Mean Distance Error Research Articles

Accurate Acupoint Localization in 2D Hand Images: Evaluating HRNet and ResNet Architectures for Enhanced Detection Performance

Introduction: This research assesses HRNet and ResNet architectures for their precision in localizing hand acupoints on 2D images, which is integral to automated acupuncture therapy. Objectives: The primary objective was to advance the accuracy of acupoint detection in traditional Korean medicine through the application of these advanced deep-learning models, aiming to improve treatment efficacy. Background: Acupoint localization in traditional Korean medicine is crucial for effective treatment, and the study aims to enhance this process using advanced deep-learning models. Methods: The study employs YOLOv3, YOLOF, and YOLOX-s for object detection within a top-down framework, comparing HRNet and ResNet architectures. These models were trained and tested using datasets annotated by technicians and their mean values, with performance evaluated based on Average Precision at two IoU thresholds. Results: HRNet consistently demonstrated lower mean distance errors across various acupoints compared to ResNet, particularly at a 256x256 pixel resolution. Notably, the HRNet-w48 model surpassed human annotators, including medical experts, in localization accuracy. Conclusion: HRNet's superior performance in acupoint localization suggests its potential to improve the precision and efficacy of acupuncture treatments. The study highlights the promising role of machine learning in enhancing traditional medical practices and underscores the importance of accurate acupoint localization in clinical acupuncture.

Underwater Line Monitoring Using Optimally Placed Inclinometers

Underwater monitoring presents challenges related to maintaining a continuous power supply and communication, necessitating the use of a smaller number of sensors to effectively cover the entire line. An underwater line tracking method is proposed to evaluate global behaviors and stresses in real time. The method employs angles at several points on the line, as well as displacements and curvatures at both ends. In this method, any line displacement, angle, and curvature are expressed as Fourier series, and Fourier coefficients are obtained by utilizing sensor data. Then, the behavior of any line location is assessed. In addition, to reduce the number of sensors and improve accuracy, optimal inclinometer locations are determined by a genetic algorithm. The proposed line tracking algorithm was validated through two numerical examples; one with an inclined tunnel and one with a marine steel catenary riser attached to a Floating Production Storage and Offloading (FPSO) vessel. Through these examples, the proposed algorithm was proven to capture global behaviors accurately when optimally located sensors are used. In the riser monitoring case, the optimized sensor placement with eight intermediate sensors achieved an average mean distance error of 1.91 m, which is lower than the 2.65 m error obtained with ten intermediate sensors without optimization.

The Diagnostic Utility of Cervical Spine Proprioception for Adolescent Concussion.

Cervical spine proprioception may be impaired after concussion. Our objective was to determine the diagnostic utility of cervical spine proprioception for adolescent concussion. Cross-sectional. Research laboratory. Adolescents ≤18 days of concussion and uninjured controls. N/A. Head repositioning accuracy (HRA) testing, a measure of cervical spine proprioception. The HRA test involved patients relocating their head back to a neutral starting position with eyes closed after maximal cervical spine flexion, extension, and right and left rotations. The overall HRA error score was the mean error (distance from the starting point to self-reported return to neutral) across 12 trials: 3 trials in each direction. We used t-tests to compare group means and logistic regression (outcome = group, predictor = HRA, covariates) to calculate odds ratios. We used a receiver operator characteristic curve to evaluate area under the curve (AUC) and calculate the optimal HRA cutpoint to distinguish concussion from controls. We enrolled and tested 46 participants with concussion (age = 15.8 ± 1.3 years, 59% female, mean = 11.3 ± 3.3 days postconcussion) and 83 uninjured controls (age = 16.1 ± 1.4 years, 88% female). The concussion group had significantly worse HRA than controls (4.3 ± 1.6 vs 2.9 ± 0.7 degrees, P < 0.001, Cohen d = 1.19). The univariable HRA model AUC was 0.81 (95% CI = 0.73, 0.90). After adjusting for age, sex, and concussion history, the multivariable model AUC improved to 0.85 (95% CI = 0.77, 0.92). The model correctly classified 80% of participants as concussion/control at a 3.5-degree cutpoint. Adolescents with concussion demonstrated worse cervical spine proprioception than uninjured controls. Head repositioning accuracy may offer diagnostic utility for subacute concussion.

Automated multivolume placental reconstruction using three-dimensional power Doppler ultrasound and infrared camera tracking.

Placental insufficiency contributes to many obstetric pathologies however there is no bedside clinical tool to evaluate placental perfusion. We have developed a method to acquire multiple three-dimensional power Doppler ultrasound (3D PD-US) volumes of placental vasculature with infrared camera tracking providing global coordinates. These are automatically reconstructed ('stitched') into a model of the entire placenta. The purpose of the study was to evaluate the accuracy of automated reconstruction in an US phantom and apply this technique to human placentas. A custom-designed acrylic phantom was constructed with dimensions mimicking a third-trimester placenta, containing 12 quadrilateral towers of varying heights submersed in tissue-mimicking solution. Triplicated three-dimensional ultrasound volumes of this phantom were acquired at three different acquisition angles using infrared camera tracking. Data was transformed into a three-dimensional cartesian volume and automatically stitched. A single-centre, cross-sectional feasibility study was conducted on uncomplicated second-to-third trimester singleton pregnancies using standardised obstetric settings. Multiple 3D PD-US and grayscale volumes of the placenta were acquired with infrared camera tracked coordinates. Volumes were stitched to create a model of placental vasculature. 6 phantom datasets were reconstructed at each of 3 volume angles with a median of 9 volumes required. Perfect volume alignment occurred in 66.7% of 648 datapoints. Mean distance error for volume misalignment was 2.92mm. Measurements of 210 distances in each stitched volume (2160 total distances) differed an average of 1.51mm from true measurements. These compare favourably with recent literature, though for a substantially larger phantom. 17 participants were scanned with 92% reconstruction success per placental volume set and 100% participant achievability. Median reconstruction time was 10 minutes. Placental vasculature was qualitatively assessed to be present, continuous, and detailed throughout. Volume measurement of entire segmented placentas was highly repeatable (ICC 0.96). We present an automated method to model the entire structure and vasculature of second-to-third trimester placentas, with verified accuracy and clinical feasibility for grayscale and power Doppler. This study builds the foundation to develop a practical screening tool for detecting placental insufficiency, and expansion to adult organ perfusion evaluation. This article is protected by copyright. All rights reserved.

Evaluation of a Fully Digital, In-House Virtual Surgical Planning Workflow for Bimaxillary Orthognathic Surgery

The advantages of virtual surgical planning (VSP) for orthognathic surgery are clear. Previous studies have evaluated in-house VSP; however, few fully digital, in-house protocols for orthognathic surgery have been studied. The purpose of this study was to evaluate the difference between the virtual surgical plan and actual surgical outcome for orthognathic surgery using a fully digital, in-house VSP workflow. This is a prospective cohort study from September 2020 to November 2022 of patients at the Victoria General Hospital in Halifax, NS, Canada who underwent bimaxillary orthognathic surgery. Patients were excluded if they had previously undergone orthognathic surgery or were diagnosed with a craniofacial syndrome. The primary outcome variables were the mean 3-dimensional (3D) (Euclidean) distance error, as well as mean error and mean absolute error in the transverse (x axis), vertical (y axis), and anterior-posterior (z axis) dimensions. Covariates included age, sex, and surgical sequence (mandible-first or maxilla-first). The primary outcome was tested using Z and t critical value confidence intervals. The P value was set at .05. The 3D distance error for mandible-first and maxilla-first groups was compared using a 2-sample t-test as well as analysis of variance. The study sample included 52 subjects (24 males and 28 females) with a mean age of 27.7 (± 12.1) years. Forty three subjects underwent mandible-first surgery and 9 maxilla-first surgery. The mean absolute distance error was largest in the anterior-posterior dimension for all landmarks (except posterior nasal spine, left condyle, and gonion) and exceeded the threshold for clinical acceptability (2mm) in 16 of 23 landmarks. Additionally, mean distance error in the anterior-posterior dimension was negative for all landmarks, indicating deficient movement in that direction. The effect of surgical sequence on 3D distance error was not statistically significant (P=.37). In general, the largest contributor to mean 3D distance error was deficient movement in the anterior-posterior direction. Otherwise, mean absolute distance error in the vertical and transverse dimensions was clinically acceptable (< 2mm). These findings were felt to be valuable for treatment planning purposes when using a fully digital, in-house VSP workflow.

Low-Power Approximate Unsigned and Signed Multipliers with Configurable Error Recovery

In the field of Digital Signal Processing and similar applications, Approximate circuits are being explored as a means to enhance performance and energy efficiency by sacrificing a degree of accuracy. Within these circuits, Multipliers play a crucial role and are being investigated for their potential impact on overall system optimization. In this paper, a novel approximate multiplier with a low power consumption and a short critical path is proposed for high-performance DSP applications. This multiplier leverages a newly designed approximate adder that limits its carry propagation to the nearest neighbours for fast partial product accumulation. Different levels of accuracy can be achieved by using either OR gates or the proposed approximate adder in a configurable error recovery. The multipliers using these two error reduction strategies are referred to as Approximate Multiplier 1 (AM1) and Approximate Multiplier 2 (AM2), respectively. Both AM1 and AM2 have a low mean error distance, i.e., most of the errors are not significant in magnitude. Compared with a Unsigned multiplication multiplier, with the signed multiplication. The signed multiplication tends to have lower power consumption is observed. By utilizing an appropriate error recovery, the proposed approximate multipliers achieve similar processing accuracy as traditional exact multipliers, but with significant improvements in power.

A Hierarchy-Aware Geocoding Model Based on Cross-Attention within the Seq2Seq Framework

Geocoding converts unstructured geographic text into structured spatial data, which is crucial in fields such as urban planning, social media spatial analysis, and emergency response systems. Existing approaches predominantly model geocoding as a geographic grid classification task but struggle with the output space dimensionality explosion as the grid granularity increases. Furthermore, these methods generally overlook the inherent hierarchical structure of geographical texts and grids. In this paper, we propose a hierarchy-aware geocoding model based on cross-attention within the Seq2Seq framework, incorporating S2 geometry to model geocoding as a task for generating grid labels and predicting S2 tokens (labels of S2 grids) character-by-character. By incorporating a cross-attention mechanism into the decoder, the model dynamically perceives the address contexts at the hierarchical level that are most relevant to the current character prediction based on the input address text. Results show that the proposed model significantly outperforms previous approaches across multiple metrics, with a median and mean distance error of 41.46 m and 93.98 m, respectively. Furthermore, our method achieves superior results compared to others in regions with sparse data distribution, reducing the median and mean distance error by 16.27 m and 7.52 m, respectively, suggesting that our model has effectively mitigated the issue of insufficient learning in such regions.

Proprioceptive acuity for landmarks on the hand and digits.

Previous work using visually guided reaches to localize landmarks on a hidden hand has suggested that proprioceptive acuity for hand targets is low and representation of hand dimensions is highly distorted (e.g., hand width estimated to be 60% wider than actual hand width). We re-examined these issues using a pure proprioceptive task in which 20 blindfolded subjects reached in a single movement without terminal corrections to touch the right index-tip to landmarks of the left hand placed in various locations in 3D space. Subjects were also tested with vision allowed to estimate minimal errors. Based on previous reports of high proprioceptive acuity for some hand landmarks, we hypothesized that the proprioceptive representation of the hand was much less distorted than described previously and that errors were not correlated with target hand location. Mean distance errors in proprioceptively guided reaches to the landmarks averaged less than 3cm and were only 0.5-1.3cm larger than when vision was allowed. Errors were not correlated with hand location in most subjects. Distortions of hand width averaged less than 20% wider than actual width and were not correlated with hand location in most subjects. We conclude that relatively accurate proprioceptive awareness of locations of hand/digit structures and dimensions is available for use in control of hand movements, which are executed largely subconsciously. Studying acuity of proprioception using conscious perceptual tasks and involving vision may not provide accurate measures of proprioceptive acuity as used by the motor system.

Performance Evaluation of LMI Based on Low-Frequency Three-Dimensional Total Lightning Flash Location Data

At present, there is still some uncertainty in the evaluation of the performance of the Fengyun 4A Lightning Mapping Imager (LMI), which is mainly limited by the detection performance of the reference detection system and the suitability of the evaluation method. In this paper, a one-to-one performance evaluation of the LMI was performed based on total lightning flash data from the lightning Low-Frequency Electric field Detection Array (LFEDA). It was found that there were significant systematic biases in the discharge results detected via LMI, with a median of −0.946 s, −0.0817°, and −0.0245° in time bias, longitude bias, and latitude bias, respectively. The evaluation results after removing the systematic biases indicated that the relative detection efficiency for flashes of LMI was 17.6%, the mean and median time errors were both 0.647 s, and the mean and median distance errors were 6.09 km and 5.02 km, respectively. The relative detection efficiency for groups of LMI was 9.8%, the mean and median time errors were 0.674 s and 0.660 s, and the mean and median distance errors were 7.19 km and 6.54 km, respectively. The detection efficiency of LMI for both flashes and groups at nighttime was significantly higher than its detection efficiency during the daytime. The relative detection efficiency for flashes of LMI at nighttime was 26.5%, while during the daytime it was 14.4%. The relative detection efficiency for groups of LMI at nighttime was 16.2%, while during the daytime it was only 7.4%. The spatial accuracy for both flashes and groups was always better during the daytime than at nighttime.

Automatic 3D Lamina Curve Extraction from Freehand 3D Ultrasound Data using Sequential Localization Recurrent Convolutional Networks.

Freehand 3D ultrasound imaging is emerging as a promising modality for regular spine exams due to its non-invasiveness and affordability. The laminae landmarks play a critical role in depicting the 3D shape of the spine. However, the extraction of the 3D lamina curves from transverse ultrasound sequences presents a challenging task, primarily attributed to the presence of diverse contrast variations, imaging artifacts, the complex surface of vertebral bones, and the difficulties associated with probe manipulation. This paper proposes Sequential Localization Recurrent Convolutional Networks (SL-RCN), a novel deep learning model that takes the contextual relationships into account and embeds the transformation matrix feature as a 3D knowledge base to enhance accurate ultrasound sequence analysis. The assessment involved the analysis of 3D ultrasound sequences obtained from 10 healthy adult human participants, covering both the lumbar and thoracic regions. The performance of SL-RCN is evaluated through 7-fold cross-validation, employing the leave-one-participant-out strategy. The validity of the AI model training is assessed on test data from 3 participants. Normalized Discrete Fréchet Distance (NDFD) is employed as the main metric to evaluate the disparity of the extracted 3D lamina curves. In contrast to our previous 2D image analysis method, SL-RCN generates reduced left/right mean distance errors from 1.62/1.63mm to 1.41/1.40mm, and NDFDs from 0.5910/0.6389 to 0.4276/0.4567. The increase in the mean NDFD value from 7-fold cross-validation to the test-data experiment is less than 0.05. The experiments demonstrate the SL-RCN's capability in extracting accurate paired smooth lamina landmark curves.

Automatic cardiothoracic ratio calculation based on lung fields abstracted from chest X-ray images without heart segmentation.

The cardiothoracic ratio (CTR) based on postero-anterior chest X-rays (P-A CXR) images is one of the most commonly used cardiac measurement methods and an indicator for initially evaluating cardiac diseases. However, the hearts are not readily observable on P-A CXR images compared to the lung fields. Therefore, radiologists often manually determine the CTR's right and left heart border points of the adjacent left and right lung fields to the heart based on P-A CXR images. Meanwhile, manual CTR measurement based on the P-A CXR image requires experienced radiologists and is time-consuming and laborious. Based on the above, this article proposes a novel, fully automatic CTR calculation method based on lung fields abstracted from the P-A CXR images using convolutional neural networks (CNNs), overcoming the limitations to heart segmentation and avoiding errors in heart segmentation. First, the lung field mask images are abstracted from the P-A CXR images based on the pre-trained CNNs. Second, a novel localization method of the heart's right and left border points is proposed based on the two-dimensional projection morphology of the lung field mask images using graphics. The results show that the mean distance errors at the x-axis direction of the CTR's four key points in the test sets T1 (21 × 512 × 512 static P-A CXR images) and T2 (13 × 512 × 512 dynamic P-A CXR images) based on various pre-trained CNNs are 4.1161 and 3.2116 pixels, respectively. In addition, the mean CTR errors on the test sets T1 and T2 based on four proposed models are 0.0208 and 0.0180, respectively. Our proposed model achieves the equivalent performance of CTR calculation as the previous CardioNet model, overcomes heart segmentation, and takes less time. Therefore, our proposed method is practical and feasible and may become an effective tool for initially evaluating cardiac diseases.

Intelligent electromagnetic navigation system for robot-assisted intraoral osteotomy in mandibular tumor resection: a model experiment.

Mandibular tumor surgery necessitates precise osteotomies based on tumor boundaries; however, conventional osteotomies often lack accuracy in predicting osteotomy positions and planes, potentially leading to excessive resection of normal bone tissues or residual tumors, thus compromising postoperative quality of life and clinical outcomes. Robotic-assisted surgery (RAS) augmented with artificial intelligence (AI) offers precise localization capabilities, aiding surgeons in achieving accurate osteotomy positioning. This study aimed to evaluate the feasibility and accuracy of a robotic magnetic navigation system for positioning and osteotomy in an intraoral surgical trial of a mandibular tumor model. Patient computed tomography (CT) imaging data of mandibular chin and body tumors were utilized to create 3D printed models, serving as study subjects for mandibular tumor resection. Ten pairs of models were printed for the experimental and control groups. The experimental group (EG) underwent osteotomy using a robot-assisted surgical navigation system, performing osteotomy under robotic navigation following alignment based on preoperative design. The control group (CG) underwent traditional surgery, estimating osteotomy position empirically according to preoperative design. Postoperative CT scans were conducted on both models, and actual postoperative results were compared to preoperative design. Osteotomy accuracy was evaluated by positional and angular errors between preoperatively designed and actual osteotomy planes. For ten randomly selected spots on the left and right sides, respectively, the EG group had mean distance errors of 0.338mm and 0.941mm. These values were obtained from the EG group. In the EG group, on the left side, the mean angular errors were 14.741 degrees, while on the right side, they were 13.021 degrees. For the 10 randomly selected spots on the left and right sides, respectively, the CG had mean distance errors of 1.776mm and 2.320mm. This is in contrast to the results obtained by the EG. It was determined that the left side had a mean angle error of 16.841 degrees, while the right side had an error of 18.416 degrees in the CG group. The above results indicated significantly lower point errors of bilateral osteotomy planes in the experimental group compared to the control group. This study demonstrates the feasibility of electromagnetic navigation robot-assisted intraoral osteotomy for mandibular tumors and suggests that this approach can enhance the precision of clinical surgery.

Identifying wind turbines from multiresolution and multibackground remote sensing imagery

The wind energy industry has expanded in recent years. Promotion of the Sustainable Development Goals (SDGs) is expected to further increase the scale of the wind energy industry. Determining the location and quantity of wind turbines is crucial for monitoring the development status of the wind energy industry and evaluating wind energy production. In this study, we propose a method for simultaneously detecting and positioning wind turbines in remote sensing images, namely, Wind Turbine YOLO (WT-YOLO), based on the You Only Look Once version 5 model (YOLOv5). The wind turbine hub, base, and shadow hub are treated as key points in the proposed method. Regression terms are incorporated into the head of the YOLOv5 basic framework to predict the location of these three key points. The base point is utilized to determine the exact position of the wind turbine. A multibackground and multiresolution wind-turbine image dataset is constructed by sampling high-resolution images from Google Earth. The WT-YOLO method outperforms existing methods in both wind turbine detection and positioning on different types of land cover backgrounds and multiresolution images of the constructed dataset. In the spatial resolution range of 0.6 m to 5.4 m, WT-YOLO exhibits enhanced wind-turbine detection, where the average precision (AP) is 5.92 % to 15.43 % higher than that of existing wind-turbine detection methods. Wind-turbine positioning by WT-YOLO has a mean distance error (MDE) that is 16.06 m to 21.59 m lower than that of existing wind-turbine positioning methods. A comparative analysis showed that the shadow and the three key points are effective features for wind-turbine detection. The proposed WT-YOLO model can support detection, positioning and counting for wind turbines worldwide.

Advancing construction site workforce safety monitoring through BIM and computer vision integration

Ensuring a safe work environment is crucial for construction projects. It is essential that workforce monitoring is both efficient and non-intrusive to the ongoing construction activities. This paper introduces a method that integrates building information modeling (BIM) and computer vision to monitor workforce safety hazards at construction sites in real time. Despite the rising adoption of BIM and computer vision individually within the construction sector, the potential of their integrated application as a cohesive system for workforce safety monitoring remains unexplored. While BIM provides rich 3D semantic information about the construction site, computer vision captures real-time field data. The system was tested using a realistic construction simulation, and the accuracy of the position estimate was evaluated in a real-world interior environment, yielding a mean error distance (MED) of 13.2 cm. Overall, the findings have substantial significance for the construction industry to help minimize accidents and enhance overall worker safety.

Out-of-the-box calving-front detection method using deep learning

Abstract. Glaciers across the globe react to the changing climate. Monitoring the transformation of glaciers is essential for projecting their contribution to global mean sea level rise. The delineation of glacier-calving fronts is an important part of the satellite-based monitoring process. This work presents a calving-front extraction method based on the deep learning framework nnU-Net, which stands for no new U-Net. The framework automates the training of a popular neural network, called U-Net, designed for segmentation tasks. Our presented method marks the calving front in synthetic aperture radar (SAR) images of glaciers. The images are taken by six different sensor systems. A benchmark dataset for calving-front extraction is used for training and evaluation. The dataset contains two labels for each image. One label denotes a classic image segmentation into different zones (glacier, ocean, rock, and no information available). The other label marks the edge between the glacier and the ocean, i.e., the calving front. In this work, the nnU-Net is modified to predict both labels simultaneously. In the field of machine learning, the prediction of multiple labels is referred to as multi-task learning (MTL). The resulting predictions of both labels benefit from simultaneous optimization. For further testing of the capabilities of MTL, two different network architectures are compared, and an additional task, the segmentation of the glacier outline, is added to the training. In the end, we show that fusing the label of the calving front and the zone label is the most efficient way to optimize both tasks with no significant accuracy reduction compared to the MTL neural-network architectures. The automatic detection of the calving front with an nnU-Net trained on fused labels improves from the baseline mean distance error (MDE) of 753±76 to 541±84 m. The scripts for our experiments are published on GitHub (https://github.com/ho11laqe/nnUNet_calvingfront_detection, last access: 20 November 2023). An easy-access version is published on Hugging Face (https://huggingface.co/spaces/ho11laqe/nnUNet_calvingfront_detection, last access: 20 November 2023).

Design and Analysis of High Performance Heterogeneous Block-based Approximate Adders

Approximate computing is an emerging paradigm to improve the power and performance efficiency of error-resilient applications. As adders are one of the key components in almost all processing systems, a significant amount of research has been carried out toward designing approximate adders that can offer better efficiency than conventional designs; however, at the cost of some accuracy loss. In this article, we highlight a new class of energy-efficient approximate adders, namely, Heterogeneous Block-based Approximate Adders (HBAAs), and propose a generic configurable adder model that can be configured to represent a particular HBAA configuration. An HBAA, in general, is composed of heterogeneous sub-adder blocks of equal length, where each sub-adder can be an approximate sub-adder and have a different configuration. The sub-adders are mainly approximated through inexact logic and carry truncation. Compared to the existing design space, HBAAs provide additional design points that fall on the Pareto-front and offer a better quality-efficiency tradeoff in certain scenarios. Furthermore, to enable efficient design space exploration based on user-defined constraints, we propose an analytical model to efficiently evaluate the Probability Mass Function (PMF) of approximation error and other error metrics, such as Mean Error Distance (MED), Normalized Mean Error Distance (NMED), and Error Rate (ER) of HBAAs. The results show that HBAA configurations can provide around 15% reduction in area and up to 17% reduction in energy compared to state-of-the-art approximate adders.

DAEM: A Data- and Application-Aware Error Analysis Methodology for Approximate Adders

Approximate adders are some of the fundamental arithmetic operators that are being employed in error-resilient applications, to achieve performance/energy/area gains. This improvement usually comes at the cost of some accuracy and, therefore, requires prior error analysis, to select an approximate adder variant that provides acceptable accuracy. Most of the state-of-the-art error analysis techniques for approximate adders assume input bits and operands to be independent of one another, while some also assume the operands to be uniformly distributed. In this paper, we analyze the impact of these assumptions on the accuracy of error estimation techniques, and we highlight the need to address these assumptions, to achieve better and more realistic quality estimates. Based on our analysis, we propose DAEM, a data- and application-aware error analysis methodology for approximate adders. Unlike existing error analysis models, we neither assume the adder operands to be uniformly distributed nor assume them to be independent. Specifically, we use 2D joint input probability mass functions (PMFs), populated using sample data, in order to incorporate the data and application knowledge in the analysis. These 2D joint input PMFs, along with 2D error maps of approximate adders, are used to estimate the error PMF of an adder network. The error PMF is then utilized to compute different error measures, such as the mean squared error (MSE) and mean error distance (MED). We evaluate the proposed error analysis methodology on audio and video processing applications, and we demonstrate that our methodology provides error estimates having a better correlation with the simulation results, as compared to the state-of-the-art techniques.

Efficient and low-cost approximate multipliers for image processing applications

The image processing can be implemented in an error-tolerant mode with acceptable accuracy. In digital image processing, the adders and multipliers are the most important components. In this paper, two approximate multipliers based on an approximate 4:2 compressor and two 4-bit full adders with minimizing computational costs are proposed. The first and second proposed approximate multipliers are designed by using the proposed imprecise 4-bit full adders and 4:2 compressor, respectively. In the structure of the proposed multipliers, we have adders for the addition of two multiple-bit numbers. In the adder structure, the full adders are designed based on two modified half adders and one NAND gate. The proposed approximate 4:2 compressor and two approximate 4-bit full adders are compared from hardware and accuracy point of view such as gate count, area-delay product (ADP), error rate (ER), mean error distance (MED), mean relative error distance (MRED), and normalized error distance (NED). The efficacy of proposed multipliers in image processing applications such as image multiplication is evaluated using MATLAB. The results show the proposed approximate multipliers are comparable in terms of area, delay, and mean structural similarity index metric (MSSIM) parameter with other works.

Using a New Deep Learning Method for 3D Cephalometry in Patients With Hemifacial Microsomia.

Deep learning algorithms based on automatic 3D cephalometric marking points about people without craniomaxillofacial deformities have achieved good results. However, there has been no previous report about hemifacial microsomia (HFM). The purpose of this study is to apply a new deep learning method based on a 3D point cloud graph convolutional neural network to predict and locate landmarks in patients with HFM based on the relationships between points. The authors used a PointNet++ model to investigate the automatic 3D cephalometry. And the mean distance error (MDE) of the center coordinate position and the success detection rate (SDR) were used to evaluate the accuracy of systematic labeling. A total of 135 patients were enrolled. The MDE for all 32 landmarks was 1.46 ± 1.308 mm, and 10 landmarks showed SDRs at 2 mm over 90%, and only 4 landmarks showed SDRs at 2 mm under 60%. Compared with the manual reproducibility, the standard distance deviation and coefficient of variation values for the MDE of the artificial intelligence system was 0.67 and 0.43, respectively. In summary, our training sets were derived from HFM computed tomography to achieve accurate results. The 3D cephalometry system based on the graph convolutional network algorithm may be suitable for the 3D cephalometry system in HFM cases. More accurate results may be obtained if the HFM training set is expanded in the future.

Joint Calibration Method for Robot Measurement Systems.

Robot measurement systems with a binocular planar structured light camera (3D camera) installed on a robot end-effector are often used to measure workpieces' shapes and positions. However, the measurement accuracy is jointly influenced by the robot kinematics, camera-to-robot installation, and 3D camera measurement errors. Incomplete calibration of these errors can result in inaccurate measurements. This paper proposes a joint calibration method considering these three error types to achieve overall calibration. In this method, error models of the robot kinematics and camera-to-robot installation are formulated using Lie algebra. Then, a pillow error model is proposed for the 3D camera based on its error distribution and measurement principle. These error models are combined to construct a joint model based on homogeneous transformation. Finally, the calibration problem is transformed into a stepwise optimization problem that minimizes the sum of the relative position error between the calibrator and robot, and analytical solutions for the calibration parameters are derived. Simulation and experiment results demonstrate that the joint calibration method effectively improves the measurement accuracy, reducing the mean positioning error from over 2.5228 mm to 0.2629 mm and the mean distance error from over 0.1488 mm to 0.1232 mm.