Training Accuracy Research Articles

Object classification through heterogeneous fog with a fast data-driven algorithm using a low-cost single-photon avalanche diode array

This study presents a framework for classifying a wooden mannequin’s poses using a single-photon avalanche diode (SPAD) array in dynamic and heterogeneous fog conditions. The target and fog generator are situated within an enclosed fog chamber. Training datasets are continuously collected by configuring the temporal and spatial resolutions on the sensor's firmware, utilizing a low-cost SPAD array sensor priced below $5, consisting of an embedded SPAD array and diffused VCSEL laser. An extreme learning machine (ELM) is trained for rapid pose classification, as a benchmark against CNN. We quantitatively justify the selection of nodes in the hidden layer to balance the computing speed and accuracy. Results demonstrate that ELM can accurately classify mannequin poses when obscured by dynamic heavy fog to 35 cm away from the sensor, enabling real-time applications in consumer electronics. The proposed ELM achieves 90.65% and 89.58% accuracy in training and testing, respectively. Additionally, we demonstrate the robustness of both ELM and CNN as the fog density increases. Our study also discusses the sensor’s current optical limitations and lays the groundwork for future advancements in sensor technology.

Fuzzy lattices assisted EJAYA Q-learning for automated pulmonary diseases classification

This work proposes a novel technique called Enhanced JAYA (EJAYA) assisted Q-Learning for the classification of pulmonary diseases, such as pneumonia and tuberculosis (TB) sub-classes using chest x-ray images. The work introduces Fuzzy lattices formation to handle real time (non-linear and non-stationary) data based feature extraction using Schrödinger equation. Features based adaptive classification is made possible through the Q-learning algorithm wherein optimal Q-values selection is done via EJAYA optimization algorithm. Fuzzy lattice is formed using x-ray image pixels and lattice Kinetic Energy (K.E.) is calculated using the Schrödinger equation. Feature vector lattices having highest K.E. have been used as an input features for the classifier. The classifier has been employed for pneumonia classification (normal, mild and severe) and Tuberculosis detection (presence or absence). A total of 3000 images have been used for pneumonia classification yielding an accuracy, sensitivity, specificity, precision and F-scores of 97.90%, 98.43%, 97.25%, 97.78% and 98.10%, respectively. For Tuberculosis 600 samples have been used. The achived accuracy, sensitivity, specificity, precision and F-score are 95.50%, 96.39%, 94.40% 95.52% and 95.95%, respectively. Computational time are 40.96 and 39.98 s for pneumonia and TB classification. Classifier learning rate (training accuracy) for pneumonia classes (normal, mild and severe) are 97.907%, 95.375% and 96.391%, respectively and for tuberculosis (present and absent) are 96.928% and 95.905%, respectively. The results have been compared with contemporary classification techniques which shows superiority of the proposed approach in terms of accuracy and speed of classification. The technique could serve as a fast and accurate tool for automated pneumonia and tuberculosis classification.

Evaluating the performance of a non-uniform squash function in Capsule networks for early diabetic retinopathy detection using fundus image analysis

Diabetic retinopathy, known as one of the most crucial neurovascular barriers to diabetes, has affected the retina, leading to a potential risk of blindness. The harm is not isolated to personal health, with negative effects felt at the economic and social level in entire communities. The consequences of vision loss secondary to diabetic retinopathy on quality of life, independent living, and employment can be significant. It is necessary to adopt a comprehensive approach for effective management of underlying diabetes, which includes early detection, timely intervention, supportive care, and management of comorbid conditions. Critical to this is a precise classification of its stages for disease analysis. A manual diagnosis, on the other hand, requires qualified individuals to recognize the subtle nuances of the condition, making it a lengthy process. This study aims to develop a computer-assisted system for analyzing retinal fundus images to identify and classify different stages of diabetic retinopathy. A dataset of eye fundus images was used by the proposed network to classify the presence and severity of blindness and its stages. The proposed model uses a non-uniform squash function in capsule neural networks. The motive behind employing a non-uniform squash function was to encourage the model to allocate greater attention and stability to intricate images throughout the training phase. The proposed model's efficiency is assessed meticulously with pre-trained CNN methods like InceptionV3, VGG16, VGG19, and Xception. The proposed capsule network model has been revamped to address the shortcomings of prior research with care. The method stands out with an impressive 99.84 % training accuracy and 98.68 % validation accuracy. Diabetic retinopathy, though rare in occurrence, poses significant challenges due to the high cost of diagnosis, rendering it inaccessible to rural and remote communities with limited access to tests and check-ups.

Emulating Low-Power Synaptic Plasticity in a Solution-Processed Oxide-Based Long Retention Memory Transistor with High Learning Accuracy.

The exploration of synaptic plasticity in metal-oxide-based ferroelectric thin-film transistors has been limited. As a perovskite ferroelectric material, LiNbO3 is widely studied; but its potential use as a neuromorphic device, like synaptic transistors, has not been realized. In this study, a solution-processed ferroelectric thin-film transistor (FeTFT) with an alternating layer of LiNbO3 and Li5AlO4 as a gate dielectric has been fabricated. This configuration reduces the depolarization field by leveraging the large ionic polarization of Li+ ions in the Li5AlO4 layer, while the wide bandgap helps mitigate the leakage current. FeTFT exhibits impressive transistor performance, including a saturation mobility of 0.478 cm2V-1 s-1, an on/off ratio of 3.08 × 103, and a low trap-state density of 1.3 × 1013 cm-2. Moreover, the device demonstrates good memory retention, retaining information for nearly 1 day. It successfully emulates synaptic plasticity, specifically short-term plasticity and long-term plasticity. Besides, a 94% training accuracy has been achieved through artificial neural network simulation. Notably, the FeTFT consumes minimal power, with energy consumption of approximately 3.09 nJ per synaptic event, which is remarkably low compared to other reported solution-processed FeTFT devices.

Image Enhancement Using Bidimensional Empirical Mode Decomposition and Morphological Operations for Brain Tumor Detection and Classification.

The three steps of brain image processing – preprocessing, segmentation, and classification are becoming increasingly important in patient care. The aim of this article is to present a proposed method in the mentioned three-steps, with emphasis on the preprocessing step, which includes noise removal and contrast enhancement. The fast and adaptive bidimensional empirical mode decomposition and the anisotropic diffusion equation as well as the modified combination of top-hat and bottom-hat transforms are used for noise reduction and contrast enhancement. Fast C-means clustering with enhanced image is used to detect tumors and the tumor cluster corresponds to the maximum centroid. Finally, Ensemble learning is used for classification. The Figshare brain tumor dataset contains magnetic resonance images used for data selection. The optimal parameters for both noise reduction and contrast enhancement are investigated using a tumor contaminated with Gaussian noise. The results are evaluated against state-of-the-art results and qualitative performance metrics to demonstrate the dominance of the proposed approach. The fast C-means algorithm is applied to detect tumors using twelve enhanced images. The detected tumors were compared to the ground truth and showed an accuracy and specificity of 99% each, and a sensitivity and precision of 90% each. Six statistical features are retrieved from 150 enhanced images using wavelet packet coefficients at level 4 of the Daubechies 4 wavelet function. These features are used to develop the classifier model using ensemble learning to create a model with training and testing accuracy of 96.7% and 76.7%, respectively. When this model is applied to classify twelve detected tumor images, the accuracy is 75%; there are three misclassified images, all of which belong to the pituitary disease group. Based on the research, it appears that the proposed approach could lead to the development of computer-aided diagnosis (CADx) software that physicians can use as a reference for the treatment of rain tumor.

Deep learning-based digitization of Kurdish text handwritten in the e-government system

Many government institutions in developing countries such as the Kurdistan Region of Iraq (KRI) keep a variety of paper-based records that are available in printed or handwritten format. The need for technology that turns handwritten writing into digital text is therefore highly demanded. E-government in developed and developing countries is a crucial facilitator for the provision of such services. This paper aims to develop a deep learning model based on the mask region convolutional neural network (mask-RCNN) to effectively digitize kurdish handwritten text recognition (KHTR). In this research, typical datasets, which includes the isolated handwritten Central Kurdish character images, an extensive database of 40,410 images, and 390 native writers have been produced to determine the developed approach’s performance in terms of identification rates. This approach achieves adequate outcomes in terms of training time and accuracy. The proposed model gives higher performance for detection, localization, and recognition when using a dataset containing many challenges, the results were 80%, 96%, and 87.6 for precision, recall, and F-score respectively. The findings revealed that the proposed model obtained better results compared to other similar works. The accuracy of optical character recognition (OCR) is more than 99%.

Real-Time CNN Training and Compression for Neural-Enhanced Adaptive Live Streaming.

We propose a real-time convolutional neural network (CNN) training and compression method for delivering high-quality live video even in a poor network environment. The server delivers a low-resolution video segment along with the corresponding CNN for super resolution (SR), after which the client applies the CNN to the segment in order to recover high-resolution video frames. To generate a trained CNN corresponding to a video segment in real-time, our method rapidly increases the training accuracy by promoting the overfitting property of the CNN while also using curriculum-based training. In addition, assuming that the pretrained CNN is already downloaded on the client side, we transfer only residual values between the updated and pretrained CNN parameters. These values can be quantized with low bits in real time while minimizing the amount of loss, as the distribution range is significantly narrower than that of the updated CNN. Quantitatively, our neural-enhanced adaptive live streaming pipeline (NEALS) achieves higher SR accuracy and a lower CNN compression loss rate within a constrained training time compared to the state-of-the-art CNN training and compression method. NEALS achieves 15 to 48% higher quality of the user experience compared to state-of-the-art neural-enhanced live streaming systems.

Remote sensing image road network detection based on channel attention mechanism

Extracting and detecting road network consistency from high-resolution remote sensing images has been a hot and difficult problem in the computer vision. Although it has made significant progress, there is still a phenomenon of high training accuracy but unsatisfactory actual extraction and detection results. The attention mechanism is one of the efficient and practical mechanisms in deep learning. It improves the performance of deep learning by selectively focusing on a portion of all information while ignoring other visible information, while effectively utilizing computing resources. Numerous experiments have also confirmed that the attention mechanism is resource-saving and effective. Its plug and play feature brings great convenience to programmers. In order to provide better road network detection results and solve the above problem, this paper combines the channel attention mechanism with ResNet and proposes SE-ResNet and ECA-ResNet for remote sensing image road network detection, making networks extract and learn road network features and ignore some non-road network features. The experimental results show that on the Massachusetts roads (MR) and CHN6-CUG roads datasets, ECA-ResNet and SE-ResNet based on channel attention mechanism perform similar to LeNet7 and ResNet in terms of accuracy, loss, accuracy convergence, and loss convergence, and even increase a certain computational burden. However, their final road network detection results (including road network detection pixel count, precision, recall, accuracy, IOU, F1 score, and actual road network detection result) of the former are significantly better than those of the latter. The channel attention mechanism makes the deep neural network pay more attention to the extraction and learning of road network features, while ignoring the extraction and learning of some non-road network features, which improves the accuracy of containing road network samples and reduces the accuracy of not containing road network samples. Therefore, the performance of ECA-ResNet and SE-ResNet is similar to that of LeNet7 and ResNet in the accuracy, loss, accuracy convergence and loss convergence, but the final road network detection results of ECA-ResNet and SE-ResNet are significantly better than those of LeNet7 and ResNet. Therefore, the proposed ECA-ResNet and SE-ResNet have broad application prospects in road network detection, especially ECA-ResNet.

Implementation of Convolutional Neural Networks (CNN) in An Emotion Detection System for Measuring Learning Concentration Levels

Technological advancements have had a significant impact on the education sector, including the application of Convolutional Neural Networks (CNN) for facial image analysis. This research aims to implement CNN to measure students' learning concentration levels. The FER2013 dataset, which includes seven emotion classifications and comprises 28,709 images for training data, is used as the database. The data is processed through rescaling and augmentation to prepare the CNN model. The model consists of several convolutional layers, pooling layers, and fully connected layers designed to extract crucial features from facial images. Evaluation results demonstrate a very high accuracy of 94.95% on training data, indicating that the model effectively recognizes complex patterns within the data. Although there is a higher loss value of 157% and a decreased accuracy of 62.75% on validation data, this suggests that the model possesses a strong foundational capability and can still be improved through further adjustments. With high accuracy in training and promising validation results, the model shows substantial potential for real-world application, where it can assist teachers in understanding students' emotional responses in real-time. The implementation of CNN aids educators in comprehending students' emotional responses and adapting their teaching methods more effectively, thereby creating a more conducive learning environment and enhancing students' academic and social development. These findings also open opportunities for further research to improve the performance and generalization of the model on unseen data, making this technology an increasingly reliable tool in education

Landsat-based spatiotemporal estimation of subtropical forest aboveground carbon storage using machine learning algorithms with hyperparameter tuning.

The aboveground carbon storage (AGC) in forests serves as a crucial metric for evaluating both the composition of the forest ecosystem and the quality of the forest. It also plays a significant role in assessing the quality of regional ecosystems. However, current technical limitations introduce a degree of uncertainty in estimating forest AGC at a regional scale. Despite these challenges, remote sensing technology provides an accurate means of monitoring forest AGC. Furthermore, the implementation of machine learning algorithms can enhance the precision of AGC estimates. Lishui City, with its rich forest resources and an approximate forest coverage rate of 80%, serves as a representative example of the typical subtropical forest distribution in Zhejiang Province. Therefore, this study uses Landsat remote sensing images, employing backpropagation neural network (BPNN), random forest (RF), and categorical boosting (CatBoost) to model the forest AGC of Lishui City, selecting the best model to estimate and analyze its forest AGC spatiotemporal dynamics over the past 30 years (1989-2019). The study shows that: (1) The texture information calculated based on 9×9 and 11×11 windows is an important variable in constructing the remote sensing estimation model of the forest AGC in Lishui City; (2) All three machine learning techniques are capable of estimating forest AGC in Lishui City with high precision. Notably, the CatBoost algorithm outperforms the others in terms of accuracy, achieving a model training accuracy and testing accuracy R2 of 0.95 and 0.83, and RMSE of 2.98 Mg C ha-1 and 4.93 Mg C ha-1, respectively. (3) Spatially, the central and southwestern regions of Lishui City exhibit high levels of forest AGC, whereas the eastern and northeastern regions display comparatively lower levels. Over time, there has been a consistent increase in the total forest AGC in Lishui City over the past three decades, escalating from 1.36×107 Mg C in 1989 to 6.16×107 Mg C in 2019. This study provided a set of effective hyperparameters and model of machine learning suitable for subtropical forests and a reference data for improving carbon sequestration capacity of subtropical forests in Lishui City.

Multi‐data classification detection in smart grid under false data injection attack based on Inception network

AbstractDuring operation, the smart grid is subject to different false data injection attacks (FDIA). If the different kinds of FDIAs and typical failures have been detected, the system operator can develop various defenses to protect the smart grid in multiple categories. Therefore, this article aims to propose a multi‐data classification detection model to differentiate the data of regular operation, faults, and FDIAs when the smart grid suffers FDIAs. Due to the unbalanced number of different kinds of samples in the dataset, Affinitive Borderlinen SMOTE is used to pre‐process the data by oversampling to improve the training accuracy. A multi‐data detection model based on the Inception network is established, and the overall structure of the network and the individual Inception modules are given. A small power system is an example of simulating a smart grid suffering from FDIAs. The designed classification detection model is simulated, validated, and compared with two‐dimensional convolutional neural networks and existing research results. The qualitative analysis of the evaluation metrics can show that the Inception network model has high accuracy and real‐time performance for detecting different data.

Schizophrenia diagnosis using the GRU-layer's alpha-EEG rhythm's dependability

Verifying schizophrenia (SZ) can be assisted by deep learning techniques and patterns in brain activity observed in alpha-EEG recordings. The suggested research provides evidence of the reliability of alpha-EEG rhythm in a Gated-Recurrent-Unit-based deep-learning model for investigating SZ. This study suggests Rudiment Densely-Coupled Convolutional Gated Recurrent Unit (RDCGRU) for the various EEG-rhythm-based (gamma, beta, alpha, theta, and delta) diagnoses of SZ. The model includes multiple 1-D-Convolution (Con-1-D) folds with steps greater than 1, which enables the model to programmatically and effectively learn how to reduce the incoming signal. The Con-1-D layers and numerous Gated Recurrent Unit (GRU) layers comprise the Exponential-Linear-Unit activation function. This powerful activation function facilitates in-deep-network training and improves classification performance. The Densely-Coupled Convolutional Gated Recurrent Unit (DCGRU) layers enable RDCGRU to address the training accuracy loss brought on by vanishing or exploding gradients, and this might make it possible to develop intense, deep versions of RDCGRU for more complex problems. The sigmoid activation function is implemented in the digital (binary) classifier's output nodes. The RDCGRU deep learning model attained the most excellent accuracy, 88.88 %, with alpha-EEG rhythm. The research achievements: The RDCGRU deep learning model's GRU cells responded superiorly to the alpha-EEG rhythm in EEG-based verification of SZ.

A bidirectional domain separation adversarial network based transfer learning method for near-infrared spectra

Traditional calibration transfer (CT) methods usually fail to adapt the source domain model to the target domain because of changes associated with the instrument, detection environment, sample composition, or sample type. Most deep transfer learning (DTL) methods have shown reliable capability in extracting domain-invariant representations. However, factors unique to the spectral domain limit their development and application, including difficulty in obtaining labeled samples, numerous spectral perturbation factors, and the scarcity of spectral databases. To address these challenges, we here introduce the Bidirectional Domain-Separating Adversarial Network (BiDSAN), based on a deep adversarial network transfer architecture, for learning and adapting to distribution/domain differences between sample spectra. BiDSAN incorporates Wasserstein divergence (w-div) and dynamic adversarial factor into the domain critic (discriminator) to accelerate the extraction of spectral shared components. Additionally, the use of the spectral signal distortion ratio as a loss function and the introduction of a bidirectional mechanism further enhanced the training accuracy and speed of spectral models. Calibration analyses were tested on pharmaceutical and two sets of wood datasets, corresponding to three common domain/distribution difference scenarios discussed in the study. Compared to existing CT and DTL methods, the proposed BiDSAN method demonstrated superior spectral signal calibration capability and stability, achieving unsupervised domain adaptation for similar samples and, for the first time, obtaining ideal prediction results in cross-species spectral domain adaptation. Our method represents significant potential for using DTL approaches in spectral domain adaptation.

Tomato Leaf Disease Classification by Combining EfficientNetv2 and a Swin Transformer

Recently, convolutional neural networks (CNNs) and self-attention mechanisms have been widely applied in plant disease identification tasks, yielding significant successes. Currently, the majority of research models for tomato leaf disease recognition rely solely on traditional convolutional models or Transformer architectures and fail to capture both local and global features simultaneously. This limitation may result in biases in the model’s focus, consequently impacting the accuracy of disease recognition. Consequently, models capable of extracting local features while attending to global information have emerged as a novel research direction. To address these challenges, we propose an Eff-Swin model that integrates the enhanced features of the EfficientNetV2 and Swin Transformer networks, aiming to harness the local feature extraction capability of CNNs and the global modeling ability of Transformers. Comparative experiments demonstrate that the enhanced model has achieved a further increase in training accuracy, reaching an accuracy rate of 99.70% on the tomato leaf disease dataset, which is 0.49~3.68% higher than that of individual network models and 0.8~1.15% higher than that of existing state-of-the-art combined approaches. The results show that integrating attention mechanisms into convolutional models can significantly enhance the accuracy of tomato leaf disease recognition while also offering the great potential of the Eff-Swin backbone with self-attention in plant disease identification.

SILP: Enhancing skin lesion classification with spatial interaction and local perception

Because of the harmful effects of ultraviolet rays and global environmental factors, the number of patients with skin lesions is increasing. If left untreated, skin lesions may lead to skin cancer. However, limited access to specialized medical care remains a challenge in certain regions. Therefore, there is an urgent need for an efficient, accurate, and accessible tool to identify suspicious lesions. Although there are many classification models for skin lesions, there is still room for improvement in terms of accuracy. To enhance the accuracy of skin lesion classification, a novel system named SILP is proposed in this study. There are two modules in SILP: the LPM and the SIM. Additionally, we have changed the activation function from GELU to SiLU to improve both training time and accuracy. SILP, along with several other models, has been tested on two public skin lesion datasets. The results demonstrate that our proposed system outperforms the state-of-the-art skin lesion classification model, not only in terms of accuracy but also in various other evaluation metrics.

Biofeedback and Exercise Load Affect Accuracy of Tongue Strength Exercise Performance.

Rehabilitative exercises require precise movement coordination and target accuracy for optimal effectiveness. This paper explores the impact of tongue strength exercises (TSE) performance accuracy on exercise outcomes, adherence, and participant confidence and motivation. An 8-week randomized clinical trial included 84 typically aging participants divided into four groups defined by access to biofeedback (present/absent) and TSE intensity dosing (maximal/submaximal) during a home exercise program (HEP). Retention, training, and HEP accuracy were tracked at biweekly visits and during HEP for participants with access to a biofeedback device. Associations with tongue strength outcomes, participant factors, biofeedback, and intensity dosing were analyzed. Exercise accuracy measures did not contribute to tongue strength outcomes at the end of 8weeks. Increased training accuracy (less practice required to achieve competency) was associated with higher participant confidence and better adherence to the HEP. The presence of biofeedback was associated with reduced adherence but better retention accuracy, while maximal intensity was associated with improvements in all accuracy measures compared to submaximal intensity exercise. These findings in typically aging participants suggest the need for tailored approaches in swallowing-related exercise programs, given the effects of biofeedback and exercise intensity on motor learning and exercise retention. Accuracy performance and its effect on clinical outcomes warrants study in clinical populations with dysphagia and with various rehabilitative approaches.Trial Registration Clincialtrials.gov: NCT04809558.

Computer-Aided Diagnosis of Diabetic Retinopathy Lesions Based on Knowledge Distillation in Fundus Images

At present, the early diagnosis of diabetic retinopathy (DR), a possible complication of diabetes due to elevated glucose concentrations in the blood, is usually performed by specialists using a manual inspection of high-resolution fundus images based on lesion screening, leading to problems such as high work-intensity and accessibility only in specialized health centers. To support the diagnosis of DR, we propose a deep learning-based (DL) DR lesion classification method through a knowledge distillation (KD) strategy. First, we use the pre-trained DL architecture, Inception-v3, as a teacher model to distill the dataset. Then, a student model, also using the Inception-v3 model, is trained on the distilled dataset to match the performance of the teacher model. In addition, a new combination of Kullback–Leibler (KL) divergence and categorical cross-entropy (CCE) loss is used to measure the difference between the teacher and student models. This combined metric encourages the student model to mimic the predictions of the teacher model. Finally, the trained student model is evaluated on a validation dataset to assess its performance and compare it with both the teacher model and another competitive DL model. Experiments are conducted on the two datasets, corresponding to an imbalanced and a balanced dataset. Two baseline models (Inception-v3 and YOLOv8) are evaluated for reference, obtaining a maximum training accuracy of 66.75% and 90.90%, respectively, and a maximum validation accuracy of 35.94% and 81.52%, both for the imbalanced dataset. On the other hand, the proposed DR classification model achieves an average training accuracy of 99.01% and an average validation accuracy of 97.30%, overcoming the baseline models and other state-of-the-art works. Experimental results show that the proposed model achieves competitive results in DR lesion detection and classification tasks, assisting in the early diagnosis of diabetic retinopathy.

Machine learning and fluorosensing for estimation of maize nitrogen status at early growth-stages

Potential of mobile fluorescence sensor measurements have been in focus for quantifying plant nitrogen (N) variability early in the crop growing season. Real time estimation of such N status indicators at field scale would enable precision management of N fertilizers. In standard practice, linear regression analysis involves the use of several fluorescence channels and indices as predictive variables for estimating plant nitrogen content. Considering the multi-collinearity between these predictor variables, the conventional regression analysis (multiple linear regression) often fails to deliver a good range of prediction accuracies. Hence, machine learning regression techniques are utilized in this study to estimate N status indicators i.e., %N, above ground biomass, and N uptake at V6 and V9 growth stages of maize across three site-years in 2012 and 2013 crop growing seasons. The Multiplex®3 (FORCE-A) portable active fluorescence system was used to capture fluorescence information. Derived indices including four N balance indices (NBI_R, NBI_B, NBI_B, and NBI1), two chlorophyll indices (CHL and CHL1), and one flavonoid index (FLAV) were used as predictors. The independent site data were first utilized in a Support Vector Regression (SVR) model to assess the training and test accuracies in estimation of N status indicators considering a comparative analysis between V6 and V9 growth stages. The current research also involved assessing how well the machine learning-trained model could be applied to a different dataset and validated its performance in a cross-site experimental setting. Subsequently, cross-site comparisons of nitrogen status estimates were conducted to recommend the selection of machine learning strategies. These strategies include (1) Partial Least Square Regression, (2) Support Vector Regression, (3) Gaussian Process Regression, (4) Random Forest Regression, and (5) Artificial Neural Network. The comparative investigation demonstrated promising accuracy in estimating plant nitrogen content, above-ground biomass, and nitrogen uptake at the V6 stages of maize, with correlation coefficients in the moderate range (r = 0.72 ± 0.03) and Root Mean Square Error. Superior prediction accuracies were obtained at V9 growth stages than at V6. Among the various machine learning models assessed, Support Vector Regression consistently outperformed the others across the three site-year datasets, delivering error estimates well within acceptable ranges and demonstrating the lowest RMSE values for maize nitrogen indicators.

Machine learning models for differential diagnosing HER2-low breast cancer: A radiomics approach.

To develop machine learning models based on preoperative dynamic enhanced magnetic resonance imaging (DCE-MRI) radiomics and to explore their potential prognostic value in the differential diagnosis of human epidermal growth factor receptor 2 (HER2)-low from HER2-positive breast cancer (BC). A total of 233 patients with pathologically confirmed invasive breast cancer admitted to our hospital between January 2018 and December 2022 were included in this retrospective analysis. Of these, 103 cases were diagnosed as HER2-positive and 130 cases were HER2 low-expression BC. The Synthetic Minority Oversampling Technique is employed to address the class imbalance problem. Patients were randomly split into a training set (163 cases) and a validation set (70 cases) in a 7:3 ratio. Radiomics features from DCE-MRI second-phase imaging were extracted. Z-score normalization was used to standardize the radiomics features, and Pearson's correlation coefficient and recursive feature elimination were used to explore the significant features. Prediction models were constructed using 6 machine learning algorithms: logistic regression, random forest, support vector machine, AdaBoost, decision tree, and auto-encoder. Receiver operating characteristic curves were constructed, and predictive models were evaluated according to the area under the curve (AUC), accuracy, sensitivity, and specificity. In the training set, the AUC, accuracy, sensitivity, and specificity of all models were 1.000. However, in the validation set, the auto-encoder model's AUC, accuracy, sensitivity, and specificity were 0.994, 0.976, 0.972, and 0.978, respectively. The remaining models' AUC, accuracy, sensitivity, and specificity were 1.000. The DeLong test showed no statistically significant differences between the machine learning models in the training and validation sets (Z = 0, P = 1). Our study investigated the feasibility of using DCE-MRI-based radiomics features to predict HER2-low BC. Certain radiomics features showed associations with HER2-low BC and may have predictive value. Machine learning prediction models developed using these radiomics features could be beneficial for distinguishing between HER2-low and HER2-positive BC. These noninvasive preoperative models have the potential to assist in clinical decision-making for HER2-low breast cancer, thereby advancing personalized clinical precision.

A precise grape yield prediction method based on a modified DCNN model

This paper presents an innovative method for predicting grape yield using a modified deep convolutional neural network (M−DCNN). This approach integrates advanced image processing techniques with deep learning algorithms to analyze high-resolution images of vineyards, enabling precise yield quantification. The M−DCNN model incorporates novel convolutional layers and activation functions to enhance accuracy and generalization capabilities. Extensive experiments were conducted using a varied dataset of grape varieties and vineyard conditions. The method involves image processing to identify grape clusters and monitor their growth at different stages. We evaluated the performance of several algorithms, including K-nearest neighbors (KNN), multinomial logistic regression (MLR), support vector machines (SVM), random forests (RF), deep convolutional neural network (DCNN), and M−DCNN, throughout the training and testing phases. The M−DCNN achieved the highest training accuracy at 97.35%, which represents a 0.85% improvement over the standard DCNN. Furthermore, in detecting grape berries, our method attained an average F1 score of 73.32% and coefficients of determination exceeding 0.98, showcasing high accuracy and precision. The results indicate that the M−DCNN model surpasses traditional yield estimation methods and existing machine learning models, affirming its potential to significantly enhance precision agriculture. This method offers a robust tool for accurate grape yield forecasting, facilitating informed decision-making in viticulture.