Classification Ability Research Articles

Forest canopy stratification based on fused, imbalanced and collinear LiDAR and Sentinel-2 metrics

Knowledge about the forest canopy stratification is of essential importance for forest management and planning. Collecting structural information (e.g. natural regeneration) still depends on cost and labour intensive forest inventories with a coarse spatio-temporal resolution. Remote sensing partly overcomes these limitations and particularly active sensors of type light detection and ranging (LiDAR) have proven their great potential of separating forest strata. The applicability of LiDAR metrics for the differentiation of the spruce dominated forest strata in Central Germany has not been tested yet. Additionally, studying the potential of Sentinel-2 metrics for the classification of forest strata is lacking too.In this study, we investigated the capabilities of six different classification approaches for the differentiation of five forest strata that are typical for the study region. Reference data were derived from forest inventory measurements surveyed on a dense 200 × 200 m grid. The six classification approaches were trained with fused and un-fused LiDAR and Sentinel-2 inferred metrics. The classification results were compared using the overall mean accuracy, sensitivity and specificity via receivers operating characteristics of multi-class problems. We were interested in the classification abilities of Sentinel-2 metrics due to the obvious advantages of Sentinel-2 based metrics (free of charge, high spatio-temporal coverage). We assumed that the canopy structure determines the reflection on stand level and thus might facilitate the classification of different canopy strata. Beforehand, it was important to examine the influence of distinctly imbalanced and collinear reference data on the classification results.We found that the Random Forest classifier most accurately separated the five forest strata with a mean overall accuracy of 83.3% (Kappa = 76.2%). These values were achieved from balanced training data and the classification capability was confirmed by classification results from an independent test data set. Fused predictors of active (LiDAR) and passive (Sentinel-2) remote sensing revealed no substantial improvement in the classification accuracy due to the dominant role of LiDAR metrics. Herein, we identified that especially the height variability, top height, portion of LiDAR-returns between 2 m and 10 m and the standard deviation of the return number between the 25th and 50th height percentile, predominately contributed to the classification accuracy. Classification results purely based on Sentinel-2 metrics revealed a rather small overall mean accuracy of 54.7%. The metrics (e.g. median, variance, entropy) were derived from Sentinel-2 indices, covering the visible and near to short infrared spectrum. Variable importance computations unraveled a detectable but minor contribution of MSI, TCG, NDVI to the classification result. Finally, our data driven observations illustrated serious drawbacks associated to data imbalance, collinearity and autocorrelation and presented practical guidance to cope with these issues.

Nesnelerin İnternetinde Karar Ağacı Tabanlı Saldırı Tespit Yöntemi

Developments in computer and network technologies have also positively affected internet technology. With the development of the Internet, the concept of IoT (Internet of Things) has been invented. Nowadays, IoT devices provide convenience in many areas, and the positive effects of IoT-based systems increase people's quality of life. People want to remotely monitor and manage smart cities, smart homes, and other platforms. However, IoT systems have many vulnerabilities and thus have become the target of attackers. Detecting such attacks and preventing security vulnerabilities will further increase the rate of use of IoT technology. In this work, an intelligent intrusion detection system (IDS) for IoT devices has been suggested. The presented intelligent IDS for IoT devices have been developed on a big attack dataset and this dataset contains 3,668,443 observations. In prior works which used this dataset, researchers worked on a binary classification problem (attacked and normal). However, this research aims to classify the attack types, hence, nine categories have been used. In order to propose a prompt responded IDS model, a fast classifier which is a decision tree (DT) has been employed. Our proposal attained 97.43% classification accuracy on this dataset using 10-fold cross-validation. This accuracy rate frankly demonstrates the classification ability of our proposed IDS model for IoT devices.

Design of Semiautomatic Digital Creation System for Electronic Music Based on Recurrent Neural Network.

Semiautomated digital creation is increasingly important in the manipulation of electronic music. How to realize the learning of local effective features of audio data is a difficult point in the current research field. Based on recurrent neural network theory, this paper designs a semiautomatic digital creation system for electronic music for digital manipulation and genre classification. The recurrent neural network improves the transmission of electronic music information between the input and output of the network by adopting dense connections consistent with DenseNet and adopts an inception-like structure for the autonomous selection of effective recursive nuclear electronic music categories. In the simulation process, the prediction method based on semiautomatic digital audio clips is also adopted, which pays more attention to the learning of local effective features of audio data, which gives the model the ability to create audio samples of different lengths and improves the model's support for creative tasks in different scenarios. It includes the determination of the number of neurons, the selection of the function of neurons, the determination of the connection method, and the specific learning algorithm rules, and then the training samples are formed. The experimental results show that the recurrent neural network exhibits powerful feature extraction ability and classification ability of music information. The 10-fold cross-validation on GTZAN dataset and ISMIR2004 dataset has obtained 88.7% and 87.68%, surpassing similar ones. The model has reached a leading level. After further use of the MSD (Million Song Dataset) dataset for pre-semiautomatic training, the model effect has been further greatly improved. The accuracy rate on the dataset has been increased to 91.0% and 89.91%, respectively, which has improved the semiautomatic number and creative advancement.

Open surgery tool classification and hand utilization using a multi-camera system.

The goal of this work is to use multi-camera video to classify open surgery tools as well as identify which tool is held in each hand. Multi-camera systems help prevent occlusions in open surgery video data. Furthermore, combining multiple views such as a top-view camera covering the full operative field and a close-up camera focusing on hand motion and anatomy may provide a more comprehensive view of the surgical workflow. However, multi-camera data fusion poses a new challenge: A tool may be visible in one camera and not the other. Thus, we defined the global ground truth as the tools being used regardless their visibility. Therefore, tools that are out of the image should be remembered for extensive periods of time while the system responds quickly to changes visible in the video. Participants (n = 48) performed a simulated open bowel repair. A top-view and a close-up cameras were used. YOLOv5 was used for tool and hand detection. A high-frequency LSTM with a 1-second window at 30 frames per second (fps) and a low-frequency LSTM with a 40-second window at 3 fps were used for spatial, temporal, and multi-camera integration. The accuracy and F1 of the six systems were: top-view (0.88/0.88), close-up (0.81,0.83), both cameras (0.9/0.9), high-fps LSTM (0.92/0.93), low-fps LSTM (0.9/0.91), and our final architecture the multi-camera classifier(0.93/0.94). Since each camera in a multi-camera system may have a partial view of the procedure, we defined a 'global ground truth.' Defining this at the data labeling phase emphasized this requirement at the learning phase, eliminating the need for any heuristic decisions. By combining a system with a high fps and a low fps from the multiple camera array, we improved the classification abilities of the global ground truth.

A new lateral geniculate nucleus pattern-based environmental sound classification using a new large sound dataset

Background and purposeOne of the essential purposes of sound classification is to achieve similar/over classification ability of the human auditory system (HAS). A new dataset and a biologically inspired feature extraction function have been proposed to realize this aim. We have developed a highly accurate sound classification architecture using the proposed biological-inspired feature extraction function. Materials and methodsIn this research, a new environmental sound classification (ESC) dataset has been collected as a testbed, and this dataset contains 5000 sounds with 50 classes. Moreover, the collected ESC sound dataset is balanced. A new hand-modeled sound classification model has been proposed to classify sounds of this dataset. This model consists of (i) feature generation using a new lateral geniculate nucleus pattern (LGNPat), statistical moments and discrete wavelet transform (DWT), (ii) loop-based (iterative) neighborhood component analysis (INCA) based feature selection, (iii) classification using the selected features by k nearest neighbors (kNN) with 10-fold cross-validation. The proposed sound classification architecture is an extendable model. In this aspect, a new generation of hand-modeled sound/one-dimensional signal classification methods can be proposed. ResultsThe presented hand-modeled learning method was applied to the ESC dataset acquired, and our LGNPat-based model attained 93.34% classification. ConclusionsThe computed 93.34% on the collected ESC dataset has demonstrated the success of our model. Moreover, the collected dataset has been publicly published. In this aspect, the published dataset can be used to improve advanced sound classification models.

A feature selection method via relevant-redundant weight

Feature selection is a crucial preprocessing technique in data mining and machine learning and has attracted increasing attentions. However, the relevance of existing methods only contains limited information. To enhance the classification ability of filter and information theory-based feature selection methods and reduce the redundancy of the selected subset, a relevant-redundant weight-based feature criterion (FSRRW) is proposed. In this paper, a feature relevant-redundant weight (RRW) is constructed to extract the important relevant and redundant information. Then, a novel feature relevance is defined based on the weight, which contains more comprehensive information from the dynamically changing features. Additionally, a feature evaluation criterion is presented via maximizing the feature relevance and minimizing the feature redundancy. The proposed algorithm and seven compared methods are tested on 20 benchmark datasets. Extensive experiments demonstrate that the proposed criterion exhibits better feature screening abilities, effectively facilitates classification, and has preferable applicability and robustness.

Music Emotion Classification Method Based on Deep Learning and Explicit Sparse Attention Network.

In order to improve the accuracy of music emotion recognition and classification, this study combines an explicit sparse attention network with deep learning and proposes an effective emotion recognition and classification method for complex music data sets. First, the method uses fine-grained segmentation and other methods to preprocess the sample data set, so as to provide a high-quality input data sample set for the classification model. The explicit sparse attention network is introduced into the deep learning network to reduce the influence of irrelevant information on the recognition results and improve the emotion classification and recognition ability of music sample data set. The simulation experiment is based on the actual data set of the network. The experimental results show that the recognition accuracy of the proposed method is 0.71 for happy emotions and 0.688 for sad emotions. It has a good ability of music emotion recognition and classification.

Identifying out of distribution samples for skin cancer and malaria images

Deep neural networks have shown promising results in disease detection and classification using medical image data. However, they still suffer from the challenges of handling real-world scenarios especially reliably detecting out-of-distribution (OoD) samples. We propose an approach to robustly classify OoD samples in skin and malaria images without the need to access labeled OoD samples during training. Specifically, we use metric learning along with logistic regression to force the deep networks to learn much rich class representative features. To guide the learning process against the OoD examples, we generate in distribution’s similar-looking examples by either removing class-specific salient regions in the image or permuting image parts and distancing them away from in-distribution (ID) samples. During inference time, the K-reciprocal nearest neighbor is employed to detect out-of-distribution samples. For skin cancer OoD detection, we employ two standard benchmark skin cancer ISIC datasets as ID, and six different datasets with varying difficulty levels were taken as out of distribution. For malaria OoD detection, we use the BBBC041 malaria dataset as ID and five different challenging datasets as out of distribution. We achieve state-of-the-art results, improving 5% and 4% in TNR@ TPR95% over the previous state-of-the-art OoD detection methods for skin cancer and malaria images while preserving the ID classification ability.

A Comparison of Three Research Methods: Logistic Regression, Decision Tree, and Random Forest to Reveal Association of Type 2 Diabetes with Risk Factors and Classify Subjects in a Military Population

Background: Type 2 diabetes mellitus (T2DM) is one of the major non-communicable diseases, causing morbidity and mortality worldwide. There is no study on T2DM status in Iran Army Forces. Objectives: We aimed to measure the prevalence of T2DM in this population and identify variables associated with T2DM risk in order to classify individuals. Methods: Data from 3661 Iran Army Ground Forces were employed. Characteristics of the subjects with and without T2DM were compared. We examined the classification ability of logistic regression with two tree-based supervised learning algorithms, decision tree and random forest (RF). The ethical committee of AJA University of Medical Sciences approved this study by the approval code 995685. Results: The prevalence of T2DM was 3% less than in the general population. Our results showed that the incidence of T2DM increases as subjects become older. The proportions of staff members with T2DM were more than the other military ranks. T2DM is more common in obese and overweight groups. The highest prevalence of T2DM is in the subjects with high levels of lipid profile. The areas below the receiver operating characteristic curve for logistic regression, decision tree, and RF were 73.8%, 77.1%, and 97.1%, respectively. Conclusions: Age, body mass index, total cholesterol, low-density lipoprotein cholesterol, and triglyceride are associated with T2DM risk. The RF has superior classification performance in comparison with logistic regression and decision tree.

Machine Learning-Assisted FTIR Analysis of Circulating Extracellular Vesicles for Cancer Liquid Biopsy.

Extracellular vesicles (EVs) are abundantly released into the systemic circulation, where they show remarkable stability and harbor molecular constituents that provide biochemical information about their cells of origin. Due to this characteristic, EVs are attracting increasing attention as a source of circulating biomarkers for cancer liquid biopsy and personalized medicine. Despite this potential, none of the discovered biomarkers has entered the clinical practice so far, and novel approaches for the label-free characterization of EVs are highly demanded. In this regard, Fourier Transform Infrared Spectroscopy (FTIR) has great potential as it provides a quick, reproducible, and informative biomolecular fingerprint of EVs. In this pilot study, we investigated, for the first time in the literature, the capability of FTIR spectroscopy to distinguish between EVs extracted from sera of cancer patients and controls based on their mid-IR spectral response. For this purpose, EV-enriched suspensions were obtained from the serum of patients diagnosed with Hepatocellular Carcinoma (HCC) of nonviral origin and noncancer subjects. Our data point out the presence of statistically significant differences in the integrated intensities of major mid-IR absorption bands, including the carbohydrate and nucleic acids band, the protein amide I and II bands, and the lipid CH stretching band. Additionally, we used Principal Component Analysis combined with Linear Discriminant Analysis (PCA-LDA) for the automated classification of spectral data according to the shape of specific mid-IR spectral signatures. The diagnostic performances of the proposed spectral biomarkers, alone and combined, were evaluated using multivariate logistic regression followed by a Receiving Operator Curve analysis, obtaining large Areas Under the Curve (AUC = 0.91, 95% CI 0.81–1.0). Very interestingly, our analyses suggest that the discussed spectral biomarkers can outperform the classification ability of two widely used circulating HCC markers measured on the same groups of subjects, namely alpha-fetoprotein (AFP), and protein induced by the absence of vitamin K or antagonist-II (PIVKA-II).

Adaptive graph regularized and label embedded dictionary learning for pattern classification

During the last decade, graph regularized dictionary learning (DL) models have obtained a lot of attention due to their flexible and discriminative ability in nonlinear pattern classification. However, the conventional graph-regularized methods construct a fixed affinity matrix for nearby samples in high-dimensional data space, which is vulnerable to noisy and redundant sample features. Furthermore, the discrimination of the graph regularized representation is not fully explored with the supervised classifier learning framework. To remedy these limitations, we propose an adaptive graph-regularized and label-embedded DL model for pattern classification. Especially, the affinity graph construction in low-dimensional representation space and the discriminative sparse representation is simultaneously learned in a unified framework for mutual promotion. More concretely, we iteratively update the sample similarity weight matrix in representation space to enhance the model robustness and further impose a supervised label-embedding term on sparse representation to enhancing its discriminative capability for classification. The dictionary orthonormal constraint is also considered to eliminate the redundant atoms and enhance the model discrimination. An efficient alternating direction solution with guaranteed convergence is proposed for the nonconvex and unsmooth model. Experimental results on five benchmark datasets verify the effectiveness of the proposed model.

A fast neighborhood classifier based on hash bucket with application to medical diagnosis

In the medical diagnosis, expensive costs will increase significantly with the increment of medical information, and they can be reduced by data mining methods. The neighborhood classifier, as one of the extensions of the neighborhood rough set, has become an intuitive and effective classification method in data mining. However, there are still some defects which limit its performance. On the one hand, most existing neighborhood classifiers suffer from high computation complexity to obtain the neighborhood of samples; on the other hand, the difference between samples in the neighborhood is ignored when classifying samples, which diminishes the classification ability of the model. Therefore, in this paper, hash buckets and distance voting rule are introduced to solve the above problems, and a fast neighborhood classifier based on hash bucket (FNC-HB) is proposed. First, all samples are mapped to corresponding buckets through the hash function. Next, for any test sample, a bucket-based adaptive neighborhood classification radius is defined, in which the artificial parameter has been eliminated. Then, to avoid the indiscernibility of traditional voting rule when predicting labels of the test sample, a new rule called distance voting rule is defined. Finally, experimental results show that FNC-HB has better classification performance and computation efficiency, which is feasible and effective.

Minimal EEG channel selection for depression detection with connectivity features during sleep

Background and objectivesSleeping cortical electroencephalogram (EEG) has the potential for depression detection, for different sleep structure and cortical connection have been proved in depressed patients. However, the operation of multi-channel sleep EEG recording is cumbersome and requires laboratory equipment and professional sleep technician. Here, we focus on the depression detection using minimal sleep EEG channels. MethodsSixteen channels of EEG data of 30 patients with depression and 30 age-matched normal controls were recorded during sleep. Power spectral density of each single EEG channel was calculated, followed by measuring the symbolic transfer entropy (STE) and weighed phase lag index (WPLI) between EEG channel pairs in various frequency bands. Thereafter, these features were evaluated by F-score in the two-way classification (depression vs. control) of 30-s sleep EEG segments. Based on the F-score, entropy method was introduced to calculate the weight which could further assess the classification ability of various EEG channels or channel pairs. Finally, machine learning was implemented to verify the important EEG channels or channel pairs in depression diagnosis. ResultsThe features characterizing the inter-hemispheric connectivity in the posterior lobe, especially in the temporal lobe, showed high classification capacity. The classification accuracy of using two and four EEG channels in the temporal lobe were 97.96% and 99.61%, respectively. ConclusionsThis study showed the possibility of using only a few sleep EEG channels for depression screening, which may greatly facilitate the diagnosis of depression outside the hospital.

Land use/land cover and change detection mapping in Rahuri watershed area (MS), India using the google earth engine and machine learning approach

The change detection and land use and land cover (LULC) maps are more important powerful forces behind numerous ecological systems and fallow land. The current research focuses on demarcating the spatiotemporal LULC changes, NDVI and change detections maps. These effects directly affect the ecosystem, land resources, cropping pattern and agriculture. LULC assessment and surveillance are essential for long-term planning and sustainable use of natural resources. However, we have developed the soft computing machine learning algorithm for mapping land use and land cover based on the Google earth engine (GEE) platform and change detection mapping done by SAGA GIS software. It is significantly used for ecological safety and planning under various climate variations. To accurately describe the land use and land cover classes with changes are identified in the area. This area exclusively uses the multitemporal Landsat-5 (30 m) and Sentinel-2 (10 m) imageries in LULC mapping. The GEE is a cloud-computing platform with the prevailing classification ability of random forest (RF) models to make five-year interval LULC maps for 2010, 2015 and 2020. To unique multiple RF models established as a classifier in the algorithm created by JavaScript and GEE. SAGA GIS has provided the best platform for detecting changes in land use and land cover classes. NDVI maps are created based on the cloud-based platform. These maps value ranges between −0.68 to −0.15, 0.76 to −0.29 and 0.66 to −0.11 in 2010, 2015 and 2020. Experimental outcomes indicate five classes such as water bodies, built up, barren, cropland and fallow land during 2010, 2015 and 2020. The overall accuracy of User and Producer for 2010, 2015 and 2019 years in between 86.23%, 88.34%, 85.53% and 92.51%, 94.34% and 91.54%, respectively. We have observed that (2010, 2015 − 2020) agriculture and built-up land increased by 1040.76 ha, 1246.32 ha, 1500.93 ha and 34.96 ha, 37.08 ha, 42.58 ha, respectively. Other side degraded land, fallow land, waterbodies areas (953.19 ha, 679.23 ha, 937.24 ha and 1385.73 ha, 1513.53 ha, 991.08 ha and 32.85 ha, 21.33 ha, 25.66 ha) are increased during the year of 2010, 2015 and 2020, respectively. While results have been done by GEE cloud platform and remote sensing data, this developed algorithm easily classified the land use maps from Landsat-5 and Sentinel-2 TM imagery in the machine learning approach. The determined 30-m and 10-m three-year LULC maps are made-up to deliver vital data on the changes, monitoring and understanding of which types of LULC classes and changes have occupied a place in the Rahuri area.

Featherweight assisted vulnerability discovery

Predicting vulnerable source code helps to focus attention on those parts of the code that need to be examined with more scrutiny. Recent work proposed the use of function names as semantic cues that can be learned by a deep neural network (DNN) to aid in the hunt for vulnerability of functions. Combining identifier splitting, which splits each function name into its constituent words, with a novel frequency-based algorithm, we explore the extent to which the words that make up a function's name can predict potentially vulnerable functions. In contrast to *lightweight* predictions by a DNN that considers only function names, avoiding the use of a DNN provides *featherweight* predictions. The underlying idea is that function names that contain certain "dangerous" words are more likely to accompany vulnerable functions. Of course, this assumes that the frequency-based algorithm can be properly tuned to focus on truly dangerous words. Because it is more transparent than a DNN, the frequency-based algorithm enables us to investigate the inner workings of the DNN. If successful, this investigation into what the DNN does and does not learn will help us train more effective future models. We empirically evaluate our approach on a heterogeneous dataset containing over 73000 functions labeled vulnerable, and over 950000 functions labeled benign. Our analysis shows that words alone account for a significant portion of the DNN's classification ability. We also find that words are of greatest value in the datasets with a more homogeneous vocabulary. Thus, when working within the scope of a given project, where the vocabulary is unavoidably homogeneous, our approach provides a cheaper, potentially complementary, technique to aid in the hunt for source-code vulnerabilities. Finally, this approach has the advantage that it is viable with orders of magnitude less training data.

Multi-classification for high-dimensional data using probabilistic neural networks

Multi-classification tasks need sufficient information provided by the input data, whereas the input data lying in the high-dimensional space presents too sparse distributions to afford rich information, which creates trouble for multi-classification tasks. To address this, this paper proposed a probabilistic neural network approach. Firstly, the classification probability of high-dimensional data is calculated by using a novel Equation for Naïve Bayesian in order to reduce misclassification probability caused by the high dimension of the input data. Then, the calculated classification probability is fused into the cost function of the neural networks, forming the probabilistic neural networks. Results on high-dimensional real-world datasets and the National Surface Water Quality data provided by Environmental Monitoring Station, China show that the proposed method outperforms comparison methods in both classification ability and resisting the curse of dimensionality. We find that neural networks hardly obtain those advanced classification results upon a high-dimensional space because of the curse of dimensionality, but calculating classification probability of the input data can be confident to help neural networks to be resistance to the negative effects caused by high dimension of the input data.

Adversarial training for prostate cancer classification using magnetic resonance imaging.

To use adversarial training to increase the generalizability and diagnostic accuracy of deep learning models for prostate cancer diagnosis. This multicenter study retrospectively included 396 prostate cancer patients who underwent magnetic resonance imaging (development set, 297 patients from Shanghai Jiao Tong University Affiliated Sixth People's Hospital and Eighth People's Hospital; test set, 99 patients from Renmin Hospital of Wuhan University). Two binary classification deep learning models for clinically significant prostate cancer classification [PM1, pretraining Visual Geometry Group network (VGGNet)-16-based model 1; PM2, pretraining residual network (ResNet)-50-based model 2] and two multiclass classification deep learning models for prostate cancer grading (PM3, pretraining VGGNet-16-based model 3; PM4: pretraining ResNet-50-based model 4) were built using apparent diffusion coefficient and T2-weighted images. These models were then retrained with adversarial examples starting from the initial random model parameters (AM1, adversarial training VGGNet-16 model 1; AM2, adversarial training ResNet-50 model 2; AM3, adversarial training VGGNet-16 model 3; AM4, adversarial training ResNet-50 model 4, respectively). To verify whether adversarial training can improve the diagnostic model's effectiveness, we compared the diagnostic performance of the deep learning methods before and after adversarial training. Receiver operating characteristic curve analysis was performed to evaluate significant prostate cancer classification models. Differences in areas under the curve (AUCs) were compared using Delong's tests. The quadratic weighted kappa score was used to verify the PCa grading models. AM1 and AM2 had significantly higher AUCs than PM1 and PM2 in the internal validation dataset (0.84 vs. 0.89 and 0.83 vs. 0.87) and test dataset (0.73 vs. 0.86 and 0.72 vs. 0.82). AM3 and AM4 showed higher κ values than PM3 and PM4 in the internal validation dataset {0.266 [95% confidence interval (CI): 0.152-0.379] vs. 0.292 (95% CI: 0.178-0.405) and 0.254 (95% CI: 0.159-0.390) vs. 0.279 (95% CI: 0.163-0.396)} and test set [0.196 (95% CI: 0.029-0.362) vs. 0.268 (95% CI: 0.109-0.427) and 0.183 (95% CI: 0.015-0.351) vs. 0.228 (95% CI: 0.068-0.389)]. Using adversarial examples to train prostate cancer classification deep learning models can improve their generalizability and classification abilities.

Mental Workload Classification Method Based on EEG Cross-Session Subspace Alignment

Electroencephalogram (EEG) signals are sensitive to the level of Mental Workload (MW). However, the random non-stationarity of EEG signals will lead to low accuracy and a poor generalization ability for cross-session MW classification. To solve this problem of the different marginal distribution of EEG signals in different time periods, an MW classification method based on EEG Cross-Session Subspace Alignment (CSSA) is presented to identify the level of MW induced in visual manipulation tasks. The Independent Component Analysis (ICA) method is used to obtain the Independent Components (ICs) of labeled and unlabeled EEG signals. The energy features of ICs are extracted as source domains and target domains, respectively. The marginal distributions of source subspace base vectors are aligned with the target subspace base vectors based on the linear mapping. The Kullback–Leibler (KL) divergences between the two domains are calculated to select approximately similar transformed base vectors of source subspace. The energy features in all selected vectors are trained to build a new classifier using the Support Vector Machine (SVM). Then it can realize MW classification using the cross-session EEG signals, and has good classification accuracy.

StoHisNet: A hybrid multi-classification model with CNN and Transformer for gastric pathology images

Background and objectivesGastric cancer has high morbidity and mortality compared to other cancers. Accurate histopathological diagnosis has great significance for the treatment of gastric cancer. With the development of artificial intelligence, many researchers have applied deep learning for the classification of gastric cancer pathological images. However, most studies have used binary classification on pathological images of gastric cancer, which is insufficient with respect to the clinical requirements. Therefore, we proposed a multi-classification method based on deep learning with more practical clinical value. MethodsIn this study, we developed a novel multi-scale model called StoHisNet based on Transformer and the convolutional neural network (CNN) for the multi-classification task. StoHisNet adopts Transformer to learn global features to alleviate the inherent limitations of the convolution operation. The proposed StoHisNet can classify the publicly available pathological images of a gastric dataset into four categories —normal tissue, tubular adenocarcinoma, mucinous adenocarcinoma, and papillary adenocarcinoma. ResultsThe accuracy, F1-score, recall, and precision of the proposed model in the public gastric pathological image dataset were 94.69%, 94.96%, 94.95%, and 94.97%, respectively. We conducted additional experiments using two other public datasets to verify the generalization ability of the model. On the BreakHis dataset, our model performed better compared with other classification models, and the accuracy was 91.64%. Similarly, on the four-classification task on the Endometrium dataset, our model showed better classification ability than others with accuracy of 81.74%. These experiments showed that the proposed model has excellent ability of classification and generalization. ConclusionThe StoHisNet model had high performance in the multi-classification on gastric histopathological images and showed strong generalization ability on other pathological datasets. This model may be a potential tool to assist pathologists in the analysis of gastric histopathological images.

Performance Evaluation of Pre-Trained Convolutional Neural Network Model for Skin Disease Classification

Indonesia is a tropical country that has various skin diseases. Tinea versicolor, ringworm, and scabies are the most common types of skin diseases suffered by the people of Indonesia. The classification of the three skin diseases can be automatically completed by artificial intelligence and deep learning technology because the classification process using an expert will require a lot of money and time. The challenge in classifying skin diseases is in the process of collecting data. Because health data cannot be obtained freely, there must be approval from the patient or hospital. Therefore, to overcome the limited amount of data, Pre-Trained CNN is used. The Pre-Trained CNN model has many patterns from thousands of images, so we do not need many images to train the model. In this study, a comparison of five pre-trained CNN models was conducted, namely VGGNet16, MobileNetV2, InceptionResNetV2, ResNet152V2, and DenseNet201. The aim is to find out which CNN model can produce the best performance in classifying skin diseases with a limited amount of image data. The test results show that the ResNet152V2 model has the best classification ability with the highest accuracy, precision, recall, and F1 score values, namely 95.84%, 0.963, 0.96, and 0.956. As for the training execution time, the ResNet152V2 model has the fastest time to achieve 95% accuracy. That's happened because the addition of the dropout parameter is 20%.