Hierarchical Machine Learning-Based Growth Prediction Model of Panax ginseng Sprouts in a Hydroponic Environment.

Due to an increase in interest towards functional and health-related foods, Panax ginseng sprout has been in the spotlight since it contains a significant amount of saponins which have anti-cancer, -stress, and -diabetic effects. To increase the amount of production as well as decrease the cultivation period, sprouted ginseng is being studied to ascertain its optimal cultivation environment in hydroponics. Although there are studies on functional components, there is a lack of research on early disease prediction along with productivity improvement. In this study, the ginseng sprouts were cultivated in four different hydroponic conditions: control treatment, hydrogen-mineral treatment, Bioblock treatment, and highly concentrated nitrogen treatment. Physical properties were measured, and environmental data were acquired using sensors. Using three algorithms (artificial neural networks, support vector machines, random forest) for germination and rottenness classification, and leaf number and length of stem prediction models, we propose a hierarchical machine learning model that predicts the growth outcome of ginseng sprouts after a week. Based on the results, a regression model predicts the number of leaves and stem length during the growth process. The results of the classifier models showed an F1-score of germination classification of about 99% every week. The rottenness classification model showed an increase from an average of 83.5% to 98.9%. Predicted leaf numbers for week 1 showed an average nRMSE value of 0.27, which decreased by about 33% by week 3. The results for predicting stem length showed a higher performance compared to the regression model for predicting leaf number. These results showed that the proposed hierarchical machine learning algorithm can predict germination and rottenness in ginseng sprout using physical properties.

Variational autoencoder-based estimation of chronological age and changes in morphological features of teeth.

This study led to the development of a variational autoencoder (VAE) for estimating the chronological age of subjects using feature values extracted from their teeth. Further, it determined how given teeth images affected the estimation accuracy. The developed VAE was trained with the first molar and canine tooth images, and a parallel VAE structure was further constructed to extract common features shared by the two types of teeth more effectively. The encoder of the VAE was combined with a regression model to estimate the age. To determine which parts of the tooth images were more or less important when estimating age, a method of visualizing the obtained regression coefficient using the decoder of the VAE was developed. The developed age estimation model was trained using data from 910 individuals aged 10-79. This model showed a median absolute error (MAE) of 6.99 years, demonstrating its ability to estimate age accurately. Furthermore, this method of visualizing the influence of particular parts of tooth images on the accuracy of age estimation using a decoder is expected to provide novel insights for future research on explainable artificial intelligence.

Egg Freshness Prediction Model Using Real-Time Cold Chain Storage Condition Based on Transfer Learning.

Maintaining and monitoring the quality of eggs is a major concern during cold chain storage and transportation due to the variation of external environments, such as temperature or humidity. In this study, we proposed a deep learning-based Haugh unit (HU) prediction model which is a universal parameter to determine egg freshness using a non-destructively measured weight loss by transfer learning technique. The temperature and weight loss of eggs from a laboratory and real-time cold chain environment conditions are collected from ten different types of room temperature conditions. The data augmentation technique is applied to increase the number of the collected dataset. The convolutional neural network (CNN) and long short-term memory (LSTM) algorithm are stacked to make one deep learning model with hyperparameter optimization to increase HU value prediction performance. In addition, the general machine learning algorithms are applied to compare HU prediction results with the CNN-LSTM model. The source and target model for stacked CNN-LSTM used temperature and weight loss data, respectively. Predicting HU using only weight loss data, the target transfer learning CNN-LSTM showed RMSE value decreased from 6.62 to 2.02 compared to a random forest regressor, respectively. In addition, the MAE of HU prediction results for the target model decreased when the data augmentation technique was applied from 3.16 to 1.39. It is believed that monitoring egg freshness by predicting HU in a real-time cold chain environment can be implemented in real-life by using non-destructive weight loss parameters along with deep learning.

Developing and Validating a Method for Separating Flavonoid Isomers in Common Buckwheat Sprouts Using HPLC-PDA.

Buckwheat sprouts that are synthesized during the germination process are rich in flavonoids, including orientin, vitexin, rutin, and their isomers (isoorientin, isovitexin, and quercetin-3-O-robinobioside, respectively). The purpose of this study was to optimize and validate an analytical method for separating flavonoid isomers in common buckwheat sprout extract (CSE). Factors, such as range, linearity, precision, accuracy, limit of detection, and limit of quantification, were evaluated for each standard using high-performance liquid chromatography (HPLC). On the basis of resolution and symmetry, a column temperature of 40 °C with 0.1% (v/v) acidic water and acetonitrile as mobile phases, at a flow rate of 1 mL min-1 were determined to be the optimal analytical conditions. Calibration curves for orientin, isoorientin, vitexin, isovitexin, and rutin exhibited good linearity with correlation coefficients of 0.9999 over the 6.25-100.00 µg mL-1 range. Recovery values of 96.67-103.60% confirmed that the method was accurate for all flavonoids. The relative standard deviations of intra-day repeatability and inter-day reproducibility confirmed method preciseness, with values of less than 5.21% and 5.40%, respectively. The developed method was used to analyze flavonoids in CSE, with isomers satisfactorily separated and simultaneously quantified. We demonstrated that the developed HPLC method can be used to monitor flavonoids in buckwheat sprouts.

Improving the ground reaction force prediction accuracy using one-axis plantar pressure: Expansion of input variable for neural network.

In this study, we describe a method to predict 6-axis ground reaction forces based solely on plantar pressure (PP) data obtained from insole type measurement devices free of space limitations. Because only vertical force is calculable from PP data, a wavelet neural network derived from a non-linear mapping function was used to obtain 3-axis ground reaction force in medial-lateral (GRFML), anterior-posterior (GRFAP) and vertical (GRFV) and 3-axis ground reaction moment in sagittal (GRFS), frontal (GRFF) and transverse (GRFT) data for the remaining axes and planes. As the prediction performance of nonlinear models depends strongly on input variables, in this study, three input variables - accumulated PP with respect to time, center of pressure (COP) pattern, and measurements of the opposite foot, which are calculable only with a PP device - were considered in order to improve prediction performance. To conduct this study, the golf swing motions of 80 subjects were characterized as unilateral movement and GRF patterns as functions of individual characteristics. The prediction model was verified with 5-fold cross-validation utilizing the measured values of two force plates. As a result, prediction model (correlation coefficient, r=0.73-0.97) utilized accumulated PP and PP data of the opposite foot and showed the highest prediction accuracy in left-foot GRFV, GRMF, GRMT and right-foot GRFAP, GRFML, GRMF, GRMT. Likewise, another prediction model (r=0.83-0.98) utilized accumulated PP and COP patterns as input and showed the best accuracy in left-foot GRFAP, GRFML, GRMS and right-foot GRFV, GRMS. New methods based on the findings of the present study are expected to help resolve problems such as spatial limitation and limited analyzable motions in existing GRF measurement processes.

Predicting Complete Ground Reaction Forces and Moments During Gait With Insole Plantar Pressure Information Using a Wavelet Neural Network.

In general, three-dimensional ground reaction forces (GRFs) and ground reaction moments (GRMs) that occur during human gait are measured using a force plate, which are expensive and have spatial limitations. Therefore, we proposed a prediction model for GRFs and GRMs, which only uses plantar pressure information measured from insole pressure sensors with a wavelet neural network (WNN) and principal component analysis-mutual information (PCA-MI). For this, the prediction model estimated GRFs and GRMs with three different gait speeds (slow, normal, and fast groups) and healthy/pathological gait patterns (healthy and adolescent idiopathic scoliosis (AIS) groups). Model performance was validated using correlation coefficients (r) and the normalized root mean square error (NRMSE%) and was compared to the prediction accuracy of the previous methods using the same dataset. As a result, the performance of the GRF and GRM prediction model proposed in this study (slow group: r = 0.840-0.989 and NRMSE% = 10.693-15.894%; normal group: r = 0.847-0.988 and NRMSE% = 10.920-19.216%; fast group: r = 0.823-0.953 and NRMSE% = 12.009-20.182%; healthy group: r = 0.836-0.976 and NRMSE% = 12.920-18.088%; and AIS group: r = 0.917-0.993 and NRMSE% = 7.914-15.671%) was better than that of the prediction models suggested in previous studies for every group and component (p < 0.05 or 0.01). The results indicated that the proposed model has improved performance compared to previous prediction models.

Prediction of ground reaction forces during gait based on kinematics and a neural network model.

Kinetic information during human gait can be estimated with inverse dynamics, which is based on anthropometric, kinematic, and ground reaction data. While collecting ground reaction data with a force plate is useful, it is costly and requires regulated space. The goal of this study was to propose a new, accurate methodology for predicting ground reaction forces (GRFs) during level walking without the help of a force plate. To predict GRFs without a force plate, the traditional method of Newtonian mechanics was used for the single support phase. In addition, an artificial neural network (ANN) model was applied for the double support phase to solve statically indeterminate structure problems. The input variables of the ANN model, which were selected to have both dependency and independency, were limited to the trajectory, velocity, and acceleration of the whole segment's mass centre to minimise errors. The predicted GRFs were validated with actual GRFs through a ten-fold cross-validation method, and the correlation coefficients (R) for the ground forces were 0.918 in the medial-lateral axis, 0.985 in the anterior-posterior axis, and 0.991 in the vertical axis during gait. The ground moments were 0.987 in the sagittal plane, 0.841 in the frontal plane, and 0.868 in the transverse plane during gait. The high correlation coefficients(R) are due to the improvement of the prediction rate in the double support phase. This study also proved the possibility of calculating joint forces and moments based on the GRFs predicted with the proposed new hybrid method. Data generated with the proposed method may thus be used instead of raw GRF data in gait analysis and in calculating joint dynamic data using inverse dynamics.

Upper torso and pelvis linear velocity during the downswing of elite golfers.

BACKGROUND: During a golf swing, analysis of the movement in upper torso and pelvis is a key step to determine a motion control strategy for accurate and consistent shots. However, a majority of previous studies that have evaluated this movement limited their analysis only to the rotational movement of segments, and translational motions were not examined. Therefore, in this study, correlations between translational motions in the 3 axes, which occur between the upper torso and pelvis, were also examined. METHODS: The experiments were carried out with 14 male pro-golfers (age: 29 ± 8 years, career: 8.2 ± 4.8years) who registered in the Korea Professional Golf Association (KPGA). Six infrared cameras (VICON; Oxford Metrics, Oxford, UK) and SB-Clinc software (SWINGBANK Ltd, Korea) were used to collect optical marker trajectories. The center of mass (CoM) of each segment was calculated based on kinematic principal. In addition, peak value of CoM velocity and the time that each peak occurred in each segment during downswing was calculated. Also, using cross-correlation analysis, the degree of coupling and time lags of peak values occurred between and within segments (pelvis and upper torso) were investigated. RESULTS: As a result, a high coupling strength between upper torso and pelvis with an average correlation coefficient = 0.86 was observed, and the coupling between segments was higher than that within segments (correlation coefficient = 0.81 and 0.77, respectively). CONCLUSIONS: Such a high coupling at the upper torso and pelvis can be used to reduce the degree of motion control in the central nervous system and maintain consistent patterns in the movement. The result of this study provides important information for the development of optimal golf swing movement control strategies in the future.