Mean Absolute Error Range Research Articles

A deep learning model for brain age prediction using minimally preprocessed T1w images as input.

In the last few years, several models trying to calculate the biological brain age have been proposed based on structural magnetic resonance imaging scans (T1-weighted MRIs, T1w) using multivariate methods and machine learning. We developed and validated a convolutional neural network (CNN)-based biological brain age prediction model that uses one T1w MRI preprocessing step when applying the model to external datasets to simplify implementation and increase accessibility in research settings. Our model only requires rigid image registration to the MNI space, which is an advantage compared to previous methods that require more preprocessing steps, such as feature extraction. We used a multicohort dataset of cognitively healthy individuals (age range = 32.0-95.7 years) comprising 17,296 MRIs for training and evaluation. We compared our model using hold-out (CNN1) and cross-validation (CNN2-4) approaches. To verify generalisability, we used two external datasets with different populations and MRI scan characteristics to evaluate the model. To demonstrate its usability, we included the external dataset's images in the cross-validation training (CNN3). To ensure that our model used only the brain signal on the image, we also predicted brain age using skull-stripped images (CNN4). The trained models achieved a mean absolute error of 2.99, 2.67, 2.67, and 3.08 years for CNN1-4, respectively. The model's performance in the external dataset was in the typical range of mean absolute error (MAE) found in the literature for testing sets. Adding the external dataset to the training set (CNN3), overall, MAE is unaffected, but individual cohort MAE improves (5.63-2.25 years). Salience maps of predictions reveal that periventricular, temporal, and insular regions are the most important for age prediction. We provide indicators for using biological (predicted) brain age as a metric for age correction in neuroimaging studies as an alternative to the traditional chronological age. In conclusion, using different approaches, our CNN-based model showed good performance using one T1w brain MRI preprocessing step. The proposed CNN model is made publicly available for the research community to be easily implemented and used to study ageing and age-related disorders.

How to improve the accuracy of height data from bird tracking devices? An assessment of high-frequency GPS tracking and barometric altimetry in field conditions

BackgroundIn the context of rapid development of wind energy infrastructure, information on the flight height of birds is vital to assess their collision risk with wind turbines. GPS tags potentially represent a powerful tool to collect flight height data, yet GPS positions are associated with substantial vertical error. Here, we assessed to what extent high-frequency GPS tracking with fix intervals of 2–3 s (GPS remaining turned on between fixes), or barometric altimetry using air pressure loggers integrated in GPS tags, improved the accuracy of height data compared to standard low-frequency GPS tracking (fix interval ≥ 5 min; GPS turned off between fixes).ResultsUsing data from 10 GPS tag models from three manufacturers in a field setting (194 tags deployed on free-living raptors), we estimated vertical accuracy based on periods when the birds were stationary on the ground (true height above ground was approximately zero), and the difference between GPS and barometric height in flight. In GPS height data, vertical accuracy was mainly driven by noise (little bias), while in barometric data, it was mostly affected by bias (little noise). In high-frequency GPS data, vertical accuracy was improved compared to low-frequency data in each tag model (mean absolute error (AE) reduced by 72% on average; range of mean AE 2–7 vs. 7–30 m). In barometric data, vertical accuracy did not differ between high- and low-frequency modes, with a bias of − 15 to − 5 m and mean AE of 7–15 m in stationary positions. However, the median difference between GPS and barometric data was smaller in flight positions than in stationary positions, suggesting that the bias in barometric height data was smaller in flight. Finally, simulations showed that the remaining vertical error in barometric and high-frequency GPS data had little effect on flight height distributions and the proportion of positions within the collision risk height range, as opposed to the extensive noise found in low-frequency GPS data in some tag models.ConclusionsBarometric altimetry may provide more accurate height data than standard low-frequency GPS tracking, but it involves the risk of a systematic error. Currently, high-frequency GPS tracking provides highest vertical accuracy and may thus substantially advance the study of wind turbine collision risk in birds.

Robust dynamic control algorithm for uncertain powered wheelchairs based on sliding neural network approach

<abstract><p>The dynamic model of mobile wheelchair technology requires developing and implementing an intelligent control system to improve protection, increasing performance efficiency, and creating precise maneuvering in indoor and outdoor spaces. This work aims to design a robust tracking control algorithm based on a reference model for operating the kinematic model of powered wheelchairs under the variation of system parameters and unknown disturbance signals. The control algorithm was implemented using the pole placement method in combination with the sliding mode control (PP-SMC) approach. The design also adopted a neural network approach to eliminate system uncertainties from perturbations. The designed method utilized the sinewave signal as an essential input signal to the reference model. The stability of a closed-loop control system was achieved by adopting the Goa reaching law. The performance of the proposed tracking control system was evaluated in three scenarios under different conditions. These included assessing the tracking under normal operation conditions, considering the tracking performance by changing the dynamic system's parameters and evaluating the control system in the presence of uncertainties and external disturbances. The findings demonstrated that the proposed control method efficiently tracked the reference signal within a small error based on mean absolute error (MAE) measurements, where the range of MAE was between 0.08 and 0.12 in the presence of uncertainties or perturbations.</p></abstract>

Validating the accuracy of multiple sediment fingerprinting methods

AbstractSediment source fingerprinting provides an effective and reliable way for targeted control of soil erosion. However, the accuracy of source contribution predictions of different fingerprinting methods has not been well verified, which affects the further application of these methods. Here, compared with the real sediment sources measured by indoor simulated scouring experiments, we verified the accuracy of two tracers (geochemical elements and n‐alkanes) and three mixing models (Walling model, MixSIAR, and Landwehr model). The results showed that geochemical properties have high accuracy in identifying sediment sources, with the mean absolute error (MAE) ranging from 3.5%–5.9%, while poor results were shown for n‐alkanes (MAE values 7.5%–45.7%). The Walling model and MixSIAR showed higher accuracy of contribution prediction, and the MAE was 3.5%–5.9% and 0.7%–3.5%, respectively. Although Landwehr also presented high accuracy of model output, it was relatively less robust compared with the former two mixing models (the MAE range was 3.8%–9.7%). Furthermore, the mixing model with three correction factors (particle size, organic matter (OM), and particle size with OM) showed poor output results for scouring experiments. This study provides a reference for the selection and determination of tracers, mixing models, and correction factors in field experiments.

The Effect of Mandibular Angulation on Preoperative Assessment of Dental Implant Insertion at Premolar Region: CBCT Study.

Backgrounds This study investigated the effect of mandibular angulation on the perioperative evaluation of the implant placement at the premolar region on panoramic reconstructed images using cone beam computed tomography (CBCT). Methods A total of six dried anonymous human mandibles was included. Two implants were inserted in the left and right first premolar region. CBCT scans were obtained from each mandible at the standard position, +20° extension, and -20° flexion. The distance of the implant from the anterior loop of the inferior alveolar nerve and mental foramen was measured. Mean absolute error (MAE) of the distance of the implant from both anatomical landmarks was measured. The Wilcoxon matched-pair signed-rank test was used for the comparison of the measurements. All data were analyzed with the Stata program (version 15.1). Results No statistically significant differences were found between the distance of the implant from the mental foramen and the anterior loop of the inferior alveolar nerve up to 20° extension and flexion at both sides of the mandible. (All p > 0.1) However, a variable range of MAE (SD) in the distance of the implant from different anatomical landmarks was found (0.9 ± 0.7 to 3.3 ± 2.1). Conclusions We found no statistical difference in measurements of the distance of the implant from two anatomical landmarks at different head positions up to 20° extension and flexion. However, clinically, variable range in the distance of the implant from anatomical landmarks should be considered. Our findings could alert dentists of the possibility of error up to 20° extension and flexion on the perioperative evaluation of dental implant placement.

A DNN-Based UVI Calculation Method Using Representative Color Information of Sun Object Images.

As outdoor activities are necessary for maintaining our health, research interest in environmental conditions such as the weather, atmosphere, and ultraviolet (UV) radiation is increasing. In particular, UV radiation, which can benefit or harm the human body depending on the degree of exposure, is recognized as an essential environmental factor that needs to be identified. However, unlike the weather and atmospheric conditions, which can be identified to some extent by the naked eye, UV radiation corresponds to wavelength bands that humans cannot recognize; hence, the intensity of UV radiation cannot be measured. Recently, although devices and sensors that can measure UV radiation have been launched, it is very difficult for ordinary users to acquire ambient UV radiation information directly because of the cost and inconvenience caused by operating separate devices. Herein, a deep neural network (DNN)-based ultraviolet index (UVI) calculation method is proposed using representative color information of sun object images. First, Mask-region-based convolutional neural networks (R-CNN) are applied to sky images to extract sun object regions and then detect the representative color of the sun object regions. Then, a deep learning model is constructed to calculate the UVI by inputting RGB color values, which are representative colors detected later along with the altitude angle and azimuth of the sun at that time. After selecting each day of spring and autumn, the performance of the proposed method was tested, and it was confirmed that accurate UVI could be calculated within a range of mean absolute error of 0.3.

7. An algorithm for using deep learning convolutional neural networks with 3-dimensional depth sensor imaging in scoliosis detection

<h3>BACKGROUND CONTEXT</h3> Timely intervention in growing individuals, such as brace treatment, relies on early detection of adolescent idiopathic scoliosis (AIS). To this end, several screening methods have been implemented. However, these methods have limitations in predicting the Cobb angle. <h3>PURPOSE</h3> This study aimed to evaluate the performance of a 3-dimensional depth sensor imaging system with a deep learning algorithm (DLA), in predicting the Cobb angle in AIS. <h3>STUDY DESIGN/SETTING</h3> Retrospective analysis of prospectively collected, consecutive, nonrandomized series of patients at 5 scoliosis centers in Japan. <h3>PATIENT SAMPLE</h3> One hundred sixty human subjects suspected to have AIS were included. <h3>OUTCOME MEASURES</h3> Patient demographics, radiographic measurements, and predicted Cobb angle derived from the DLA were the outcome measures for this study. <h3>METHODS</h3> One hundred sixty data files were shuffled into 5 datasets with 32 data files at random (dataset 1, 2, 3, 4 and 5) and 5-fold cross validation was performed. The relationships between the actual and predicted Cobb angles were calculated using Pearson's correlation coefficient analyses. The prediction performances of the network models were evaluated using mean absolute error (MAE) and root mean square error (RMSE) between the actual and predicted Cobb angles. The shuffling into five datasets and five-fold cross validation was conducted ten times. There were no study-specific biases related to conflicts of interest. <h3>RESULTS</h3> The mean age was 14.7 ± 2.4 years, and the average Cobb angle was 30° (range: 0˚ to 64˚). The range of correlation coefficients was 0.87 to 0.89 in the five-fold cross validation with 10 repeats. The correlation between the actual Cobb angle and the mean predicted Cobb angle with 10 repeats was 0.91. The range of MAEs was 4.4˚ to 4.7˚in the 5-fold cross validation with 10 repeats. The range of RMSEs was 5.8˚ to 6.3˚. The MAE and RMSE between the actual Cobb angle and the mean predicted Cobb angle with 10 repeats was 4.0˚ and 5.4˚ respectively. At a Cobb angle of 10˚, which confirms a diagnosis of scoliosis, the mean predicted Cobb angle with 10 repeats showed a sensitivity of 0.99, a specificity of 0.42, PPV of 0.95, NPV of 0.71, accuracy of 0.94, PLR of 1.69, and an NLR of 0.03. <h3>CONCLUSIONS</h3> The 3D depth sensor imaging system with its newly innovated CNN for regression is objective and has significant ability to predict the Cobb angle in children and adolescents. This system is expected to be used for screening scoliosis in clinics or physical examination at schools. In addition, this is also a possible surrogate for radiographs to monitor curve progression, preventing unnecessary X-rays for mild case of scoliosis. <h3>FDA DEVICE/DRUG STATUS</h3> This abstract does not discuss or include any applicable devices or drugs.

Riparian Zone Nitrogen Management through the Development of the Riparian Ecosystem Management Model (REMM) in a Formerly Glaciated Watershed of the US Northeast

The Riparian Ecosystem Management Model (REMM) was developed, calibrated and validated for both hydrologic and water quality data for eight riparian buffers located in a formerly glaciated watershed (upper Pawcatuck River Watershed, Rhode Island) of the US Northeast. The Annualized AGricultural Non-Point Source model (AnnAGNPS) was used to predict the runoff and sediment loading to the riparian buffer. Overall, results showed REMM simulated water table depths (WTDs) and groundwater NO3-N concentrations at the stream edge (Zone 1) in good agreement with measured values. The model evaluation statistics showed that, hydrologically REMM performed better for site 1, site 4, and site 8 among the eight buffers, whereas REMM simulated better groundwater NO3-N concentrations in the case of site 1, site 5, and site 7 when compared to the other five sites. The interquartile range of mean absolute error for WTDs was 3.5 cm for both the calibration and validation periods. In the case of NO3-N concentrations prediction, the interquartile range of the root mean square error was 0.25 mg/L and 0.69 mg/L for the calibration and validation periods, respectively, whereas the interquartile range of d for NO3-N concentrations was 0.20 and 0.48 for the calibration and validation period, respectively. Moreover, REMM estimation of % N-removal from Zone 3 to Zone 1 was 19.7%, and 19.8% of N against actual measured 19.1%, and 26.6% of N at site 7 and site 8, respectively. The sensitivity analyses showed that changes in the volumetric water content between field capacity and saturation (soil porosity) were driving water table and denitrification.

Dynamic toll prediction using historical data on toll roads: Case Study of the I-66 Inner Beltway

Providing the users of a dynamic tolling system with predictions of tolling prices and the travel time difference between the toll road and the alternative routes enables them to make their travel decisions before starting their trip. This study aims to provide accurate predictions of tolling price through training and testing random forest, multilayer perceptron, and long short-term memory models and compare them with the current situation that the best prediction is extending the current toll to the next timesteps. The prediction time horizon includes five 6-minute time intervals ahead of the present time. The prediction performance of models over the testing set reveals that while all the models were significantly better than the base model, the random forest outperforms all models. For instance, while in the trained models, the mean absolute error range is from $1.5 to $2.5 for the next six minutes to the next 30 minutes, respectively, the same measure in the base model is in the range of $2.5 to $6. The prediction of travel time difference along the toll road and its alternative route with the shortest travel time revealed that the multilayer perceptron performs marginally better than the base model. However, due to a relatively stable travel time difference, the current travel time difference is an acceptable prediction for the next 30 minutes prediction horizon.

A two-state, multi-criteria Markov chain model for stochastic solar generators

ABSTRACT This paper presents a method of building a two-state multi-criteria Markov chain for stochastic solar generators. The design is based on different criteria which are defined using some statistical thresholds such as quartiles and averages. These thresholds have been used for classifying the state of all individual data samples. The proposed model is analyzed via computer simulation on Matlab environment using the historical data recorded over a period of 29 years by three meteorological stations in Algeria (Tamanrasset, Ghardaïa, and Oran). In order to validate and compare the performance of our model, several statistical tests were performed. The root mean square error and mean absolute error range, respectively, from 0.18 to 0.26 h and −0.10 h to 0.16. The obtained results proved that the model is effective to predict the data. Moreover, this study can be applied in its methodology not only in Algeria sites but also in other countries of the world and can serve as a reference for different climates, which are Mediterranean coastal climate, desert climate, and Sahel climate. Hence, it is useful to the designers of solar energy systems.

An innovative dynamic model for an integrated solar combined cycle power plant under off-design conditions

The integrated solar combined cycle power plants are currently the most efficient way of converting solar energy into electricity. Increasing the thermal efficiency of these plants depends strongly on the dynamic characteristics of off-design conditions and the operational control strategies. An innovative mathematical model has been presented to investigate the dynamic behavior of an Integrated Solar Combined Cycle power plant in Hassi R'mel, Algeria under off-design conditions in this study. The proposed model is analyzed via computer simulation on the Matlab environment using the recent operating parameters of the plant and the meteorological data recorded by the Abener hybrid solar-gas central of Hassi R'mel. The simulation results indicate that the experimental data of the outlet temperature of the solar field and the power generated by the plant verify the predictions found by the model. The root mean square error and mean absolute error range respectively between 4.03 and 4.12 °C and 1.45–1.80 °C for the outlet temperature of the solar cycle and 1.61–1.72 MW and 0.67–0.75 MW for the net power under specified experimental conditions. It is revealed that the off-design conditions, particularly the direct normal irradiance and the wind speed, affect the Integrated Solar Combined Cycle performance significantly. The presented model is an effective tool that can provide a wealth of information to power plant operators and designers during operational plant design, and management to mitigate the adverse impact of possible off-design conditions on power production and to boost power production by management strategies.

Analysis on Precipitable Water Vapor over the Tibetan Plateau Using FengYun-3A Medium Resolution Spectral Imager Products

The Tibetan Plateau is the largest and highest plateau in the world, and its complex terrain affects the distribution of precipitable water vapor (PWV) in the atmosphere, which plays an important role in the weather and climate of East Asia. In this paper, the characteristics of PWV over the Tibetan Plateau are studied using the FengYun-3A Medium Resolution Spectral Imager (MERSI) water vapor products, which are retrieved from the MERSI raw images of Chinese second-generation polar orbit meteorological satellite. Firstly, the accuracy of the MERSI 5-minute water vapor product is validated using three referenced water vapor data from TERRA/MODIS, ground-based GPS, and AERONET sun photometer over the Tibetan Plateau. Then, the spatial distribution and seasonal variation of PWV over the plateau are analyzed, and the effects of topographic factors on PWV are discussed. The results indicate that the MERSI 5-minute water vapor product has a good accuracy over the Tibetan Plateau, which the mean absolute error of MERSI water vapor product is in the range of 28.91%-37.54%, the mean absolute error range between 1.87 and 2.76 millimeter (mm), and the mean bias is between -1.14 and 0.64 mm comparing three referenced data. The PWV content appears as a typical spatial pattern over the Tibetan Plateau where there is a decrease from east to west of the Tibetan Plateau with increasing elevation, with the highest values over the south of Tibet. A second pattern also appears over the eastern part of the Tibetan Plateau, where the PWV content in the Qaidam Basin and the south of Tarim Basin are also considerably high. The seasonal variation of PWV content over the Tibetan Plateau presents to be highest in summer, followed by autumn and spring, and lowest in winter. The PWV content changes periodically during the year, which fits with a quadratic polynomial over monthly scales. The topographical factors of the Tibetan Plateau were found to affect the water vapor, where the altitude and latitude are negatively correlated with water vapor, while the slope and longitude show a positive correlation with water vapor; however, the aspect does not appear to have any significant influence on water vapor.

Prediction of land surface temperatures for surface urban heat island assessment over Chandigarh city using support vector regression model

Rapid urbanization is one of the primary reasons for changing the local climate, and there is a high impact on the surrounding areas. Chandigarh is one of the fastest developing cities in India showing rapidly urbanizing agglomeration. Due to the rapid urbanization, natural land surfaces are being replaced by the anthropogenic materials which negatively impacts the ecosystem resulting in urban heat island (UHI) effect. Land surface temperature (LST) is the primary and vital step for the analysis of UHI effect. The present study has been conducted to predict the LSTs for the assessment of UHI effect of the area surrounding Chandigarh city. Remote sensing data from Moderate-Resolution Imaging Spectroradiometer (MODIS) sensor and Advanced Spaceborne Thermal Emission and Reflection Radiometer Global Digital Elevation Model (ASTER GDEM) have been used for the prediction of LST. In the study, support vector regression (SVR) model has been developed from LST values of previous three years along with enhanced vegetation index (EVI), road density (RD) and elevation as input parameters to predict LST. The results of the SVR model have been validated using the data of the year 2014. A comparison of the model estimated LST and measured LST indicates that the range of mean absolute error (MAE) and mean absolute percentage error (MAPE) varies between 0.521 K and 0.525 K and 0.181–0.187%, respectively. Hence, SVR model can be used as a significant tool to predict LST for the assessment of heat island effect at any location. From the sensitivity analysis, it is observed that LST was ultimately the most sensitive to the RD compared to EVI and elevation. The SVR model has been compared with artificial neural networks (ANN) model to estimate the skill score factor of the SVR model (Forecasted) with reference to the ANN (Referred) model. Skill scores calculated for the periods show positive values which clearly depicts the efficacy of SVR model compared to ANN model for better LST prediction.

Evaluation of 16 genotype-guided Warfarin Dosing Algorithms in 310 Korean Patients Receiving Warfarin Treatment: Poor Prediction Performance in VKORC1 1173C Carriers

PurposeThe purpose of this study was to evaluate the performance of 16 previously published warfarin dosing algorithms in Korean patients. MethodsThe 16 algorithms were selected through a literature search and evaluated using a cohort of 310 Korean patients with atrial fibrillation or cerebral infarction who were receiving warfarin therapy. FindingsA large interindividual variation (up to 11-fold) in warfarin dose was observed (median, 25 mg/wk; range, 7–77 mg/wk). Estimated dose and actual maintenance dose correlated well overall (r range, 0.52–0.73). Mean absolute error (MAE) of the 16 algorithms ranged from –1.2 to –20.1 mg/wk. The percentage of patients whose estimated dose fell within 20% of the actual dose ranged from 1.0% to 49%. All algorithms showed poor accuracy with increased MAE in a higher dose range. Performance of the dosing algorithms was worse in patients with VKORC1 1173TC or CC than in total (r range, 0.38–0.61 vs 0.52–0.73; MAE range, –2.6 to –28.0 mg/wk vs –1.2 to –20.1 mg/wk). ImplicationsThe algorithms had comparable prediction abilities but showed limited accuracy depending on ethnicity, warfarin dose, and VKORC1 genotype. Further studies are needed to develop genotype-guided warfarin dosing algorithms with greater accuracy in the Korean population.

Prediction of surface temperatures for the assessment of urban heat island effect over Ahmedabad city using linear time series model

Ahmedabad is one of the fastest growing metropolitan cities in India showing rapidly urbanizing agglomeration. Due to the rapid urbanization natural land surfaces are being replaced by anthropogenic materials which negatively impact the environment resulting in urban heat island (UHI) effect. Land surface temperature (LST) is the primary and key step for the analysis of UHI effect. The present study has been carried out to predict the land surface temperatures for the assessment of UHI effect of area surrounding Ahmedabad city. Remote sensing data from Moderate-Resolution Imaging Spectroradiometer (MODIS) and Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) sensors has been used for the present study for the prediction of LST. Linear time series (LTS) model has been developed from LST values of 10 years along with enhanced vegetation index (EVI), road density (RD) and elevation as other parameters to predict LST for the year subsequent to the 10 year data period. The results of the model have been validated using the observed data of year 2014. A comparison of the model predicted LST and observed LST shows that the range of mean absolute error (MAE) and mean absolute percentage error (MAPE) varies between 0.23K–0.31K and 0.08%–0.10%, respectively. High correlation exists between the model predicted and observed LST with an average R2 value of 0.96. Hence, LTS model can be used as a significant tool to predict LST for the assessment of UHI effect at any location.