Polynomial Curve Fitting Method Research Articles

Impact of Data Corruption and Operating Temperature on Performance of Model-Based SoC Estimation

Electric vehicles (EVs) are becoming popular around the world. Making a lithium battery (LIB) pack with a robust battery management system (BMS) for an EV to operate under different complex environments is both a challenge and a requirement for engineers. A BMS can intelligently manage LIB systems by estimating the battery state of charge (SoC). Due to the nonlinear characteristics of LIB, influenced by factors such as the harsh environment and data corruption caused by electromagnetic interference (EMI) inside electric vehicles, SoC estimation should consider available capacity, model parameters, operating temperature and reductions in data sampling time. The widely used model-based algorithms, such as the extended Kalman filter (EKF) have limitations. Therefore, a detailed review of the balance between temperature, data sampling time, and different model-based algorithms is necessary. Firstly, a state of charge—open-circuit voltage (SoC-OCV) curve of LIB is obtained by the polynomial curve fitting (PCF) method. Secondly, a first-order RC (1-RC) equivalent circuit model (ECM) is applied to identify the battery parameters using a forgetting factor-based recursive least squares algorithm (FF-RLS), ensuring accurate internal battery parameters for the next step of SoC estimation. Thirdly, different model-based algorithms are utilized to estimate the SoC of LIB under various operating temperatures and data sampling times. Finally, the experimental data by dynamic stress test (DST) is collected at temperatures of 10 °C, 25 °C, and 40 °C, respectively, to verify and analyze the impact of operating temperature and data sampling time to provide a practical reference for the SoC estimation.

Synthetic minority oversampling and iterative fluorescence-suppression integrated algorithm for Raman spectrum pesticide detection system

Raman spectrum preprocessing method for automatic denoising and suppression of the fluorescent background. In this method, noise is reduced using wavelet transform, and a modified polynomial curve fitting method is implemented such that an algorithm can independently identify the optimal curve parameters for fluorescent background suppression. To address the problem of imbalanced datasets, the present study employed a synthetic minority oversampling technique to increase the volume of data in minority classes. This technique enables the prediction of pesticides that are otherwise difficult to detect, and the prediction accuracy is comparable to that of detection with large data volumes. The proposed convolutional neural network model was verified to accurately identify the type of single pesticides and composition of mixed pesticides. The prediction accuracy for mixed pesticides reached 99.1%.

Computer-aided detection of retinopathy of prematurity severity assessment via vessel tortuosity measurement in preterm infants' fundus images.

To develop a computer-aided diagnostic system for retinopathy of prematurity (ROP) disease using retinal vessel morphological features. A total of 200 fundus images from 136 preterm infants with stage 1 to 3 ROP were analysed. Two methods were developed to measure vessel tortuosity: the peak-and-valley method and the polynomial curve fitting method. Correlations between temporal artery tortuosity (TAT) and temporal vein tortuosity (TVT) with ROP severity were investigated, and vessel tortuosity relationships with vessel angles (TAA and TVA) and vessel widths (TAW and TVW). A separate dataset from Japan containing 126 images from 97 preterm patients was used for verification. Both methods identified similar tortuosity in images without ROP and mild ROP cases. However, the polynomial curve fit method demonstrated enhanced tortuosity detection in stages 2 and 3 ROP compared to the peak and valley method. A strong positive correlation was revealed between ROP severity and increased arterial and venous tortuosity (P < 0.0001). A significant negative correlation between TAA and TAT (r = -0.485, P < 0.0001) and TVA and TVT (r = -0.281, P < 0.0001), and a significant positive correlation between TAW and TAT (r = 0.204, P value = 0.0040) were identified. Similar results were found in the test dataset from Japan. ROP severity was associated with increased retinal tortuosity and retinal vessel width while displaying a decrease in retinal vascular angle. This quantitative analysis of retinal vessels provides crucial insights for advancing ROP diagnosis and understanding its progression.

Optimization of laser-texturing process parameters of Ti6Al4V alloys using metaheuristic algorithms

Manufacturing microscale textures can reduce friction and wear compared with smooth surfaces. In laser surface texturing, the resulting texture geometry is directly affected by laser parameters such as laser power, scanning speed and overlap rate. Therefore, finding the optimal parameter settings to improve the process performance is crucial. One of the biggest challenges encountered during the optimization stage is accurately determining the functional relationship between surface quality and process parameters. Another significant problem is solving the nonlinear equation system. This study aims to investigate the effect of laser-textured surfaces on the tribological behavior of titanium Ti6Al4V. Optimization of laser texture process parameters using metaheuristic algorithms is a hot topic. The first contribution of this study is the modeling of the contact angle, kurtosis and skewness quality parameters of Ti6Al4V titanium alloy based on the polynomial curve fitting method as a function of laser power, laser speed and overlap rate. Another significant contribution is to optimize the proposed nonlinear regression models using metaheuristic approaches such as genetic algorithm, particle swarm optimization, whale optimization algorithm, gray wolf optimization, equilibrium optimizer, teaching-learning-based optimization, student psychology-based optimization (SPBO), slime mould algorithm and social network search. The reason for using many metaheuristics is to test the effectiveness of different metaheuristic methods in solving this problem. Analysis of variance was performed to evaluate experimental data and compare the success of optimization methods. The results revealed that the proposed metaheuristic methods can be applied to the related problem. It has been observed that the methods give identical results. Among these methods, the SPBO method provided the best performance in the optimization of all laser output parameters.

Winter wheat irrigation parameters in the Brazilian Cerrado

Wheat is the second most produced cereal in the world. The gap between production and consumption in Brazil shows that wheat production must be increased to meet the domestic demand. In this context, the objective of this study was to determine the crop coefficient of wheat and the ideal moment for irrigation, based on the soil water depletion rate (f). The study was carried out in the tropics, on an Oxisoil (Typic Haplustox), in an experimental field of Embrapa Cerrados in Planaltina, Federal District of Brazil. The experimental design consisted of randomized blocks with three replications. The treatments were differentiated by the soil water depletion index (f) of 0.2, 0.4, 0.6, and 0.8. In other words, irrigation was initiated when 20, 40, 60 and 80% of the available soil water capacity (AWC) at a soil depth of 0.4 m was consumed by the crop. Wheat yield and yield components were evaluated. The phenological development was monitored based on Zadok’s scale, in days after emergence, and the management parameters crop coefficient (Kc) and irrigation timing were determined. The results indicated a soil water depletion index of 0.4 as the best indicator for beginning wheat irrigation. Under this condition, Kc can be estimated daily as a function of days after emergence (DAE) by the quadratic polynomial regression or curve fitting method, adjusted as follows: Kc = −0.000298 × DAE² + 0.036695 × DAE + 0.332486. Alternatively, the means of Kc of each phenological crop phase can be used: Kc = 0.45 (establishment); 0.81 (tillering); 1.21 (booting); 1.43 (flowering); 1.27 (grain formation); and 0.89 (maturation). Thus, there is a limit of consumption of soil available water, at which wheat still achieves the full yield potential, with a rational use of water which varies according to the development phase.

Improvement of DBSCAN Algorithm Based on K-Dist Graph for Adaptive Determining Parameters

For the shortcomings of an unstable clustering effect and low accuracy caused by the manual setting of the two parameters Eps and MinPts of the DBSCAN (density-based spatial clustering of applications with noise) algorithm, this paper proposes an adaptive determination method for DBSCAN algorithm parameters based on the K-dist graph, noted as X-DBSCAN. The algorithm uses the least squares polynomial curve fitting method to fit the curve in the K-dist graph to generate a list of candidate Eps parameters and uses the mathematical expectation method and noise reduction threshold to generate the corresponding MinPts parameter list. According to the clustering results of each group of parameters in the Eps and MinPts parameter lists, a stable range of cluster number changes is found, and the MinPts and Eps corresponding to the maximum K value in the stable range are selected as the optimal algorithm parameters. The optimality of this parameter was verified using silhouette coefficients. A variety of experiments were designed from multiple angles on the artificial dataset and the UCI real dataset. The experimental results show that the clustering accuracy of X-DBSCAN was 21.83% and 15.52% higher than that of DBSCAN on the artificial and real datasets, respectively. The X-DBSCAN algorithm was also superior to other algorithms through comprehensive evaluation and analysis of various clustering indicators. In addition, experiments on four synthetic Gaussian datasets of different dimensions showed that the average clustering indices of the proposed algorithm were above 0.999. The X-DBSCAN algorithm can select parameters adaptively in combination with the characteristics of the dataset; the clustering effect is better, and clustering process automation is realized.

Key Technologies of Collaborative Target Detection between Self-propelled Anti-aircraft Weapons

In the case of a lack of target information, a self-propelled anti-aircraft weapon can complete its own combat functions by using the detection information of other vehicles through detection coordination technology. When using the detection coordination technology to process the detection information of other vehicles into information that can be used by the vehicle, the relative position and attitude of the vehicle and data transmission delay will affect the accuracy of the processed target information. With the system of detection collaborative technology expounded, emphasis is attached to the research on the detection coordination method based on pure angles without considering distance. A mathematical model of detection coordination technology is created, the approximate statistical law of the processed detection coordination information is obtained through the Monte Carlo method, and the influence of various factors on transmission error analyzed. In addition, the polynomial curve fitting method is used for time alignment, and the unscented Kalman filter method based on singular value decomposition adopted to reduce the information transfer error. The simulation results show the the method is effective for that end.

Robust fault-tolerant control design for polynomial fuzzy systems

This paper focuses on the robust H∞ fault-tolerant control issue for a family of polynomial fuzzy systems subject to actuator faults and external disturbances. Especially, asynchronous fuzzy membership function technique is considered to minimize the expenses of controller implementation. To derive the robust stability conditions, actuator fault factors are taken into the account in the closed-loop system formulation. Moreover, to utilize the exact shape information of fuzzy membership functions in the stability criterion, they are approximated as sum-of-square polynomials with the aid of polynomial curve fitting method. In addition, to reduce the conservatism, a generalized polynomial Lyapunov function is considered. However, there are some non-convex terms in the stability conditions due to the construction of the generalized Lyapunov function that consists of all state-dependent polynomial matrix. Thus, some existing convex methods for general polynomial fuzzy systems are rendered inapplicable. To deal with this issue, the actual non-convex stability conditions converts into a convex optimization problem by introducing some additional convex constraints. Moreover, the membership-function-dependent robust stability conditions of the closed-loop system are provided as sum-of-square convex polynomial constraints. Lastly, the usefulness and superiority of the designed controller are illustrated by the simulation examples.

Measurement precision enhancement of surface plasmon resonance based angular scanning detection using deep learning

Angular scanning-based surface plasmon resonance measurement has been utilized in label-free sensing applications. However, the measurement accuracy and precision of the surface plasmon resonance measurements rely on an accurate measurement of the plasmonic angle. Several methods have been proposed and reported in the literature to measure the plasmonic angle, including polynomial curve fitting, image processing, and image averaging. For intensity detection, the precision limit of the SPR is around 10–5 RIU to 10–6 RIU. Here, we propose a deep learning-based method to locate the plasmonic angle to enhance plasmonic angle detection without needing sophisticated post-processing, optical instrumentation, and polynomial curve fitting methods. The proposed deep learning has been developed based on a simple convolutional neural network architecture and trained using simulated reflectance spectra with shot noise and speckle noise added to generalize the training dataset. The proposed network has been validated in an experimental setup measuring air and nitrogen gas refractive indices at different concentrations. The measurement precision recovered from the experimental reflectance images is 4.23 × 10–6 RIU for the proposed artificial intelligence-based method compared to 7.03 × 10–6 RIU for the cubic polynomial curve fitting and 5.59 × 10–6 RIU for 2-dimensional contour fitting using Horner's method.

Colorimetric Parameters Correlated with the Variation in the Marker Constituent Contents During the Stir-fry Processing of Schizonepetae Spica

Schizonepetae Spica (SS), the dried spike of Schizonepeta tenuifolia Briq., is often consumed in beverages, sauce, beneficial tea, or herbal medicine. Raw SS has the action of dispelling the common cold and fever. Schizonepetae Spica Carbonisata (SSC; raw SS processed by stir-frying until the surface becomes blackish brown and the interior turns dark yellow) has particularly strong efficacy in arresting bleeding from bloody stools and metrorrhagia. In this paper, a high-performance liquid chromatography diode array detector method and colorimetry were employed to determine 6 nonvolatile marker constituents, and the color parameters (L*, a*, and b*) of SS during stir-fry processing, respectively. According to Pearson correlation analysis, L*, a*, and b* showed significant correlations with the levels of luteolin-7-O-β-D-glucoside, apigenin-7-O-β-D-glucoside, hesperidin, and rosmarinic acid, while a* was significantly correlated with the amounts of luteolin and apigenin. The corresponding regression models were good fits for the color parameters as respective functions of the 6 marker constituents according to enter multiple linear regression and third-order polynomial curve fitting methods (adjusted R2 > 0.800). Surrogate prediction of marker compounds using color parameters provides a valuable and cost-saving tool for rapid or online monitoring of the processing degree of SS. Raw SS and SSC could be determined based on the color parameters and 6 marker constituent contents for a comprehensive quality evaluation for the first time. This paper provides scientific data to validate the relationship between color features and quality during the processing of SS.

In-Process Error-Matching Measurement and Compensation Method for Complex Mating.

This study proposed an error-matching measurement and compensation method for curve mating and complex mating. With use of polynomial curve fitting and least squares methods for error analysis, an algorithm for error identification and error compensation were proposed. Furthermore, based on the proposed method, an online error-matching compensation system with an autorevising function module for autogenerating an error-compensated NC program for machining was built. Experimental verification results showed that the proposed method can effectively improve the accuracy of assembly matching. In a curve-type mating experiment, the matching error without compensation was 0.116 mm, and it decreased to 0.048 mm after compensation. The assembly accuracy was improved by 28%. In a complex-type mating experiment, the verification results showed that the error reductions after compensation for three mating shapes (straight line, triangle, and curve shape) were 81%, 87%, and 79%, respectively. It showed that the proposed method can improve the assembly accuracy for complex mating shapes, which would also be improved without losing production efficiency.

A novel model to determine gas content in naturally fractured shale

A shale reservoir’s gas content is an indispensable, crucial parameter used in exploration, formation evaluation, block selection and exploitation. Canister gas desorption measurements are used to quantify lost gas volumes that are key to determine the gas content of the reservoir Usually, a lost gas volume is evaluated using the United States Bureau of Mines (USBM) direct method and/or a polynomial curve-fitting method. These methods extrapolate desorption data to time zero in the curve of a cumulative desorption volume versus the square root of time. Lost gas time is the time spent lifting the core out of a hole, handling and placing it inside a canister for evaluation., For cores with densely developed natural fractures, neither the USBM method nor the polynomial curve-fitting method can accurately estimate the lost gas volume if the lifting, handling and canister storage takes a long time. The lost gas period is a non-isothermal gas desorption process; the surrounding temperature keeps declining. Experiments were conducted to determine the methane desorption volumes of a Duvernay shale sample under various pressures and temperatures. Using Boyle’s law and the multilayer adsorption model, a novel mathematical model was built to calculate lost gas volume, including free gas and desorption gas. Because gas desorption is an endothermic procedure, equations for the inhaled energy were developed to overcome the van der Waals force between the molecules, deorbit the vibration equilibrium and the migration of molecules into the air. Finally, a shale sample with dense natural fractures from the Duvernay formation of Alberta, Canada, was used to validate the mathematical model. For comparison, the results from the USBM method, polynomial curve-fitting method and model’s calculations are illustrated in tables and figures. The outcomes exposed the results from the USBM method and polynomial curve-fitting method as deviating from the real gas content in the Duvernay formation without the consideration of the free gas loss. The model is reliable for performing gas content evaluations because it considers dual-pore structure, PVT behaviour and free gas escape simultaneously.

An agent‐based fast concurrent prediction scheme for transient and short‐term voltage instability

A novel agent-based online prediction method is presented in this paper, to predict the status of both transient and short-term voltage (STV) stability, against fault occurrence. In the proposed method, the trajectories of the relative frequency deviation (ΔF) and the power imbalance (ΔP) are estimated by employing the third-degree polynomial curve fitting method. By tracking the estimated trajectories on ΔF–ΔP plane and checking some simple defined rules, an early prediction of both transient and STV instability is achieved in an organized multi-agent system (MAS). In order to evaluate the performance of the proposed algorithm, the method has been tested on IEEE 39-bus system, IEEE 118-bus system and IEEE Nordic test system. Based on the obtained results, the proposed algorithm has an overall accuracy of 99.5% under symmetrical and asymmetrical faults, PMU measurement error, different operating points and topological changes.

A modified dental age assessment method for 5- to 16-year-old eastern Chinese children

ObjectivesAge estimation is widely applied in the field of orthodontics, pediatric dentistry, and forensic science. Dental age estimation by the radiological method is frequently used because of its convenience and noninvasiveness. However, there are not enough suitable methods for eastern Chinese children. This study aimed to establish a modified formula for eastern Chinese children according to the Demirjian method and then compared the accuracy of the modified method with the Demirjian method and Willems method.Materials and methodsA total of 2367 dental panoramic radiographs from individuals aged 5–16 years of eastern China were collected as samples. Age estimation was conducted using the Demirjian and Willems methods. The polynomial curve fitting method was used to modify the Demirjian method to improve its application to the eastern Chinese children. The paired t test and accuracy ratio were used to compare the applicability of the modified methods with two commonly used methods.ResultsThe mean chronological age (CA) of the subjects was 11.20 ± 3.29 years for boys and 10.99 ± 3.12 years for girls. The mean difference values between the CA and dental age (DA) (CA–DA) using the Demirjian and Willems methods were 0.73 and 0.7 for boys, respectively, and both 0.79 for girls. The modified method using the polynomial curve fitting presented a smaller underestimation compared with CA for both boys (0.04 years) and girls (0.09 years), which showed a high suitability to Chinese children to some extent.ConclusionsThe Willems method was more accurate in estimating DA compared with the Demirjian method. However, the modified method was more accurate than the two methods; therefore, it can be used in eastern Chinese children.Clinical relevanceIt was thought to be a non-invasive, convenient, and efficient method to connect DA and CA. By estimating dental age, pediatrist, and orthodontists can better understand the development of permanent teeth and provide a more accurate orthodontic treatment time and treatment plan to children patients.

Multi-Sensor Data Fusion for Remaining Useful Life Prediction of Machining Tools by IABC-BPNN in Dry Milling Operations.

Inefficient remaining useful life (RUL) estimation may cause unpredictable failures and unscheduled maintenance of machining tools. Multi-sensor data fusion will improve the RUL prediction reliability by fusing more sensor information related to the machining process of tools. In this paper, a multi-sensor data fusion system for online RUL prediction of machining tools is proposed. The system integrates multi-sensor signal collection, signal preprocess by a complementary ensemble empirical mode decomposition, feature extraction in time domain, frequency domain and time-frequency domain by such methods as statistical analysis, power spectrum density analysis and Hilbert-Huang transform, feature selection by a Light Gradient Boosting Machine method, feature fusion by a tool wear prediction model based on back propagation neural network optimized by improved artificial bee colony (IABC-BPNN) algorithm, and the online RUL prediction model by a polynomial curve fitting method. An example is used to verify whether if the prediction performance of the proposed system is stable and reliable, and the results show that it is superior to its rivals.

Effect of Self-Heating on Terahertz Double Avalanche Region Transit Time Source

This paper highlights the effect of temperature on the high-frequency performance of the double avalanche region avalanche transit time (ATT) source capable of operating in multiple numbers of frequency bands. The variance of static (DC) and large-signal properties on temperature are accounted by utilizing a well-established authors’ developed simulation method. The large-signal admittance characteristics of the source show the presence of discrete and far separated negative conductance bands which can yield a wide tuning capability of it. Here in this paper, linear changes of DC parameters as well as non-linear changes of large-signal parameters with junction temperature have been studied; with the help of utilizing linear and quadratic polynomial curve fitting method, analytical expressions are obtained which represent the DC and high-frequency parameters of the source as functions of junction temperature. These will be very much beneficial for optimization of thermal design-related issues of the ATT sources.

Unmanned Aerial Vehicle Positioning Algorithm Based on the Secant Slope Characteristics of Transmission Lines

Aiming at the problem of positioning consistency and real-time performance during power inspection of unmanned aerial vehicles (UAV), this paper proposes an UAV positioning algorithm based on the secant slope characteristics of transmission lines. The traditional correlation scan matching (CSM) positioning method is prone to mismatch under the condition of a single geometric feature. Although the inertial measurement unit (IMU) -aided CSM method improves the positioning accuracy, using the IMU for a long time to estimate the position is prone to cumulative errors and increases the computational cost to a certain extent. Our proposed method uses the least square polynomial curve fitting method to parametric reconstruction of the transmission line, which enhances the geometric characteristics of the transmission line and made up the missing part. The secant line of transmission line is extracted and its feature is parameterized to match with the reconstructed transmission line map, and the translation vector and rotation vector of UAV are estimated. Through multiple sets of experiments, it is proved that our algorithm is less than the CSM method and the IMU-aided CSM method in terms of average translation error, average rotation error, and positioning time. Our algorithm not only improves the positioning accuracy but also guarantees the real-time positioning requirements.

A Comparative Study of Charging Voltage Curve Analysis and State of Health Estimation of Lithium-ion Batteries in Electric Vehicle

Lithium-ion (Li-ion) cells degrade after repeated cycling and the cell capacity fades while its resistance increases. Degradation of Li-ion cells is caused by a variety of physical and chemical mechanisms and it is strongly influenced by factors including the electrode materials used, the working conditions and the battery temperature. At present, charging voltage curve analysis methods are widely used in studies of battery characteristics and the constant current charging voltage curves can be used to analyze battery aging mechanisms and estimate a battery’s state of health (SOH) via methods such as incremental capacity (IC) analysis. In this paper, a method to fit and analyze the charging voltage curve based on a neural network is proposed and is compared to the existing point counting method and the polynomial curve fitting method. The neuron parameters of the trained neural network model are used to analyze the battery capacity relative to the phase change reactions that occur inside the batteries. This method is suitable for different types of batteries and could be used in battery management systems for online battery modeling, analysis and diagnosis.

Performance of multiple probability distributions in generating daily precipitation for the simulation of hydrological extremes

Stochastic weather generators are statistical models widely used to produce climate time series with similar statistical properties to observed data. They are also used as downscaling tools to generate climate change scenarios for impact studies. Precipitation is one of the main variables simulated by weather generators and is also a key variable for impact studies, especially for hydrology. Precipitation is usually simulated by multiple precipitation models which have been proposed for simulating site-specific precipitation. However, these models’ performance in simulating watershed-averaged extreme precipitation, especially in representing hydrological extremes, has not been well-investigated. Accordingly, this study compares the performance of six probability distributions (exponential, gamma, skewed normal, mixed exponential, hybrid exponential/Pareto, and Weibull distributions) and a polynomial curve-fitting method in generating precipitation for simulating hydrological extremes over three basins using a set of extreme indices. The results show that except for the exponential distribution (EXP), all of the methods produce the distribution of observed precipitation at the daily, monthly and annual scales reasonably well for all three river basins. Although the three-parameter hybrid exponential/Pareto distribution (EXPP) overestimates precipitation extremes, other three-parameter models produce extremes accurately. The three-parameter mixed exponential (MEXP) distribution outperforms other models for simulating precipitation extremes. However, with respect to representing hydrological extremes, the MEXP distribution is not the best model. When simulating extreme streamflows with synthetic weather data, the EXP distribution shows the worst performance, while the curve fitting method (PN) performs the best. The inferiority of the EXPP distribution in generating extreme precipitation does not propagate to extreme flow simulations. Meanwhile, the performance of WEB is moderate in terms of representing hydrological extremes. Overall, finding the model that best reproduces precipitation for simulating hydrological extremes is not as clear-cut, since the performance of each model is extreme-indices dependent. Taking all of the indices into account, the MEXP and the PN appear to be superior in representing extreme precipitation and hydrological extremes.

Deep-learning–assisted automatic digitization of applicators in 3D CT image-based high-dose-rate brachytherapy of gynecological cancer

PurposeApplicator digitization is one of the most critical steps in 3D high-dose-rate brachytherapy (HDRBT) treatment planning. Motivated by recent advances in deep-learning, we propose a deep-learning–assisted applicator digitization method for 3D CT image-based HDRBT. This study demonstrates its feasibility and potential in gynecological cancer HDRBT. Methods and MaterialsOur method consisted of two steps. The first step used a U-net to segment applicator regions. We trained the U-net using two-dimensional CT images with a tandem-and-ovoid (T&O) applicator and corresponding applicator mask images. The second step applied a spectral clustering method and a polynomial curve fitting method to extract applicator central paths. We evaluated the accuracy, efficiency, and robustness of our method in different scenarios including other T&O cases that were not used in training, a T&O case scanned with cone-beam CT, and Y-tandem and cylinder-applicator cases. ResultsIn test cases with a T&O applicator, average 3D Dice similarity coefficient between automatic and manual segmented applicator regions was 0.93. Average distance between tip positions and average Hausdorff distance between applicator channels determined by our method and manually were 0.64 mm and 0.68 mm, respectively. Although trained only using CT images of T&O cases, our tool can also digitize Y-tandem, cylinder applicator, and T&O applicator scanned in cone-beam CT with error of tip position and Hausdorff distance <1 mm. Computation time was ∼15 s per case. ConclusionsWe have developed a deep-learning–assisted applicator digitization tool for 3D CT image–based HDRBT of gynecological cancer. The achieved accuracy, efficiency, and robustness made our tool clinically attractive.