Classification Ability Research Articles

MCMP-Net: MLP combining max pooling network for sEMG gesture recognition

sEMG Gesture Recognition is often used in wearable devices, so balancing model complexity and classification ability are important when designing models. Presently, methods employing large complex deep learning models can achieve high accuracy, but they are too computationally intensive to be deployed on wearable devices. In this study, we propose an Multilayer Perceptron (MLP) combining max pooling network (MCMP-Net), this model is lightweight and has a strong classification ability. Data in each frame is projected respectively into a latent space via MLP, and then max pooled as features to do classification, we also designed a multi-head mechanism to improve classification ability and a completion embedding mechanism to add sequential information. We evaluate the model using seven benchmark datasets: NinaPro DB1, DB2, DB4, DB5, and CapgMyo DB-a, DB-b, and DB-c, the accuracies are 91.8 %, 84.5 %, 85.5 %, 93.1 %, 90.5 %, 90.3 %, 94.5 %, respectively. Compared to state-of-the-art methods, the results show that our model is competitive in accuracy, furthermore, it is much computationally-efficient and easier to arrange in wearable devices.

A quality detection method of the unbalanced data based on the non‐parameter Log–Log prediction model with the feature extraction

In quality detection, it is important to classify and predict the unbalanced data sets with a high proportion of qualified and unqualified products. There exist already some machine learning methods available. However, these existing methods assume that the samples are evenly distributed among the different classes and ignore the unbalanced characteristics of data. In addition, existing methods cannot be directly applied to high‐dimensional data and cannot accurately express the relationship between data features and the quality of industrial engineering products. In this paper, we propose a new quality detection method of the unbalanced data by establishing a non‐parameter Log–Log classification model. The principal component analysis (PCA) is used to extract the features and reduce the dimension of the original data sets. We develop a sieve maximum likelihood algorithm to obtain the non‐parameter function classifier. The proposed method is applied to the product quality detection of industrial semiconductor manufacturing. The results show the proposed method has high detection performance and classification ability. Compared with traditional machine learning methods, the proposed method has a higher classification accuracy can better describe the relationship between product characteristics and product quality and has a strong generalization ability for different data sets.

DFS-WR: A novel dual feature selection and weighting representation framework for classification

Classification methods relying on multidimensional features have been widely applied in commerce, healthcare, and transportation. Nevertheless, haphazardly selecting inappropriate features may not only hinder the improvement of classification performance but also disrupt the convergence of classifier parameters, leading to prolonged classifier runtime and an increased storage burden. Therefore, this paper proposed a novel framework (dubbed as DFS-WR) to perform dual selection and weighted representation of multidimensional features. Firstly, DFS-WR employs Laplacian scores to conduct a de-redundant operation as the first feature selection. Subsequently, DFS-WR utilizes a correlation coefficient matrix and redundant sorting to eliminate duplicate deployment of similar features in classification through the second selection. Finally, for hierarchical expression, DFS-WR assigns weights by virtue of information gain (IG), thereby emphasizing the significance of dominant features. The performance of DFS-WR was further evaluated across four aspects: classification ability, runtime performance, storage occupancy, and portability and universality, using five public datasets and eleven widely adopted classifiers. Additionally, this paper considered the proper configuration of crucial parameters within DFS-WR and compared it with other related works. The results indicated that the dual selection strategy for eliminating redundancy and similarities among features could improve the performance of classifiers, and the weighted process based on IG outperformed unweighted and ReliefF-based weighted methods. By employing DFS-WR to multidimensional features, the classifier’s runtime and data storage capacity were effectively reduced. In addition, it was proved that the deployment of DFS-WR generally had no limitations on datasets and classifiers, underscoring its good portability and universality in practical applications.

Kidney CT Scan Image Classification Using Modified Vision Transformer

With the rising number of kidney-related health issues, early and precise diagnosis is crucial. The study aims to create a reliable method for categorizing kidney CT scan images into four groups: Cyst, Normal, Tumor, and stone. Traditional approaches usually rely on typical Machine Learning (ML) and Convolution Neural Networks (CNNs). However, in this research, the potential of a novel model called Vision Transformer (ViT) is explored. ViT was initially designed for Natural Language Processing (NLP) tasks but shows promise for medical image classification. ViT’s capabilities are enhanced by coupling it with Fully Connected Networks (FCN). This combination helps to merge the feature extraction capability of the ViT and the classification ability of the FCN, which ultimately helps to overcome the challenge of detecting kidney-related issues.

Analysis of core attribute and approximate reduct based on the three-way decision

Attribute reduction plays an important role in pattern recognition and machine learning, and the theory of rough sets has become a commonly used model for attribute reduction since its superiority in describing and quantifying vagueness and uncertainty. However, little attention has been paid to analyzing the importance of core attributes in different circumstances, and core attributes are generally all selected into the reduct without considering their differences. In this study, the role of core attributes is analyzed in terms of their impacts on classification ability, and the three-way partition of attributes is proposed to distinguish different core attributes and condition attributes. Then, a unified approximate attribute reduction framework based on the three-way decision is introduced to keep the quality and resulting classification performance of the reduct. Moreover, a general forward-adding and back-deleting heuristic algorithm is developed to effectively select important attributes and also eliminate unimportant core and condition attributes in the boundary region. Comprehensive comparative experiments and statistical significance analysis are conducted on UCI data sets. The experimental results show that our method achieves a better attribute reduction rate and classification performance and also verify that core attributes are not always indispensable.

Machine Learning Approach to Identify Dysregulation in the Plasma Proteome of CADASIL

AbstractBackgroundThe plasma proteome provides information to identify dysregulated molecular pathways underlying disease. Cerebrovascular diseases often manifest a significantly different plasma proteomic signature. Cerebral‐Autosomal‐Dominant‐Arteriopathy‐Subcortical‐Infarcts‐and‐Leukoencephalopathy (CADASIL) is one such vascular brain disease. The CADASIL plasma proteome has yet to be investigated.MethodTo investigate the plasma proteome in the context of CADASIL, we used large‐scale proteomics, measuring over 7,000 proteins via aptamer‐based technology (SomaLogic) from 53 study participants (nCADASIL = 25; nControl = 28). We employed machine learning (ML) methods as an unbiased approach to uncover disease‐associated change in proteomic networks. This approach is based on the premise that protein sets that best classify disease state could be important biological drivers of disease. We developed a novel ML method: coupling recursive feature extraction with logistic regression (LR) and XGBoost evaluators, as well maximum‐relevance‐minimum‐redundancy with Random‐Forest and F‐Statistic evaluators. The results of all four models were selectively aggregated in order to delineate the CADASIL plasma proteome signature while minimizing overfitting to the low sample size data. We developed a 45 protein model of the CADASIL proteomic signature from these results. To evaluate the classification ability, we tested the Repeated‐Stratified‐10‐fold‐Cross‐Validation accuracy of a LR classifier. Next, we investigated whether the CADASIL proteomic signature was relevant for other neurodegenerative diseases with vascular components; the classifier was applied to an Alzheimer’s Disease (AD) dataset. Finally, the functional signature of the 45 proteins was investigated through gene ontology (GO).ResultThe LR classifier in CADASIL is 99.8% accurate. In AD, the accuracy is 71.2%. This implies the CADASIL signature shares commonalities as well as pertinent differences with other neurodegenerative diseases. The GO of the 45 proteins resulted in GO terms corresponding to extracellular matrix and collagen, cellular membrane lipoproteins and glycolipids, and proteasome biological pathways.ConclusionThe proposed method allows for an initial unbiased discovery of important proteomic changes associated with disease state. The identified proteins would provide starting points for mechanistic studies. In addition, a panel of proteins could be used for screening at population level for risk stratification and more in‐depth phenotyping. Next steps include investigating the biological relevance of findings in mechanistic models for therapeutic developments.

Machine Learning Approach to Identify Dysregulation in the Plasma Proteome of CADASIL

AbstractBackgroundThe plasma proteome provides information to identify dysregulated molecular pathways underlying disease. Cerebrovascular diseases often manifest a significantly different plasma proteomic signature. Cerebral‐Autosomal‐Dominant‐Arteriopathy‐Subcortical‐Infarcts‐and‐Leukoencephalopathy (CADASIL) is one such vascular brain disease. The CADASIL plasma proteome has yet to be investigated.MethodTo investigate the plasma proteome in the context of CADASIL, we used large‐scale proteomics, measuring over 7,000 proteins via aptamer‐based technology (SomaLogic) from 53 study participants (nCADASIL = 25; nControl = 28).We employed machine learning (ML) methods as an unbiased approach to uncover disease‐associated change in proteomic networks. This approach is based on the premise that protein sets that best classify disease state could be important biological drivers of disease. We developed a novel ML method: coupling recursive feature extraction with logistic regression (LR) and XGBoost evaluators, as well maximum‐relevance‐minimum‐redundancy with Random‐Forest and F‐Statistic evaluators.The results of all four models were selectively aggregated in order to delineate the CADASIL plasma proteome signature while minimizing overfitting to the low sample size data. We developed a 45 protein model of the CADASIL proteomic signature from these results. To evaluate the classification ability, we tested the Repeated‐Stratified‐10‐fold‐Cross‐Validation accuracy of a LR classifier. Next, we investigated whether the CADASIL proteomic signature was relevant for other neurodegenerative diseases with vascular components; the classifier was applied to an Alzheimer’s Disease (AD) dataset. Finally, the functional signature of the 45 proteins was investigated through gene ontology (GO).ResultsThe LR classifier in CADASIL is 99.8% accurate. In AD, the accuracy is 71.2%. This implies the CADASIL signature shares commonalities as well as pertinent differences with other neurodegenerative diseases.The GO of the 45 proteins resulted in GO terms corresponding to extracellular matrix and collagen, cellular membrane lipoproteins and glycolipids, and proteasome biological pathways.ConclusionThe proposed method allows for an initial unbiased discovery of important proteomic changes associated with disease state. The identified proteins would provide starting points for mechanistic studies. In addition, a panel of proteins could be used for screening at population level for risk stratification and more in‐depth phenotyping. Next steps include investigating the biological relevance of findings in mechanistic models for therapeutic developments.

Stable and Fast Deep Mutual Information Maximization Based on Wasserstein Distance.

Deep learning is one of the most exciting and promising techniques in the field of artificial intelligence (AI), which drives AI applications to be more intelligent and comprehensive. However, existing deep learning techniques usually require a large amount of expensive labeled data, which limit the application and development of deep learning techniques, and thus it is imperative to study unsupervised machine learning. The learning of deep representations by mutual information estimation and maximization (Deep InfoMax or DIM) method has achieved unprecedented results in the field of unsupervised learning. However, in the DIM method, to restrict the encoder to learn more normalized feature representations, an adversarial network learning method is used to make the encoder output consistent with a priori positively distributed data. As we know, the model training of the adversarial network learning method is difficult to converge, because there is a logarithmic function in the loss function of the cross-entropy measure, and the gradient of the model parameters is susceptible to the "gradient explosion" or "gradient disappearance" phenomena, which makes the training of the DIM method extremely unstable. In this regard, we propose a Wasserstein distance-based DIM method to solve the stability problem of model training, and our method is called the WDIM. Subsequently, the training stability of the WDIM method and the classification ability of unsupervised learning are verified on the CIFAR10, CIFAR100, and STL10 datasets. The experiments show that our proposed WDIM method is more stable to parameter updates, has faster model convergence, and at the same time, has almost the same accuracy as the DIM method on the classification task of unsupervised learning. Finally, we also propose a reflection of future research for the WDIM method, aiming to provide a research idea and direction for solving the image classification task with unsupervised learning.

Estimation of Japanese word intelligibility by automatic speech recognition with noise adaptation

The purpose of this study is to estimate subjective word intelligibility using DNN-based speech recognition. The evaluation of intelligibility is important to guarantee high speech quality in ambient noise and reverberation. The Japanese Diagnostic Rhyme Test (JDRT) is one such subjective evaluation method. In this study, we estimated Japanese word intelligibility by simulating the JDRT with cloud-based speech recognition systems. Recent DNNs can improve recognition accuracy with their robust classification ability and efficient use of high-dimensional features. We built a DNN-based speech recognition system using cloud-based ASR APIs, and adaptively trained its model to several noise conditions to improve its intelligibility estimation accuracy. In this paper, we will report comparisons with other objective and subjective evaluations and discuss ideas for a more generic objective evaluation method.

Proposed Two-Steps Procedure of Classification High Dimensional Data with Regularized Logistic Regression

The field of Bioinformatics has developed in response to the rapid increase in biological data, particularly high-dimensional gene expression data. Bioinformatics utilizes optimization, computational science, and statistical methods to effectively address challenges in the field of molecular biology. Numerous genes (variables) in gene expression are irrelevant to their study. Gene selection has been demonstrated to be an effective means of enhancing the performance of numerous methods of classification. The job of acquiring significant variables via the use of ranking variable selection (RVS) techniques and then picking the most effective classifier is an enormous challenge in the context of high-dimensional data. in this study, we proposed a new ranking filter method using smooth clipped absolute deviation depending on the resampling technique (RSVS) to obtain a proficient subset of genes with strong classification abilities. This is achieved by merging A screening technique employed as a filtering method in conjunction with Regularized Logistic Regression, such as LASSO,ALASSO,ENET, and MCP. The study involved the utilization of both simulated and real datasets to conduct an empirical evaluation of the proposed approach. The findings indicated that the proposed method outperformed other established methods. it was tested using three publicly data sets about Cancer. The Results demonstrate that the suggested approach is highly effective and viable, thus showing a strong level of performance with regards to accuracy, geometric mean, and the area under the curve. Furthermore, The findings suggest that the genes most often chosen are physiologically associated with the specific form of cancer. Therefore, the method that has been suggested has potential advantages for the classification of cancer via the use of DNA gene expression data within a clinical setting.

Construction of Transformer Fault Diagnosis and Prediction Model Based on Deep Learning

The current intelligent diagnosis and prediction methods for transformer faults are prone to low diagnostic accuracy and insufficient trend prediction ability when the fault categories are imbalanced. Therefore, a fault diagnosis and prediction model for transformers was constructed using a deep learning framework. The fault diagnosis model was constructed using a focus loss stack sparse noise reduction autoencoder on the deep learning framework. The prediction model was constructed by fusing long and short term memory networks on the basis of tree structure Parzen optimization, and the two models were validated. The results obtained through validation of the diagnostic model indicate that, when the actual hidden layer is set to 3 and the quantity of neurons is 58, the model accuracy during training and testing reaches 97.5% and 92.5% respectively. After adding 0.001 times the Gaussian white noise, the model accuracy was significantly lifted, so this study set the Gaussian noise coefficient to 0.001. In the comparison with baseline models, the actual classification ability of the research model samples is strong, significantly improving the fault diagnosis ability. In the validation of the prediction model, the three error index values of the research model in the single prediction step of CH4 concentration were 0.0699, 0.0540, and 0.8481%, respectively, and proved to be were lower than in the case of the baseline model. The three error values in the two-step prediction are 0.0194, 0.0161, and 0.6535%, which are also lower than in case of the baseline model. Overall, the diagnosis and prediction model proposed in this paper can provide real-time future numerical predictions of dissolved gas analysis and monitoring data in transformer oil. Furthermore, the research outilnes the future development trend of monitoring and measurement through application of tensor flow deep learning framework in transformer fault diagnosis. The attained prediction results are innovative, and could well complete the purpose of actual transformer fault diagnosis and early warning.

Diagnosing injection-production system faults in the same well using the rough set-LVQ neural network.

This study proposed a reverse calculation model of the unique rod pump injection and production system structures in the same well to diagnose and resolve defects, after which dynamometer diagrams of the system production and injection pumps were drawn. The invariant moment feature method was applied to identify seven such characteristics in the injection pump power graph, establishing a downhole system for fault diagnosis in rod pump injection and production systems in the same well using Rough Set(RS)-Learning Vector Quantization(LVQ). On the premise of keeping the classification ability unchanged, the Self-Organizing Map(SOM) neural network was used to discretize the original feature data, while RS theory was employed for attribute reduction. After establishing the LVQ fault diagnosis subsystem, the reduced decision table was entered for learning and training. The test results confirmed the efficacy and accuracy of this method in diagnosing downhole faults in rod pump injection-production systems in the same well. After comparing the test results with the actual working conditions, it can be seen that the rod pump injection-production diagnosis system based on RS-LVQ designed in this paper has a recognition rate of 91.3% for fault types, strong recognition ability, short diagnosis time, and A certain practicality. However, the research object of fault diagnosis in this paper is a single fault, and the actual downhole fault situation is complex, and there may be two or more fault types at the same time, which has certain limitations.

Performance evaluation of backscattering coefficients and polarimetric decomposition parameters for marsh vegetation mapping using multi-sensor and multi-frequency SAR images

Wetland vegetation is the basis for wetland ecosystems to regulate climate change, carbon sequestration and maintain biodiversity. Therefore, high-precision mapping and dynamic monitoring of wetland vegetation are essential for the effective management, restoration and sustainable development of wetland ecosystems. This study addressed to explore classification performance of backscattering coefficients and polarimetric decomposition parameters of multi-frequency SAR images, including single-polarimetric X-band TerraSAR (TS), full-polarimetric C-band Radarsat-2 (RS) and L-band ALOS PALSAR-2 (PS) for marsh vegetation mapping in Honghe National Nature Reserve, Northeast China. We proposed two transfer-learning strategies, and examined the feasibility of transfer-learning vegetation classifications between optical and SAR sensors, different frequencies SAR (X-, C- and L-band) and its derivative images, respectively. This paper further compared transfer-learning classification performance of marsh vegetation from polarimetric decomposition parameter images to backscattering coefficient images under the same SAR sensors. The results indicated that: (1) The three SAR images performed good classification ability in identifying marsh vegetation with the overall accuracies (OA) ranging from 0.74 to 0.88. For the same frequency SAR images, polarimetric decomposition parameters outperformed backscattering coefficients with an OA improvement ranging from 0.24 % to 2.41 %; (2) The longer wavelength SAR images produced better classification results, and L-band PS images realized the highest classification accuracy (OA = 0.871); (3) Full-polarimetric SAR images obtained the better transfer-learning classifications (OA > 0.8), and the transfer-learning classification ability of the same frequency SAR images outperformed the different frequencies; (4) The transfer learning classification results between SAR sensors outperformed that of between optical and SAR images. The results of this study provide a scientific basis for wetland change monitoring conservation and sustainable development.

Data-independent acquisition for proteomic applications in early-stage endometrial cancer progression.

Most endometrial cancer (EC) patients are diagnosed at an early-stage (FIGO stage I or II), with a favorable prognosis. However, some high-grade, early-stage EC patients have unexpected recurrences and undesirable results, the molecular alterations that underlie these tumors are far from being fully understood. Our goal was to use proteome analysis to examine dysregulated pathways in this specific subgroup of EC. We used data-independent acquisition (DIA) quantitative proteomics to analyze cancer and matched paracancerous tissues from 20 EC patients (10 high-grade and 10 low-grade). Immunohistochemistry (IHC) analysis was used to validate protein expression of six hub genes. In total, 7107 proteins were quantified, while 225 downregulated and 366 upregulated proteins in high-grade cancer tissues, 130 downregulated and 413 upregulated proteins in high-grade paracancerous tissues. The pathway associated with oxidative phosphorylation (OXPHOS) was upregulated and have similar expression patterns in high-grade EC tissues and matched paracancerous tissues. OXPHOS-related protein, ATP5F1D showed the best classification and diagnostic ability in distinguishing high-grade from low-grade EC. In both cancer and paracancerous tissues, double-label immunofluorescence demonstrated ITGA4 and COL4A1 co-localized at the basal membrane. Our present works elucidate that metabolism reprogramming is associated with high-grade, early-stage EC, particularly OXPHOS is upregulated. Noticeably, the paracancerous tissues have undergone molecular changes similar to cancer tissues, maybe they have signal exchange via secreted proteins (ITGA4 and COL4A1). The upregulation of OXPHOS-related proteins may be the potential biomarker for EC diagnosis, and targeting OXPHOS metabolism might be an effective treatment for high-grade, early-stage EC.

Knowledge distillation for object detection based on Inconsistency-based Feature Imitation and Global Relation Imitation

Knowledge distillation (KD) is a method that transfers information from a larger network (teacher) to a smaller network (student) to obtain stronger performance without extra computational load. It has made great success in image classification, but only attains trivial improvement in object detection. In this paper, we propose a knowledge distillation scheme for object detection with Inconsistency-Based Feature Imitation (IBFI) and Global Relation Imitation (GRI) schemes. IBFI calculates the difference between the outputs of classification head and regression head to balance the classification and localization abilities of the detector. GRI enables the student to mimic the teacher’s relational information. Extensive experiments have been conducted on popular datasets including MS COCO and PASCAL VOC to validate the effectiveness of our scheme. We achieve 39.9% mAP on MS COCO by using RetinaNet + ResNet-50, which surpasses the baseline by 2.5%.

An improved multi-scale gradient generative adversarial network for enhancing classification of colorectal cancer histological images.

Deep learning-based solutions for histological image classification have gained attention in recent years due to their potential for objective evaluation of histological images. However, these methods often require a large number of expert annotations, which are both time-consuming and labor-intensive to obtain. Several scholars have proposed generative models to augment labeled data, but these often result in label uncertainty due to incomplete learning of the data distribution. To alleviate these issues, a method called InceptionV3-SMSG-GAN has been proposed to enhance classification performance by generating high-quality images. Specifically, images synthesized by Multi-Scale Gradients Generative Adversarial Network (MSG-GAN) are selectively added to the training set through a selection mechanism utilizing a trained model to choose generated images with higher class probabilities. The selection mechanism filters the synthetic images that contain ambiguous category information, thus alleviating label uncertainty. Experimental results show that compared with the baseline method which uses InceptionV3, the proposed method can significantly improve the performance of pathological image classification from 86.87% to 89.54% for overall accuracy. Additionally, the quality of generated images is evaluated quantitatively using various commonly used evaluation metrics. The proposed InceptionV3-SMSG-GAN method exhibited good classification ability, where histological image could be divided into nine categories. Future work could focus on further refining the image generation and selection processes to optimize classification performance.

The Effect of Analyzing Images Using The PlantNet Application and TGT on Motivation, Classification Abilities, and Data Presentation in Plantae Material

Learners need to observe various types of plants to achieve basic competencies in the subject of "Plantae." However, teachers find it challenging to manage learning activities outside the school environment. This research aims to: 1) analyze the influence of three different teaching methods on motivation, classification abilities, and data presentation, and 2) determine the most effective teaching method among the three.This study adopts a quantitative approach, specifically quasi-experimental research, using a non-equivalent control group design. The research population comprises all 10th-grade students in the Science Program (MIPA) at MAN 2 Banyumas, totaling six classes. The sample consists of three classes: X MIPA 2, which uses PlantNet + TGT, X MIPA 3, which uses TGT, and X MIPA 4, which employs discussion. Data collection methods include questionnaires, tests, and observation sheets. Data analysis techniques encompass statistical methods such as normality and homogeneity tests, analysis of variance (ANOVA), and post hoc tests.Based on the ANOVA analysis results for all three classes, a significance value of 0.00<0.05 was obtained. This indicates the impact of the techniques of analyzing images using the PlantNet application and TGT, TGT-based learning, and discussion-based learning on motivation, classification abilities, and data presentation. The least significant difference analysis shows that the PlantNet + TGT class had an average motivation score of 48.20, the TGT class had 36.69, and the discussion class had 28.64. The average increase in classification abilities was 35.25 for the PlantNet + TGT class, 21.18 for the TGT class, and 15.58 for the discussion class. In terms of data presentation abilities, the PlantNet + TGT class had an average increase of 25.02, the TGT class had 17.15, and the discussion class had 11.00. This indicates that the PlantNet + TGT class showed significantly higher improvement compared to the TGT and discussion classes.In conclusion, the study results indicate that all three teaching strategies, including the technique of analyzing images with the PlantNet application and TGT, TGT, and discussion, have an impact on motivation, classification abilities, and data presentation. Among these methods, teaching with the technique of analyzing images with the PlantNet application and TGT resulted in the highest improvement in motivation, classification abilities, and data presentation.

Identification of dialects and individuals of globally threatened yellow cardinals using neural networks

Audio-based analysis of bird songs has proven to be a valuable practice for the growth of knowledge in the fields of ethology and ecology. In recent years, machine learning techniques applied to audio field recordings of bird calls have yielded successful results in studying population distributions and identification of individuals for their monitoring in a variety of bird species. This offers promising possibilities in the study of social behavior, biodiversity, and conservation strategies for birds. In this work, we trained deep learning models, directly from the sonograms of audio field recordings, to investigate the statistical properties of vocalizations in an endangered bird species, the Yellow Cardinal, Gubernatrix cristata. This research marks the first successful application of this method to an endangered species. Our results indicate the presence of vocal signatures that reflect similarities in songs of individuals that inhabit the same region, determining dialects, but which also show differences between individuals. These differences can be exploited by a deep learning classifier to discriminate the bird identities through their songs. Models trained with data labeled by regions showed a good performance in the recognition of dialects with a mean accuracy of 0.84 ± 0.04, significantly higher than the accuracy obtained by chance. Precision and recall values also reflected the classifier's ability to find alike vocal patterns in the songs of neighboring individuals. Models trained with data labeled by individuals showed an accuracy of 0.63 ± 0.03, significantly higher than that obtained by chance. However, the individual discrimination model showed greater confusion with neighboring individuals. This reflects a hierarchical structure in the characteristics of the Yellow Cardinal's vocalization, where the intra-individual variability is lower than the inter-individual variability, but it is even lower than the variability obtained when individuals inhabit different regions, providing evidence of the existence of dialects. This reinforces the results of previous works but also offers an automated method for characterizing cultural units within the species. Along with genetic data, this method could help better define management units, thereby benefiting the success of reintroduction of individuals of Yellow Cardinal recovered from the illegal trade. Moreover, the novelty of individual discrimination using neural networks for the Yellow Cardinal, which has limited data availability, shows promise for non-invasive acoustic monitoring strategies with potentially relevant implications for its conservation.

Risk predicting for acute coronary syndrome with conventional and unstable plaque radiomics features from serial coronary computed tomography angiography

Abstract Background Plaques with high-risk features are responsible for most acute cardiac events. However, there are still some lesions with high-risk plaque (HRP) morphology, which will not lead to cardiac events over time. It is crucial to precisely identify features of unstable plaque and to carry out individualized treatment, eventually aiming to reduce the likelihood of plaque rupture and lower the incidence of acute coronary events. Besides, the development of quantitative radiomics in recent years may further help with the diagnostic and predictive performance of coronary CTA. Purpose To establish machine learning models by combining traditional coronary CTA features of high-risk plaque and radiomics features of unstable plaque, so as to improve the efficiency of models and to predict acute cardiac events precisely. Methods A total of 133 elderly patients (407 lesions) were retrospectively enrolled in this cohort study who underwent serial CCTA with the interval ≥ 2 years between scans. The patients were divided into two groups according to the occurrence of acute coronary events within 1 year after coronary CTA scanning . We recorded anatomical characteristics, high-risk plaque characteristics, and acute coronary events of the elderly patients. Based on the coronary CTA image cohort, we applied 3D Slicer software to segment target lesions, delineate regions of interest, and extract radiomics features. After that, we used LASSO regression to reduce the dimensionality of the radiomics features, and finally screened out unstable radiomics features. Combining the HRP features with the selected radiomics features, we established a prediction model of acute cardiac events with machine learning method. The area under the receiver operating characteristic (ROC) curve (AUC) was analyzed to compare the classification ability of conventional and radiomics models in the validation set. Results 133 elderly patients were included, with an average age of 65.8 ± 5.3 years. A novel model has been established by combining HRP features and 4 selected unstable radiomics features (2 wavelet features, 1 shape feature and 1 GrayLevel feature) with SVM machine learning method. Compared with conventional HRP model, the novel model displayed a significantly improved predictive performance in the validation set (AUC 0.784 vs. 0.624, P < 0.001). Conclusions We have established a novel model by combining the traditional HRP features of coronary CTA and radiomics features of unstable plaque, assisting in better prediction of acute coronary events in elderly . Compared with traditional HRP feature model, the prediction efficiency of novel model was significantly improved.

An Overview of Inflammatory Spondylitis for Biomedical Imaging Using Deep Neural Networks

Ankylosing Spondylitis (AS) is an axial spine inflammatory illness and also chronic that might present with a range of clinical symptoms and indicators. The illness is most frequently characterized by increasing spinal stiffness and persistent back discomfort. The affect of the sacroiliac joints, spine, peripheral joints, entheses and digits are the main cause of the illness. AS symptoms include reduced spinal mobility, aberrant posture, hip and dactylitis, enthesitis, peripheral arthritis, and buttock pain. With their exceptional picture classification ability, the diagnosis of AS illness has been transformed by deep learning techniques in artificial intelligence (AI). Despite the excellent results, these processes are still being widely used in clinical practice at a moderate rate. Due to security and health concerns, medical imaging applications utilizing deep learning must be viewed with caution. False instances, whether good or negative, have far-reaching effects on the well-being of patients and these are to be considered. These are extracted from the fact of the state-of-the-art of deep learning (DL) algorithms lack internal workings comprehension and have complicated interconnected structure, huge millions of parameters, and also a "black box" aspect compared to conventional machine learning (ML) algorithms. XAI (Explainable AI) approaches make it easier to comprehend model predictions, which promotes system reliability, speeds up the diagnosis of the AS disease, and complies with legal requirements.