preserved intact, thus further increasing the detection limit of current acoustic sensing. Specially, the designed line-defect PnC is parametrically tunable. Combined with the prior knowledge of rotating machinery fault features, such as gears and bearings, it is possible to design structures suitable for enhancing their fault features, which has great potential for practical engineering applications. The enhancement mechanism of line-defect PnC is theoretically described, and numerical simulations are also conducted to verify its ability to detect weak faults in rotating machinery. The experimental results show that by comparison with the variational modal decomposition (VMD)-based method, the proposed method exhibits superior fault feature enhancement performance under low SNR conditions. By systematically combining fault detection methods and acoustic metamaterial sensing, it can be foreseen that the proposed method shows great potential in mechanical equipment fault diagnosis.

MEMS reservoir computing system with stiffness modulation for multi-scene data processing at the edge

Reservoir computing (RC) is a bio-inspired neural network structure which can be implemented in hardware with ease. It has been applied across various fields such as memristors, and electrochemical reactions, among which the micro-electro-mechanical systems (MEMS) is supposed to be the closest to sensing and computing integration. While previous MEMS RCs have demonstrated their potential as reservoirs, the amplitude modulation mode was found to be inadequate for computing directly upon sensing. To achieve this objective, this paper introduces a novel MEMS reservoir computing system based on stiffness modulation, where natural signals directly influence the system stiffness as input. Under this innovative concept, information can be processed locally without the need for advanced data collection and pre-processing. We present an integrated RC system characterized by small volume and low power consumption, eliminating complicated setups in traditional MEMS RC for data discretization and transduction. Both simulation and experiment were conducted on our accelerometer. We performed nonlinearity tuning for the resonator and optimized the post-processing algorithm by introducing a digital mask operator. Consequently, our MEMS RC is capable of both classification and forecasting, surpassing the capabilities of our previous non-delay-based architecture. Our method successfully processed word classification, with a 99.8% accuracy, and chaos forecasting, with a 0.0305 normalized mean square error (NMSE), demonstrating its adaptability for multi-scene data processing. This work is essential as it presents a novel MEMS RC with stiffness modulation, offering a simplified, efficient approach to integrate sensing and computing. Our approach has initiated edge computing, enabling emergent applications in MEMS for local computations.

Deep learning-based automated detection and segmentation of bone and traumatic bone marrow lesions from MRI following an acute ACL tear

IntroductionTraumatic bone marrow lesions (BML) are frequently identified on knee MRI scans in patients following an acute full-thickness, complete ACL tear. BMLs coincide with regions of elevated localized bone loss, and studies suggest these may act as a precursor to the development of post-traumatic osteoarthritis. This study addresses the labour-intensive manual assessment of BMLs by using a 3D U-Net for automated identification and segmentation from MRI scans. MethodsA multi-task learning approach was used to segment both bone and BML from T2 fat‐suppressed (FS) fast spin echo (FSE) MRI sequences for BML assessment. Training and testing utilized datasets from individuals with complete ACL tears, employing a five-fold cross-validation approach and pre-processing involved image intensity normalization and data augmentation. A post-processing algorithm was developed to improve segmentation and remove outliers. Training and testing datasets were acquired from different studies with similar imaging protocol to assess the model's performance robustness across different populations and acquisition conditions. ResultsThe 3D U-Net model exhibited effectiveness in semantic segmentation, while post-processing enhanced segmentation accuracy and precision through morphological operations. The trained model with post-processing achieved a Dice similarity coefficient (DSC) of 0.75 ± 0.08 (mean ± std) and a precision of 0.87 ± 0.07 for BML segmentation on testing data. Additionally, the trained model with post-processing achieved a DSC of 0.93 ± 0.02 and a precision of 0.92 ± 0.02 for bone segmentation on testing data. This demonstrates the approach's high accuracy for capturing true positives and effectively minimizing false positives in the identification and segmentation of bone structures. ConclusionAutomated segmentation methods are a valuable tool for clinicians and researchers, streamlining the assessment of BMLs and allowing for longitudinal assessments. This study presents a model with promising clinical efficacy and provides a quantitative approach for bone-related pathology research and diagnostics.

Time-efficient filtering of imaging polarimetric data by checking physical realizability of experimental mueller matrices.

Imaging Mueller polarimetry has already proved its potential for biomedicine, remote sensing and metrology. The real-time applications of this modality require both video rate image acquisition and fast data post-processing algorithms. First, one must check the physical realizability of the experimental Mueller matrices in order to filter out non-physical data, ie to test the positive semi-definiteness of the 4 × 4 Hermitian coherency matrix calculated from the elements of corresponding Mueller matrix pixel-wise. For this purpose, we compared the execution time for the calculations of i) eigenvalues, ii) Cholesky decomposition, iii) Sylvester's criterion, and iv) coefficients of the characteristic polynomial (two different approaches) of the Hermitian coherency matrix, all calculated for the experimental Mueller matrix images (600 pixels × 700 pixels) of mouse uterine cervix. The calculations were performed using C ++ and Julia programming languages. Our results showed the superiority of the algorithm iv) based on the simplification via Pauli matrices over other algorithms for our dataset. The sequential implementation of latter algorithm on a single core already satisfies the requirements of real-time polarimetric imaging. This can be further amplified by the proposed parallelization (e.g., we achieve a 5-fold speed up on 6 cores). The source codes of the algorithms and experimental data are available at https://github.com/pogudingleb/mueller_matrices.

Long‐Term Monitoring of the Annual Irrigated Cropland Extent in Fragmented and Heterogeneous Arid Landscapes Using Machine Learning and Landsat Imagery

AbstractUnderstanding the long‐term spatiotemporal evolution of irrigated cropland is essential for water resource management, but this knowledge remains elusive in most water‐stressed arid areas. In this study, we introduced an integrated framework for long‐term and field‐scale mapping of annual irrigated cropland in arid and semiarid regions. This framework combines the k‐means algorithm with a semiautomatically trained random forest classifier for initial classification and employs the Bayesian Updating of Land Cover algorithm for subsequent postprocessing. Taking the Heihe River basin in northwestern China as the experimental area, we generated 30‐m annual irrigated cropland maps spanning from 1990 to 2020 based on Landsat imagery and the Google Earth Engine. Comprehensive validation confirmed the reliability of this approach, with the overall accuracy of the annual maps ranging from 83% to 88.3% (mean: 86.6%). Our data set provides an unprecedentedly long‐term and fine‐scale perspective for understanding the continuous spatial and temporal dynamics of irrigated cropland in the Heihe River basin, surpassing previous studies in Central Asia and northwestern China. Notably, a rapid expansion of irrigated areas is occurring in the basin, especially in the water‐stressed midstream and downstream areas. This finding points to potential ecological risks in the foreseeable future due to water resource constraints.

Pixel-wise exposure control for single-shot HDR imaging: A joint optimization approach

Dynamic range is one of the primary limitations that restricts digital image sensors from acquiring more visual information. Current high dynamic range (HDR) imaging techniques entail a trade-off between dynamic range and visual fidelity. In this work, we propose a HDR imaging method, termed PE-HDR, to achieve both a wide dynamic range and high visual fidelity without additional complex post-processing algorithms. Instead of merging a bracketed exposure sequence, the PE-HDR captures HDR images in a single shot using optical coded pixel-wise exposure control, enabling cost-effective and flexible HDR imaging. By incorporating a differentiable optical encoder and a neural network decoder, we jointly optimize the imaging pipeline from light irradiance to digital image signals, thereby refining the pixel-wise exposure control strategy and improving image fidelity. Both simulations and experiments demonstrate that the proposed method achieves a dynamic range of up to 120 dB and an excellent visual fidelity with spatial resolution of up to 2560 × 1600 pixels.

Enhancement of the Consistency and Connectivity of the Ant-tracking Algorithm via 3D U-Net with Dual-Threshold Iteration

Abstract The ant-tracking algorithm is commonly used to extract faults in geological structures. However, obtaining 3D ant-tracking data requires the calculation of various volume attributes. To obtain satisfactory data, it is necessary to iterate through the parameters of these attributes to achieve reasonable continuity. Moreover, due to the numerous parameters involved, the algorithm can produce different outputs with each execution. In this study, we aimed to enhance the performance of the ant-tracking algorithm by combining it with a basic U-Net structure. The input and corresponding labels to the model are "cubic shaped" 3D data segmented from the original 3D seismic volume to facilitate cross-validation with distinct regions. We used the label data as the single ant-tracking result to minimize the operator’s bias by executing the ant-tracking with several different parameters and executors and then taking the average. An evaluative comparison of three different loss functions (MAE, RMSE, and MSE) was conducted to identify the optimal function for training the model. Across five out of the six metrics, MSE function demonstrated predominant performance, leading to its adoption. Apart from this, a significant number of misinterpreted faults led us to propose the post-processing algorithm named "Dual-Threshold Iteration." It was initially used to extract fine blood vessels branching out from large vessels in medical image segmentation and adapted in our work to ensure a high level of continuity while ignoring worthless noise. Comparison with the F1 score and the number of 3D-connected components confirmed that the proposed method could generate reduced bias and smoothly connected fault structures.

A Face-Mask Detection Approach based on YOLO V3 applied for a New Collected Dataset

In the wake of the COVID-19 pandemic, the necessity of face-mask detection systems has become paramount to enforce safety measures and regulations. This paper presents a novel approach for face-mask detection utilizing the YOLOv3 architecture applied to a newly collected dataset. The proposed system aims to accurately detect the presence or absence of face masks in real-time scenarios. The dataset used in this research is meticulously curated to encompass diverse environmental conditions, facial expressions, and variations in mask types and orientations. Each image in the dataset is annotated with bounding boxes indicating the regions of faces and masks, facilitating supervised learning for the YOLOv3 model. Our methodology involves fine-tuning the pre-trained YOLOv3 model on the collected dataset to specialize in face-mask detection. The model is trained using a combination of techniques such as data augmentation, transfer learning, and hyperparameter optimization to enhance its performance. Furthermore, to address challenges such as occlusions, varying lighting conditions, and diverse facial orientations, we incorporate techniques like multi-scale training and post-processing algorithms. These techniques aid in improving the robustness and generalization capability of the model, making it suitable for deployment in real-world scenarios. Keywords: Face Mask Detection, YOLO V3, Computer Vision, Deep Learning, Object Detection, Dataset Collection, COVID-19, Image Processing, Real-time Monitoring.

In-situ particle analysis with heterogeneous background: a machine learning approach

We propose a novel framework that combines state-of-the-art deep learning approaches with pre- and post-processing algorithms for particle detection in complex/heterogeneous backgrounds common in the manufacturing domain. Traditional methods, like size analyzers and those based on dilution, image processing, or deep learning, typically excel with homogeneous backgrounds. Yet, they often fall short in accurately detecting particles against the intricate and varied backgrounds characteristic of heterogeneous particle–substrate (HPS) interfaces in manufacturing. To address this, we've developed a flexible framework designed to detect particles in diverse environments and input types. Our modular framework hinges on model selection and AI-guided particle detection as its core, with preprocessing and postprocessing as integral components, creating a four-step process. This system is versatile, allowing for various preprocessing, AI model selections, and post-processing strategies. We demonstrate this with an entrainment-based particle delivery method, transferring various particles onto substrates that mimic the HPS interface. By altering particle and substrate properties (e.g., material type, size, roughness, shape) and process parameters (e.g., capillary number) during particle entrainment, we capture images under different ambient lighting conditions, introducing a range of HPS background complexities. In the preprocessing phase, we apply image enhancement and sharpening techniques to improve detection accuracy. Specifically, image enhancement adjusts the dynamic range and histogram, while sharpening increases contrast by combining the high pass filter output with the base image. We introduce an image classifier model (based on the type of heterogeneity), employing Transfer Learning with MobileNet as a Model Selector, to identify the most appropriate AI model (i.e., YOLO model) for analyzing each specific image, thereby enhancing detection accuracy across particle–substrate variations. Following image classification based on heterogeneity, the relevant YOLO model is employed for particle identification, with a distinct YOLO model generated for each heterogeneity type, improving overall classification performance. In the post-processing phase, domain knowledge is used to minimize false positives. Our analysis indicates that the AI-guided framework maintains consistent precision and recall across various HPS conditions, with the harmonic mean of these metrics comparable to those of individual AI model outcomes. This tool shows potential for advancing in-situ process monitoring across multiple manufacturing operations, including high-density powder-based 3D printing, powder metallurgy, extreme environment coatings, particle categorization, and semiconductor manufacturing.