Higher Absolute Error Research Articles

Radiological age assessment based on clavicle ossification in CT: enhanced accuracy through deep learning

BackgroundRadiological age assessment using reference studies is inherently limited in accuracy due to a finite number of assignable skeletal maturation stages. To overcome this limitation, we present a deep learning approach for continuous age assessment based on clavicle ossification in computed tomography (CT).MethodsThoracic CT scans were retrospectively collected from the picture archiving and communication system. Individuals aged 15.0 to 30.0 years examined in routine clinical practice were included. All scans were automatically cropped around the medial clavicular epiphyseal cartilages. A deep learning model was trained to predict a person’s chronological age based on these scans. Performance was evaluated using mean absolute error (MAE). Model performance was compared to an optimistic human reader performance estimate for an established reference study method.ResultsThe deep learning model was trained on 4,400 scans of 1,935 patients (training set: mean age = 24.2 years ± 4.0, 1132 female) and evaluated on 300 scans of 300 patients with a balanced age and sex distribution (test set: mean age = 22.5 years ± 4.4, 150 female). Model MAE was 1.65 years, and the highest absolute error was 6.40 years for females and 7.32 years for males. However, performance could be attributed to norm-variants or pathologic disorders. Human reader estimate MAE was 1.84 years and the highest absolute error was 3.40 years for females and 3.78 years for males.ConclusionsWe present a deep learning approach for continuous age predictions using CT volumes highlighting the medial clavicular epiphyseal cartilage with performance comparable to the human reader estimate.

FPGA Implementation of High-Performance Truncated Rounding based Approximate Multiplier with High-Level Synchronous XOR-MUX Full Adder

In research and development, the most emerging field in digital signal processing and image processing is rounded-based approximated signed and unsigned multipliers. In the present research, we propose some cutting-edge, Preformation, and logic simplification technology connected to processing the Discrete cosine transform (DCT) and Discrete wavelet transform (DWT) images for sharpening. This technology will yield a truncated shifter incorporated with logical XOR-MUX Full adder techniques. A reliable and cost-effective approximate signed and unsigned multiplier was created for the rounding method. While this more advanced technology includes many approximate multipliers, it sacrifices the ability to find the closest integer of a rounded value when combining signed and unsigned capabilities, resulting in higher absolute errors than other approximate multipliers based on rounding. This proposed work will introduce a novel method of Truncated Shifter Rounding-based Approximate Multiplier integrated with a High-Level Synchronous XOR-MUX Full Adder design to minimize the number of logic gates and power consumption in the multiplier architecture. The Truncated RoBA (Rounding-based Approximate Multiplier) with XOR MUX Full Adder will reduce the logic size in the shifter and the arithmetic circuit. The work will modify this rounding-based approximate multiplier to minimize area, delay, and power consumption. This proposed architecture will be integrated with two fundamental changes: firstly, its Barrel shifter method will be replaced with a truncated shifter multiplier with XOR MUX Full Adder, and secondly, the parallel prefix Brent Kung adder will be replaced with a carrying-save adder with XOR MUX Full Adder. Finally, this architecture was designed using Verilog-HDL and synthesized using the Xilinx Vertex-5 FPGA family, targeting the device Xc7Vx485tFFg1157-1. It resulted in a reduction of area LUT (34%), power (1%), delay (32%), and error analysis (75%) when compared to the existing RoBA.

Learning time-averaged turbulent flow field of jet in crossflow from limited observations using physics-informed neural networks

Physics-informed neural networks (PINNs) are becoming popular in solving fluid mechanics problems forwardly and inversely. However, under limited observations, the application of PINNs was found to be difficult in solving the inverse problems of three-dimensional Reynolds-averaged Navier–Stokes (RANS) equations. In this study, the classical turbulent case of jet in crossflow was representatively adopted into the investigation. The dataset was obtained from a high-fidelity large-eddy simulation. The tensor-basis eddy viscosity (t-EV) model was imported first into the structure of PINNs as prior knowledge. Observations of five measured planes were preliminarily used to reconstruct the time-averaged turbulent flow field. After embedding the t-EV model, the highest absolute error and the relative L2 error of streamwise velocity were reduced by 11.1% and 31.4%, respectively. To cut down the volume of limited observations, a more effective training dataset containing only two planes and two pairs of lines was determined based on the flow characteristics (e.g., shear layer and counter-rotating vortex pair). Compared with those of five planes, the highest absolute error and the relative L2 error of streamwise velocity were further reduced by 30.0% and 6.4%, respectively. The investigation in this study provided an alternative to resolve the inverse problems of three-dimensional RANS equations with limited observations, which extended the deep learning application in fluid mechanics.

Comparison of decision tree based ensemble methods for prediction of photovoltaic maximum current

• Irradiance can be used as a direct reference generator to avoid tracking oscillations. • Extracting the parasitic resistances in Single Diode Model can assess aging. • Solar panel maximum current is directly proportional to irradiance variations. • AdaBoost is sensitive to noisy datasets and extreme outliers. The intermittent nature of the output power of photovoltaic (PV) systems, in addition to the fast-varying solar irradiance, has prompted the development of fast, accurate, and reliable forecasting techniques. This paper presents a comparative study of five ensemble machine learning methods based on bootstrap aggregating and gradient boosting for PV applications, namely AdaBoost, LightGBM, XGBoost, Random Forest, and CatBoost. A dataset of fast-varying environmental conditions was collected, and the terminal current of the experimental setup was augmented by applying a mathematical model, along with an evolutionary algorithm to extract the parasitic resistances found in the Single Diode Model (SDM) to accommodate for the aging effect. The mathematical model was evaluated for several irradiance and temperature levels against manufacturer Standard Test Conditions (STC), and then variations of environmental conditions were adjusted based on the manufacturer datasheet. CatBoost showed the lowest overall absolute error distribution (with respect to the mean and standard deviation) of all methods, and the best performance in terms of the absolute error (0.25%) and its standard deviation (0.85%) relative to the mathematical model. The AdaBoost method had the highest absolute error (34.5%) and a standard deviation of (15.8%). After hyperparameters tuning, CatBoost still outperformed other methods and showed consistency of high accuracy above 99% in performance with respect to a testing dataset, in addition, to having the largest area under the curve using the trapezoidal rule. Therefore, the CatBoost prediction method is expected to be an effective technique for maximum power point tracking schemes under fast-varying environmental conditions.

The Automatic Brain Tumor Segmentation Based on MRI Using Optimal Morphology Thresholding Methods

This article presents the segmentation of the brain tumors in the MRI images by using the optimal morphology thresholding methods. The number of patients with brain diseases is increased. Therefore, the needs for MRI are increased. Accordingly, the accurate and quick segmentation and identification methods for the brain tumors are really necessary. These also include planning and diagnosis tools for the automatic segmentation of the brain tumors in the MRI images. The optimal morphology thresholding methods are new methods that can automatically solve problems and diagnose the diseases. These consist of the RGB to Grayscale conversion process, the image quality improvement process and the optimum thresholding process, respectively. By testing the methods with the MRI images, it was found that the proposed methods could automatically segment the brain tumors in the MRI images with the shapes similar to that from the public databases. The highest absolute error was 3.96%, and the accuracy was 98.00% as compared to the other methods.

Does secondary cognitive task affect knee force production sense in young male soccer players?

Objective: The aim of the present study is to compare the knee force production sense error (KFPSE) of young soccer players under single and dual-task conditions. Materials and Methods: Twenty-four young male soccer players with a soccer license participated in this study. Knee force production sense (KFPS) was evaluated using a biofeedback device. In order to compare the KFPSE first, the bilateral KFPS of the participants was evaluated. Then the same assessment was performed concurrently with a cognitive dual task (five countdowns from 200). Results: The results revealed that, compared with the single-task condition, players had higher absolute error under the dual-task condition for both extremities (p˂0.05). Conclusion: Secondary cognitive task may be a possible critical factor reducing the accuracy of the force (re)production sense. A secondary cognitive task may adversely affect the KFPS.

Prediction of drilling induced delamination and circularity deviation in GFRP nanocomposites using deep neural network

Drilling of Glass Fiber Reinforced Polymer (GFRP) nanocomposites is most prevailing topic to understand the composite behaviour under different cutting conditions. The present study is mainly focused on prediction of drilling output responses such as delamination factor and circularity error randomly with the help of deep neural network (DNN) model. L9 orthogonal array is used for experimentation. Drilling operation is performed on 0.3 wt% multi-walled carbon nano tubes reinforced GFRPs with solid carbide, TiCN and TiAlN coated (6 mm- diameter) twist drills. Based on experimental results, two different deep neural network models are prepared with single and double hidden layers by varying node numbers such as 8, 16, 32, 64, and 128. Thrust force, Acoustic Emission RMS voltage, and drill type (coating) are considered as input to the neural network and delamination factor at exit, circularity error are treated as predicted output responses for the given network model. The revealed predicted results recommended that two hidden layers with 32 nodes network model give the lowest absolute error of 0.08% and 3.13% in delamination factor and circularity errors respectively. Similarly, the highest absolute error is identified as 4.19% in delamination factor and 13.14% in circularity error by single hidden layer with 128 nodes. Therefore, it is urged that DNN is the most suitable modelling technique for prediction of drilling responses on GFRP nano composites.

Applying GMDH artificial neural network in modeling CO2 emissions in four nordic countries

CO2 emission depends on several parameters. Due to environmental issues, it is necessary to find influential factors on CO2 emission as one of the most critical greenhouse gases. Type of utilized fuels and their share in total primary energy consumption, Gross Domestic Product (GDP) as an indicator for economic activities and the share of renewable energies play key role in the amount of CO2 emission. In the present study, Group method of data handling (GMDH) is applied in order to model CO2 emission as a function of consumption of various fuels, renewable energies and GDP. Obtained data showed that GMDH is an appropriate approach to predict CO2 emission. Comparing between actual data and GMDH output indicates that the R-squared value for the proposed model is equal to 0.998 which shows its high accuracy. In addition, it is observed that the highest absolute error by using GMDH artificial neural network is lower than 4%. The absolute relative error for more than 66% of data is lower than 1% which is another criterion demonstrating acceptable accuracy of the proposed model.

Heat transfer through window frames in EnergyPlus: model evaluation and improvement

ABSTRACTWindow frames can significantly affect the energy demand of buildings. Due to their complex non-planar geometry, two-dimensional, time-consuming simulations are necessary to estimate with precision heat transfer through frames. However, in dynamic heat transfer calculation through computer-based building performance simulation tools, simplifications are made to reduce simulation time. In particular, EnergyPlus models window frames as rectangular profiles, and uses a one-dimensional heat transfer model. In order to evaluate whether this simplification is legitimate, thermal transmittance values were calculated for a selection of frames, on one hand through two-dimensional finite elements simulation, and on the other hand, through a spreadsheet that reproduces the steps of the algorithm currently used by EnergyPlus to calculate heat transfer through window frames. The comparison of both values showed relative errors higher than 20%, including for rectangular frames (without simplifying geometry). Aluminium frames without thermal break exhibited the highest absolute errors, up to 2.98 W/m2K. An alternative algorithm was proposed, which uses compactness factors for film coefficient calculation and reduced emissivities to take into account self-viewing surfaces. The proposed model improved the estimates, reducing the absolute error to less than 0.10 W/m2K in 52% of the occurrences, and to less than 0.47 W/m2K for aluminium windows.

The empirical relationship between PM2.5 and AOD in Nanjing of the Yangtze River Delta

Ground-based Aerosol optical properties and PM2.5 concentrations in an industrial area in Nanjing were measured and analyzed from May 2013 to May 2015. The annual mean of the aerosol optical depth (AOD) at 550 nm was 1.04 ± 0.63. The annual mean of the Angstrom exponent (α) was 1.03 ± 0.23. The annual PM2.5 concentration was 62 ± 13 μg m−3. There were significant correlations between the PM2.5 concentration and the AOD. The linear regression functions of the daily PM2.5 concentration (y) with the ground-based AOD (x) and the MODIS AOD (x) were y = 55.70x + 25.60 (R2 = 0.53) and y = 63.84x + 39.33 (R2 = 0.56), respectively. The results demonstrated significant differences in the slopes and intercepts of the linear regression functions for different seasons. The distribution of PM2.5 was retrieved by the seasonal functions between the ground-based PM2.5 and the MODIS AOD on sunny days in the region. The MODIS-derived PM2.5 concentrations in the north of Jiangsu and parts of Shandong and Henan were generally higher than those in the hilly lands of the south of Anhui and the northwest of Zhejiang. There was a high-PM2.5 area anchored over the industrial area. The particle pollution was more serious in spring and winter than in autumn and summer. The MODIS-derived PM2.5 concentration in urban and industrial areas was higher than 80 μg m−3 during spring and winter. The retrieved PM2.5 values had the highest absolute error (+17 μg m−3) in summer and the lowest (+2 μg m−3) in winter, and the average relative error of the seasonal PM2.5 concentration was 15% ± 10%.

Near ground propagation model for pine tree forest environment

A propagation model experimentally derived from both free space and near ground plane earth path loss models for pine tree forest with foliage depth lower than 400m has been proposed. It is a piecewise model taking foliage depth of 200m as a boundary. The model is compared with several well-known models, namely Weissberger, ITU-R and COST235. Different from other models, the proposed one takes into account trunk height gain k. Observed average error is about 6dB for proposed model whereas it is about 32dB, 16dB and 20dB for COST235, Weissberger and ITU-R models, respectively. Also, it is observed that COST235 estimates the highest path loss among the models and it has the highest absolute errors especially in a forest with low foliage density.

Allocation of Attentional Resources toward a Secondary Cognitive Task Leads to Compromised Ankle Proprioceptive Performance in Healthy Young Adults

The objective of the present study was to determine whether increased attentional demands influence the assessment of ankle joint proprioceptive ability in young adults. We used a dual-task condition, in which participants performed an ankle ipsilateral position-matching task with and without a secondary serial auditory subtraction task during target angle encoding. Two experiments were performed with two different cohorts: one in which the auditory subtraction task was easy (experiment 1a) and one in which it was difficult (experiment 1b). The results showed that, compared with the single-task condition, participants had higher absolute error under dual-task conditions in experiment 1b. The reduction in position-matching accuracy with an attentionally demanding cognitive task suggests that allocation of attentional resources toward a difficult second task can lead to compromised ankle proprioceptive performance. Therefore, these findings indicate that the difficulty level of the cognitive task might be the possible critical factor that decreased accuracy of position-matching task. We conclude that increased attentional demand with difficult cognitive task does influence the assessment of ankle joint proprioceptive ability in young adults when measured using an ankle ipsilateral position-matching task.

Analysis of Knee Flexion Angles During 2 Clinical Versions of the Heel Raise Test to Assess Soleus and Gastrocnemius Function

Controlled laboratory study, using a repeated-measures, counterbalanced design. To provide estimates on the average knee angle maintained, absolute knee angle error, and total repetitions performed during 2 versions of the heel raise test. The heel raise test is performed in knee extension (EHRT) to assess gastrocnemius and knee flexion (FHRT) for soleus. However, it has not yet been determined whether select knee angles are maintained or whether total repetitions differ between the clinical versions of the heel raise test. Seventeen healthy males and females performed maximal heel raise repetitions in 0° (EHRT) and 30° (FHRT) of desired knee flexion. The average angle maintained and absolute error at the knee during the 2 versions, and total heel raise repetitions, were measured using motion analysis. Participants' kinematic measures were fitted into a generalized estimation equation model to provide estimates on EHRT and FHRT performance applicable to the general population. The model estimates that average angles of 2.2° and 30.7° will be maintained at the knee by the general population during the EHRT and the FHRT, with an absolute angle error of 3.4° and 2.5°, respectively. In both versions, 40 repetitions should be completed. However, the average angles maintained by participants ranged from -6.3° to 21.6° during the EHRT and from 22.0° to 43.0° during the FHRT, with the highest absolute errors in knee position being 25.9° and 33.5°, respectively. On average, select knee angles will be maintained by the general population during the select heel raise test versions, but individualized performance is variable and total repetitions do not distinguish between versions. Clinicians should, therefore, interpret select heel raise test outcomes with caution when used to respectively assess and rehabilitate soleus and gastrocnemius function.

Relationship of gender, age, and body mass index to errors in predicted kidney function

Previous studies have shown conflicting data on accuracy of equations for kidney function prediction. The present work analysed the relationship of gender, age and body mass index (BMI) to error of predictions by the Cockcroft-Gault equation (CG(eq)), the simplified equation of the Modification of Renal Diseases Study (MDRD(eq)) and the Mayo Clinic equation (Mayo(eq)). Inulin clearance (glomerular filtration rate; GFR) and other variables were measured in 380 subjects of both sexes, aged 18-88 years, with and without kidney disease. GFR was defined as low when <60 ml/min x 1.73 m2. BMI was used for definition of underweight/overweight. Relative error of predictions was used as an index of bias. It was calculated as prediction minus GFR (positive values =overestimates, negative values = underestimates) and expressed as a percentage of the GFR. Absolute error was used as an index of imprecision and was calculated as the absolute value of relative error. CG(eq) relative error was inversely associated with age and directly associated with BMI (P<0.001), but not with gender or GFR. MDRD(eq) relative error was inversely associated with female gender and GFR (P<0.001), but not with age or BMI. Mayo(eq) relative error was directly associated with male gender, BMI and GFR (P<0.01), but not with age. Absolute error was higher for CG(eq) than for MDRD(eq) but only at low GFR (P<0.001). Mayo(eq) had a higher absolute error than CG(eq) and MDRD(eq) (P<0.01). Errors of predictions varied not only with GFR but also with gender, age and BMI. Without using creatinine assay calibration, Mayo(eq) was less accurate than both MDRD(eq) and CG(eq), whereas MDRD(eq) was slightly more precise than CG(eq) but only at low GFR.