Radial Error Research Articles

Corrections on radial edge errors of charge and current density deposition for two-dimension radial-axial particle-in-cell simulations

Corrections on radial edge errors of charge and current density deposition for two-dimension radial-axial particle-in-cell simulations

Tsunami Data Assimilation Using High‐Frequency Radar‐Derived Surface Currents by Considering Beam Angle‐Dependent Measurement Error Distributions

AbstractThe application of high‐frequency radar as an instrument for assimilating tsunami‐induced current fields is garnering increasing interest. The performance of surface current velocity measurements depends on the azimuthal differences between the crossing radar beams at the measurement points. This study aimed to incorporate the measurement error distributions of the east‐west and north‐south velocity components into tsunami data assimilation based on an optimal interpolation method, assuming Gaussian noise with the time‐invariant and a uniform standard deviation (STD = 5 cm/s) of radial velocity measurements. Through the empirical orthogonal function (EOF) analysis of radar‐derived surface currents in the Kii Channel, Japan, the velocities reconstructed using higher modes (EOFs 16–274) were associated with measurement errors, portraying nonuniform distribution depending on the crossing beam angle of two radar beams. Based on independent fifteen‐time assimilation experiments for two different tsunami scenarios, for a uniform water depth of 500 m, we observed a significant improvement of up to 29% and 0.9% in the assimilation performance (on average) over the along‐coast stations for scenarios with 1‐ and 5‐m maximum initial sea surface heights, respectively. The measurement errors dependent on the crossing beam angle reduced the error‐induced tsunamis, resulting in stable assimilations, with lower STDs in the fifteen‐time assimilation performances. When the STD of Gaussian noise varies with time, it is important to consider the temporal change in the radial velocity measurement errors and/or noise‐filtering techniques, to maintain a certain level of noise intensity.

Direct torque control of bearingless induction motor based on super-twisting sliding mode control

A bearingless induction motor (BIM) is an innovative magnetic levitation motor that incorporates torque system and suspension systems. Aiming to improve the anti-interference and levitation performance of the BIM, a second-order super-twisting sliding mode direct torque control strategy (STSM-DTC) is proposed. Initially, the basic principles and convergence conditions of the super-twisting algorithm are analysed. Subsequently, corresponding sliding mode surfaces are constructed according to speed, torque, and stator flux error, and the corresponding super-twisting sliding mode controllers (STSMC) are designed in the torque system. The STSMCs are designed based on the rotor radial displacement error and further enhanced by incorporating a differential loop in the suspension system. The hyperbolic tangent function tanh(x) is used as the switching function. Simulation and experimental of the BIM STSM-DTC system show that the STSM-DTC strategy effectively improves the anti-interference ability and suspension performance of the BIM, and show better dynamic response.

Error Analysis of Non-Time-Synchronized Lightning Positioning Method

Since the non-time-synchronized lightning positioning method does not rely on the time synchronization of the stations in the positioning system, it eliminates the errors arising from the pursuit of time synchronization and potentially achieves higher positioning accuracy. This paper provides a comprehensive overview of the errors present in the three-dimensional lightning positioning system. It compares the results of traditional positioning methods with those of non-time-synchronized lightning positioning algorithms. Subsequently, a simulation analysis of the positioning errors is conducted specifically for the non-time-synchronized lightning positioning method. The results show that (1) the non-time-synchronized lightning positioning method exhibits greater errors when utilizing two randomly positioned radiation sources for location determination. Consequently, the resulting positioning outcomes only provide a general overview of the lightning discharge. (2) The positioning outcomes resemble those of the traditional method when employing a fixed-coordinate beacon point. However, the errors in the three-dimensional positional coordinates of these fixed-coordinate beacon points significantly impact the deviations in the positioning results. This impact is positively correlated with the positional error of the beacon point, considering both the orientation and magnitude. (3) Similarly to the traditional method, the farther away from the center of the positioning network, the larger the radial error. (4) The spatial position of the selected fixed-coordinate beacon point has little influence on the error.

Cone-beam CT landmark detection for measuring basal bone width: a retrospective validation study

BackgroundAccurate assessment of basal bone width is essential for distinguishing individuals with normal occlusion from patients with maxillary transverse deficiency who may require maxillary expansion. Herein, we evaluated the effectiveness of a deep learning (DL) model in measuring landmarks of basal bone width and assessed the consistency of automated measurements compared to manual measurements.MethodsBased on the U-Net algorithm, a coarse-to-fine DL model was developed and trained using 80 cone-beam computed tomography (CBCT) images. The model’s prediction capabilities were validated on 10 CBCT scans and tested on an additional 34. To evaluate the performance of the DL model, its measurements were compared with those taken manually by one junior orthodontist using the concordance correlation coefficient (CCC).ResultsIt took approximately 1.5 s for the DL model to perform the measurement task in only CBCT images. This framework showed a mean radial error of 1.22 ± 1.93 mm and achieved successful detection rates of 71.34%, 81.37%, 86.77%, and 91.18% in the 2.0-, 2.5-, 3.0-, and 4.0-mm ranges, respectively. The CCCs (95% confidence interval) of the maxillary basal bone width and mandibular basal bone width distance between the DL model and manual measurement for the 34 cases were 0.96 (0.94–0.97) and 0.98 (0.97–0.99), respectively.ConclusionThe novel DL framework developed in this study improved the diagnostic accuracy of the individual assessment of maxillary width. These results emphasize the potential applicability of this framework as a computer-aided diagnostic tool in orthodontic practice.

Rotation error separation of a UTF300 high-speed and precision motorized spindle

To improve the measurement accuracy of high-speed and precision motorized spindle rotary error as the research objective, based on the three-point method, an error separation model and an objective optimization function for noise-containing signals are developed. To improve the convergence speed, globally optimize the model objectives, and obtain the best optimization region of the sensor mounting angle, it is proposed that an enhanced adaptive particle swarm optimization algorithm be used. With the improved particle swarm algorithm, the convergence speed was greater than that of the primary particle swarm algorithm by more than 50%. The spindle radial rotation error was experimentally measured and separated using a high-speed vertical machining center, and the deviation between the separation result and the experimental rotation error was 4.5%, indicating that the separation result's accuracy was high. It also proved the correctness and feasibility of the optimization algorithm.

Development and validation of a three-dimensional wind-turbine wake model based on high-order Gaussian function

Under natural conditions, the wake velocity distribution approximates top-hat shape in the near wake center, and a Gaussian shape in the far wake. To improve the prediction accuracy of the wake distribution, this paper proposes a three-dimensional high-order Gaussian linear entrainment wake model by modifying the wake velocity distribution with high-order Gaussian function. Field tests and wind tunnel tests are then conducted. The obtained results show that the proposed model can accurately describe the radial and vertical spatial distributions of the wake flow. Under different wind speeds, the maximum and minimum radial relative errors are respectively 7.29% and 1.4%, while the maximum and minimum vertical relative errors are respectively 7.64% and 3.87%. This demonstrates that the proposed model can accurately predict the velocity distribution in the whole wake region. The proposed model is simple, has low cost, and exhibits high accuracy. Thus, it can be used in wind farm layout optimization and wake control.

Darts fast-learning reduces theta power but is not affected by Hf-tRNS: A behavioral and electrophysiological investigation

Sports trainers have recently shown increasing interest in innovative methods, including transcranial electric stimulation, to enhance motor performance and boost the acquisition of new skills during training. However, studies on the effectiveness of these tools on fast visuomotor learning and brain activity are still limited. In this randomized single-blind, sham-controlled, between-subjects study, we investigated whether a single training session, either coupled or not with 2 mA online high-frequency transcranial random noise stimulation (hf-tRNS) over the bilateral primary motor cortex (M1), would affect dart-throwing performance (i.e., radial error, arm range of motion, and movement variability) in 37 healthy volunteers. In addition, potential neurophysiological correlates were monitored before and after the training through a 32-electrode portable electroencephalogram (EEG). Results revealed that a single training session improved radial error and arm range of motion during the dart-throwing task, but not movement variability. Furthermore, after the training, resting state-EEG data showed a decrease in theta power. Radial error, arm movement, and EEG were not further modulated by hf-tRNS. This indicates that a single training session, regardless of hf-tRNS administration, improves dart-throwing precision and movement accuracy. However, it does not improve movement variability, which might require multiple training sessions (expertise resulting in slow learning). Theta power decrease could describe a more efficient use of cognitive resources (i.e., attention and visuomotor skills) due to the fast dart-throwing learning. Further research could explore different sports by applying longer stimulation protocols and evaluating other EEG variables to enhance our understanding of the lasting impacts of multi-session hf-tRNS on the sensorimotor cortex within the framework of slow learning and training assistance.

Automatic localization of anatomical landmarks in head cine fluoroscopy images via deep learning.

Fluoroscopy guided interventions (FGIs) pose a risk of prolonged radiation exposure; personalized patient dosimetry is necessary to improve patient safety during these procedures. However, current FGIs systems do not capture the precise exposure regions of the patient, making it challenging to perform patient-procedure-specific dosimetry. Thus, there is a pressing need to develop approaches to extract and use this information to enable personalized radiation dosimetry for interventional procedures. To propose a deep learning (DL) approach for the automatic localization of 3D anatomical landmarks on randomly collimated and magnified 2D head fluoroscopy images. The model was developed with datasets comprising 800000 pseudo 2D synthetic images (mixture of vessel-enhanced and non-enhancement), each with 55 annotated anatomical landmarks (two are landmarks for eye lenses), generated from 135 retrospectively collected head computed tomography (CT) volumetric data. Before training, dynamic random cropping was performed to mimic the varied field-size collimation in FGI procedures. Gaussian-distributed additive noise was applied to each individual image to enhance the robustness of the DL model in handling image degradation that may occur during clinical image acquisition in a clinical environment. The model was trained with 629370 synthetic images for approximately 275000 iterations and evaluated against a synthetic image test set and a clinical fluoroscopy test set. The model shows good performance in estimating in- and out-of-image landmark positions and shows feasibility to instantiate the skull shape. The model successfully detected 96.4% and 92.5% 2D and 3D landmarks, respectively, within a 10mm error on synthetic test images. It demonstrated an average of 3.6±2.3mm mean radial error and successfully detected 96.8% 2D landmarks within 10mm error on clinical fluoroscopy images. Our deep-learning model successfully localizes anatomical landmarks and estimates the gross shape of skull structures from collimated 2D projection views. This method may help identify the exposure region required for patient-specific organ dosimetry in FGIs procedures.

Phantom-Enhanced High Mechanical Accuracy for Frame-Based Deep Brain Stimulation.

Deep brain stimulation (DBS) is a neurosurgical procedure that depends on high-accuracy targeting of structures to implant electrodes within the brain. The positioning of these electrodes in the brain determines the long-term efficacy of treating diseases such as Parkinson's disease, essential tremor, or dystonia.Misplaced electrodes in DBS can lead to poor efficacy and stimulation-induced side effects. Widespread targeting errors and variability have been published throughout the literature. As such,improvement in targeting accuracy is needed to enhance the quality of the procedures. A stereotactic phantom was utilized to test and adjust targeting before the surgical placement in the brain for 91 sequential electrodes.The tip of the microelectrode, the first rigid point in time, was measured and compared to the planned target.The technique utilized a to-target cannulawith an XY stage that allowed x-axis and y-axis adjustments and correction for inaccuracies relative to the phantom.A calculation was developed to convert anatomical angles (sagittal and coronal) provided by commercial planning stations to spherical angles to calculate points along a trajectory.Errorcalculations included each dimensional axis, Euclidean error, and radial error. Bends in the cannula and microelectrode were observed and corrected by referencing the phantom. All 91 first-pass tracks traversed the intended target, and three electrodes required a second mapping track beyond the first penetration due to neurophysiological and intraoperative testing. The results showed overall ultra-high (0-0.5 mm) to high (>0.5-1 mm) accuracy, an average Euclidean error of 0.66±0.30 mm, and a radial error of 0.45±0.28 mm with dimensional errors of less than 0.5 mm per axis. The utilization of a stereotactic phantom is an important tool to enhance the stereotactic targeting before insertion into the brain. This phantom technique yielded ultra-high to high accuracy in error calculations.Future methods and studies shouldfocus on error minimization to enhance these DBS mechanical accuracy and correlations with clinical outcomes.

Twist Angle Error Statistical Analysis and Uncertain Influence on Aerodynamic Performance of Three-Dimensional Compressor Rotor

Twist angle errors along the blade radial direction are uncertain and affected by cutting force, tool wear, and other factors. In this paper, the measured twist angle errors of 13 sections of 72 rotor blades were innovatively analyzed to obtain the rational statistical distribution. It is surprisingly found that the under-deflection systematic deviation of twist angle errors shows a gradually increasing W-shaped distribution along the radial direction, while the scatter is nearly linear. Logically, the statistical model is established based on the linear correlation of the scatter by regression analysis to reduce variable dimension from 13 to 1. The influence of the radial non-uniform twist angle errors’ uncertainty on the aerodynamic performance of the three-dimensional compressor rotor is efficiently quantified combining the non-intrusive polynomial chaos method. The results show that the mean values of mass flow rate, total pressure ratio, and isentropic efficiency at the typical operating conditions are lower than the nominal values due to the systematic deviation, indicating that the under-deflection twist angle errors lead to the decrease in compressor thrust. The compressor’s stable operating range is more sensitive to the scatter of twist angle errors, which is up to an order of magnitude greater than that of the total pressure ratio and isentropic efficiency, indicating the compressor’s safe and stable operation risk increases. Additionally, the flow field at the tip region is significantly affected by twist angle errors, especially at the shock wave position of the near-stall condition.

The Effect of Priority Assignment the Role of Evaluator and Executive in Collaborative Training on the Acquisition of Kinematic and Accuracy of Dart Throwing Skill

Background: Collaborative training can enhance individual learning, but the effect of role assignment on skill acquisition is unclear. Objectives: Therefore, this study aims to explore the influence of the evaluator and executor roles on the accuracy and kinematic indicators of dart throwing skill. Methods: Forty-eight female university students (mean age = 21.71 ± 1.14 years) participated in the study. Following dart throwing familiarization, participants were randomly assigned to individual group (IG), first-performer (FPG), and first-evaluator groups (FEG) (as a dyad condition) with alternating role orders. All groups completed a pre-test of 10 dart throws while kinematics was recorded via video camera. Practice consisted of 20 blocks of 6 trials (n = 120 physical trials for IG; n = 60 evaluation and n = 60 physical trials for FPG and FEG). Post-tests 1 and 2 directly followed after role reversals for dyads after each 60 trials in the acquisition phase. Accuracy was assessed via mean radial and variable errors. Kinematics (elbow release angle, average angular velocity, elbow range of motion) were analyzed using Kinovea software. Results: A two-way ANOVA revealed the FPG exhibited a significantly smaller release angle compared to IG. However, no significantly differences observed between FEG and IG for kinematic variables. Accuracy improved significantly from pre-to post-test at the three groups. Individual group led to higher accuracy than FPG, and frist-performer group had higher RE than the first-evaluator. Conclusions: The findings of this study support the benefits of observational learning in the early stages of learning. Also, priority assignment the role of evaluator and executive can influences motor performance in a dyadic approach.

Automatic soft-tissue analysis on orthodontic frontal and lateral facial photographs based on deep learning.

To establish the automatic soft-tissue analysis model based on deep learning that performs landmark detection and measurement calculations on orthodontic facial photographs to achieve a more comprehensive quantitative evaluation of soft tissues. A total of 578 frontal photographs and 450 lateral photographs of orthodontic patients were collected to construct datasets. All images were manually annotated by two orthodontists with 43 frontal-image landmarks and 17 lateral-image landmarks. Automatic landmark detection models were established, which consisted of a high-resolution network, a feature fusion module based on depthwise separable convolution, and a prediction model based on pixel shuffle. Ten measurements for frontal images and eight measurements for lateral images were defined. Test sets were used to evaluate the model performance, respectively. The mean radial error of landmarks and measurement error were calculated and statistically analysed to evaluate their reliability. The mean radial error was 14.44 ± 17.20 pixels for the landmarks in the frontal images and 13.48 ± 17.12 pixels for the landmarks in the lateral images. There was no statistically significant difference between the model prediction and manual annotation measurements except for the mid facial-lower facial height index. A total of 14 measurements had a high consistency. Based on deep learning, we established automatic soft-tissue analysis models for orthodontic facial photographs that can automatically detect 43 frontal-image landmarks and 17 lateral-image landmarks while performing comprehensive soft-tissue measurements. The models can assist orthodontists in efficient and accurate quantitative soft-tissue evaluation for clinical application.

Research on Time-Reversal Focusing Imaging Method to Evaluate a Multi-Layer Armor Composite Structure

Armor composite structures have attracted interest in structural health monitoring (SHM) for their applications in damage localization. The signal propagation and the frequency dispersion features of the Lamb wave signal on thick armor composite structures are more complicated than their counterparts on other composite plates. In this research, a time-reversal localization and imaging method for impact localization of armor composite structures is proposed. First, composite sandwich structures were designed that are typically composed of ballistic-resistant ceramic materials as the face panel and a composite material as the core layer, sandwiched between metal materials serving as the backplate. The results show that the proposed method can validate the position of impact efficiently, and radial error is within 4.12 mm and 5.39 mm in single-damage and multi-damage imaging localization, respectively.

Long-term analysis of Sentinel-6A orbit determination: Insights from three years of flight data

The Sentinel-6A mission extends the set of satellites dedicated to continuous ocean altimetry measurements, which started with TOPEX/Poseidon in 1992. To utilize these measurements, high-accuracy orbit solutions with radial position errors of less than 1.5cm (RMS) are required. For precise orbit determination (POD), a dedicated GPS/Galileo triple-frequency receiver (PODRIX) is available. Complementary to this, the TriG receiver provides GPS-only observations for POD and radio occultations. Previous research has shown that reduced-dynamic orbit solutions with PODRIX GPS/Galileo measurements meet the mission requirement for radial position accuracy of 1.5 cm (RMS). However, baseline estimation between the PODRIX and TriG antennas still reveals obvious inconsistencies in the along- and cross-track directions. In this study, we present a comprehensive reprocessing of orbit solutions using three years of Sentinel-6A flight data from both receivers covering January 2021 to December 2023. Cross-comparison between both receivers shows that inconsistencies can be removed by applying a yaw bias and a timing error correction. A comparative analysis of macro-models demonstrates strong correlation of cross-track empirical accelerations with the Sun elevation above the orbital plane when using nominal surface properties reported by the manufacturer. Adjusted macro-models improve performance at the expense of requiring increased surface areas or reflectivity coefficients. In addition to improved radiation pressure models, newly estimated antenna patterns and a time-varying gravity field were used in the analysis. The reprocessed data confirm good agreement of the results of both receivers except for a relative timing error of 1.3 μs, and the yaw bias correction of −0.43° removes previously-observed systematic differences in the estimated empirical accelerations. Satellite laser ranging results based on SLRF2020 station coordinates demonstrate a 5—6 mm (1D RMS) accuracy of the GNSS-based POD solutions across the IGb14/IGS20 reference frame transition in November 2022.

Can artificial intelligence-driven cephalometric analysis replace manual tracing? A systematic review and meta-analysis.

This systematic review and meta-analysis aimed to investigate the accuracy and efficiency of artificial intelligence (AI)-driven automated landmark detection for cephalometric analysis on two-dimensional (2D) lateral cephalograms and three-dimensional (3D) cone-beam computed tomographic (CBCT) images. An electronic search was conducted in the following databases: PubMed, Web of Science, Embase, and grey literature with search timeline extending up to January 2024. Studies that employed AI for 2D or 3D cephalometric landmark detection were included. The selection of studies, data extraction, and quality assessment of the included studies were performed independently by two reviewers. The risk of bias was assessed using the Quality Assessment of Diagnostic Accuracy Studies-2 tool. A meta-analysis was conducted to evaluate the accuracy of the 2D landmarks identification based on both mean radial error and standard error. Following the removal of duplicates, title and abstract screening, and full-text reading, 34 publications were selected. Amongst these, 27 studies evaluated the accuracy of AI-driven automated landmarking on 2D lateral cephalograms, while 7 studies involved 3D-CBCT images. A meta-analysis, based on the success detection rate of landmark placement on 2D images, revealed that the error was below the clinically acceptable threshold of 2mm (1.39mm; 95% confidence interval: 0.85-1.92mm). For 3D images, meta-analysis could not be conducted due to significant heterogeneity amongst the study designs. However, qualitative synthesis indicated that the mean error of landmark detection on 3D images ranged from 1.0 to 5.8mm. Both automated 2D and 3D landmarking proved to be time-efficient, taking less than 1min. Most studies exhibited a high risk of bias in data selection (n = 27) and reference standard (n = 29). The performance of AI-driven cephalometric landmark detection on both 2D cephalograms and 3D-CBCT images showed potential in terms of accuracy and time efficiency. However, the generalizability and robustness of these AI systems could benefit from further improvement. PROSPERO: CRD42022328800.

Microelectrode Recording During Deep Brain Stimulation Does Not Consistently Represent Lead Trajectory.

Long-term outcomes in deep brain stimulation (DBS) depend on accuracy of lead placement. Microelectrode recording (MER) is a long-used adjunct to leverage neurophysiological information to confirm satisfactory trajectory of implanted electrodes. The goal of this study was to evaluate the consistency in which electrodes are placed in sampled microelectrode trajectories. This is a retrospective study using intraoperative computed tomography to measure final electrode deviation from MER probe placement during the DBS insertion targeting subthalamic nucleus. Fifteen patients had 29 DBS leads placed using MER assistance. Radial distance between the probe and the lead were measured for each patient using intraoperative imaging. In addition, the preoperative target to final lead error was measured in 14 patients undergoing subthalamic nucleus implants without the use of MER and compared with the 15 patients in which MER was used as an adjunct. There was no significant difference in the mean radial target error (1.2 vs 1.0 mm, P = .156) when comparing the leads placed with or without MER assistance, respectively. The mean difference in final position of microelectrode compared with DBS lead was 0.9 ± 0.1 (range 0.4-2.0 mm). Of all MER-assisted electrodes placed, 44.8% (13) of electrode final positions radially deviated 1.0 mm or more from the MER probe. Electrode placement may deviate significantly from MER trajectories. Given the concern that physiological data may not be representative of the final electrode trajectory, surgeons should consider using intraoperative imaging or other adjunctive techniques during DBS to confirm accuracy and satisfactory trajectory of DBS leads.

The influence of journal's roundness on rotating accuracy of aerostatic spindle and precision improvement by time-controlled grinding

The roundness of the journal directly affects the radial synchronous error (RSE) of the aerostatic spindle. However, the state-of-the-art ultra-precision grinding machine can only achieve a roundness of 0.2–0.3 µm, which limits the improvement of the rotating accuracy. Hence, novel time-controlled grinding (TCG) is proposed to further converge the rotor's roundness. To provide the machining basis and prediction, CFD is used to quantitatively analyze the effect of roundness on the flow field and a rotating trajectory model is established. Based on the model's guidance, the rotor's average roundness converged to 0.08 µm by TCG, even better than manual lapping, and the RSE decreased from ∼30 nm to ∼11 nm synchronously, proving the model's and TCG's feasibility and potential.

Sentinel-6 Michael Freilich precise orbit determination using PODRIX and TriG receiver measurements

The Sentinel-6 Michael Freilich altimetry mission flies two GNSS receivers: a primary multi-GNSS (GPS plus Galileo) PODRIX receiver and a GPS-only TriG receiver. Each of these receivers is independently capable of supporting the precise orbit determination (POD) requirement for < 1.5 cm radial rms error. In this study, we characterize the performance of single-receiver solutions and evaluate the benefits of a combined TriG and PODRIX orbit solution. The availability of both sets of receiver observations revealed a 10 mm in-track difference between orbit solutions derived independently from TriG and PODRIX tracking data. Based on satellite laser ranging (SLR) residuals, this bias has been isolated to an apparent inconsistency between the estimated TriG receiver clock and observation time-tags of approximately 1.3 μs\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\\mu \\hbox {s}$$\\end{document}, which is equivalent to a common range error of roughly 400 m in the TriG observations. After applying this calibration, the TriG and PODRIX displayed similar performance in terms of orbit overlap precision. PODRIX-Galileo observations showed lower code and phase tracking residual rms values compared to the GPS observations. Overall, processing the calibrated TriG and PODRIX observations separately results in highly accurate orbit solutions with radial orbit accuracies better than 1 cm rms as indicated by one-way SLR residual rms of 7.2 mm or better for each solution. Orbit solution accuracy is slightly improved by processing both TriG and PODRIX observations together, resulting in one-way SLR residual rms of 7.0 mm.

Automatic measurement of lower limb alignment in portable devices based on deep learning for knee osteoarthritis

BackgroundFor knee osteoarthritis patients, analyzing alignment of lower limbs is essential for therapy, which is currently measured from standing long-leg radiographs of anteroposterior X-ray (LLR) manually. To address the time wasting, poor reproducibility and inconvenience of use caused by existing methods, we present an automated measurement model in portable devices for assessing knee alignment from LLRs.MethodWe created a model and trained it with 837 conforming LLRs, and tested it using 204 LLRs without duplicates in a portable device. Both manual and model measurements were conducted independently, then we recorded knee alignment parameters such as Hip knee ankle angle (HKA), Joint line convergence angle (JCLA), Anatomical mechanical angle (AMA), mechanical Lateral distal femoral angle (mLDFA), mechanical Medial proximal tibial angle (mMPTA), and the time required. We evaluated the model’s performance compared with manual results in various metrics.ResultIn both the validation and test sets, the average mean radial errors were 2.778 and 2.447 (P<0.05). The test results for native knee joints showed that 92.22%, 79.38%, 87.94%, 79.82%, and 80.16% of the joints reached angle deviation<1° for HKA, JCLA, AMA, mLDFA, and mMPTA. Additionally, for joints with prostheses, 90.14%, 93.66%, 86.62%, 83.80%, and 85.92% of the joints reached that. The Chi-square test did not reveal any significant differences between the manual and model measurements in subgroups (P>0.05). Furthermore, the Bland-Altman 95% limits of agreement were less than ± 2° for HKA, JCLA, AMA, and mLDFA, and slightly more than ± 2 degrees for mMPTA.ConclusionThe automatic measurement tool can assess the alignment of lower limbs in portable devices for knee osteoarthritis patients. The results are reliable, reproducible, and time-saving.