Final Classification Research Articles

Genome-Wide Association Study of Conformation Traits in Brazilian Holstein Cattle.

The linear conformation of animals exerts an influence on health, reproduction, production, and welfare, in addition to longevity, which directly affects the profitability of milk-producing farms. The objectives of this study were (1) to perform genome-wide association studies (GWASs) of conformation traits, namely the Rump, Feet and Legs, Mammary System, Dairy Strength, and Final Classification traits, and (2) to identify genes and related pathways involved in physiological processes associated with conformation traits in Brazilian Holstein cattle. Phenotypic and genotypic data from 2339 Holstein animals distributed across the states of Rio Grande do Sul, Paraná, São Paulo, and Minas Gerais were used. The genotypic data were obtained with a 100 K SNP marker panel. The single-step genome-wide association study (ssGWAS) method was employed in the analyses. Genes close to a significant SNP were identified in an interval of 100 kb up- and downstream using the Ensembl database available in the BioMart tool. The DAVID database was used to identify the main metabolic pathways and the STRING program was employed to create the gene regulatory network. In total, 36 significant SNPs were found on 15 chromosomes; 27 of these SNPs were linked to genes that may influence the traits studied. Fourteen genes most closely related to the studied traits were identified, as well as four genes that showed interactions in important metabolic pathways such as myogenesis, adipogenesis, and angiogenesis. Among the total genes, four were associated with myogenesis (TMOD2, TMOD3, CCND2, and CTBP2), three with angiogenesis (FGF23, FGF1, and SCG3), and four with adipogenesis and body size and development (C5H12orf4, CCND2, EMILIN1, and FGF6). These results contribute to a better understanding of the biological mechanisms underlying phenotypic variability in conformation traits in Brazilian Holstein cattle.

MAFT-SO: A novel multi-atlas fusion template based on spatial overlap for ASD diagnosis

Autism spectrum disorder (ASD) is a common neurological condition. Early diagnosis and treatment are essential for enhancing the life quality of individuals with ASD. However, most existing studies either focus solely on the brain networks of subjects within a single atlas or merely employ simple matrix concatenation to represent the fusion of multi-atlas. These approaches neglected the natural spatial overlap that exists between brain regions across multi-atlas and did not fully capture the comprehensive information of brain regions under different atlases. To tackle this weakness, in this paper, we propose a novel multi-atlas fusion template based on spatial overlap degree of brain regions, which aims to obtain a comprehensive representation of brain networks. Specifically, we formally define a measurement of the spatial overlap among brain regions across different atlases, named spatial overlap degree. Then, we fuse the multi-atlas to obtain brain networks of each subject based on spatial overlap. Finally, the GCN is used to perform the final classification. The experimental results on Autism Brain Imaging Data Exchange (ABIDE) demonstrate that our proposed method achieved an accuracy of 0.757. Overall, our method outperforms SOTA methods in ASD/TC classification.

IMPRESS: Improved methylation profiling using restriction enzymes and smMIP sequencing, combined with a new biomarker panel, creating a multi-cancer detection assay

BackgroundDespite the worldwide progress in cancer diagnostics, more sensitive diagnostic biomarkers are needed. The methylome has been extensively investigated in the last decades, but a low-cost, bisulfite-free detection method for multiplex analysis is still lacking.MethodsWe developed a methylation detection technique called IMPRESS, which combines methylation-sensitive restriction enzymes and single-molecule Molecular Inversion Probes. We used this technique for the development of a multi-cancer detection assay for eight of the most lethal cancer types worldwide. We selected 1791 CpG sites that can distinguish tumor from normal tissue based on DNA methylation. These sites were analysed with IMPRESS in 35 blood, 111 tumor and 114 normal samples. Finally, a classifier model was built.ResultsWe present the successful development of IMPRESS and validated it with ddPCR. The final classifier model discriminating tumor from normal samples was built with 358 CpG target sites and reached a sensitivity of 0.95 and a specificity of 0.91. Moreover, we provide data that highlight IMPRESS’s potential for liquid biopsies.ConclusionsWe successfully created an innovative DNA methylation detection technique. By combining this method with a new multi-cancer biomarker panel, we developed a sensitive and specific multi-cancer assay, with potential use in liquid biopsies.

Circulating Interleukin-22 in Patients with Acute Myocardial Infarction Undergoing Primary Percutaneous Coronary Intervention.

Acute coronary syndrome (ACS) represents an important clinical manifestation of coronary artery disease (CAD) and is characterized by a particularly poor prognosis. Myocardial reperfusion through primary percutaneous coronary intervention (PPCI) is imperative in the event of acute ST elevation myocardial infarction (STEMI). Interleukin-22 (IL-22) regulates immune and inflammatory responses. This interleukin has been described in the scenario of the CAD, but there are no data in patients with STEMI undergoing PPCI. Objectives: The goals of this study were to investigate the differences in circulating IL-22 levels between patients with STEMI undergoing PPCI and healthy controls and to determine whether these differences were associated with the culprit coronary artery, door-to-balloon time (DBT), final angiographic result, CAD classification, and presence of diabetes mellitus (DM). Methods: A total of 280 participants were recruited, comprising 210 STEMI cases and 70 healthy controls. Participants underwent clinical and angiographic evaluations, and serum IL-22 levels were measured using an enzyme-linked immunosorbent assay (ELISA). Data analysis was performed using the Mann-Whitney and Fisher tests, with p < 0.05 indicating significance. Results: Serum IL-22 levels were lower in cases (149.63, 84.99-294.56) than in the controls (482.67, 344.33-641.00); p < 0.001. Lower IL-22 levels were associated with the right coronary artery (RCA) (144.57, 70.84-242.43; 146.00, 63.60-279.67; 191.71, 121.80-388.97); p = 0.033. IL-22 was lower with shorter DBT (≤60 min, 106.00, 49.60-171.71; >60 min, 153.00, 88.86-313.60); p = 0.043. Conclusions: IL-22 levels were significantly lower in patients with STEMI than in healthy controls.

Investigation of emergency department abandonment rates using machine learning algorithms in a single centre study

A critical problem that Emergency Departments (EDs) must address is overcrowding, as it causes extended waiting times and increased patient dissatisfaction, both of which are immediately linked to a greater number of patients who leave the ED early, without any evaluation by a healthcare provider (Leave Without Being Seen, LWBS). This has an impact on the hospital in terms of missing income from lost opportunities to offer treatment and, in general, of negative outcomes from the ED process. Consequently, healthcare managers must be able to forecast and control patients who leave the ED without being evaluated in advance. This study is a retrospective analysis of patients registered at the ED of the “San Giovanni di Dio e Ruggi d’Aragona” University Hospital of Salerno (Italy) during the years 2014–2021. The goal was firstly to analyze factors that lead to patients abandoning the ED without being examined, taking into account the features related to patient characteristics such as age, gender, arrival mode, triage color, day of week of arrival, time of arrival, waiting time for take-over and year. These factors were used as process measures to perform a correlation analysis with the LWBS status. Then, Machine Learning (ML) techniques are exploited to develop and compare several LWBS prediction algorithms, with the purpose of providing a useful support model for the administration and management of EDs in the healthcare institutions. During the examined period, 688,870 patients were registered and 39188 (5.68%) left without being seen. Of the total LWBS patients, 59.6% were male and 40.4% were female. Moreover, from the statistical analysis emerged that the parameter that most influence the abandonment rate is the waiting time for take-over. The final ML classification model achieved an Area Under the Curve (AUC) of 0.97, indicating high performance in estimating LWBS for the years considered in this study. Various patient and ED process characteristics are related to patients who LWBS. The possibility of predicting LWBS rates in advance could be a valid tool quickly identifying and addressing “bottlenecks” in the hospital organization, thereby improving efficiency.

Explainable machine learning for assessing upper respiratory tract of racehorses from endoscopy videos

Laryngeal hemiplegia (LH) is a major upper respiratory tract (URT) complication in racehorses. Endoscopy imaging of horse throat is a gold standard for URT assessment. However, current manual assessment faces several challenges, stemming from the poor quality of endoscopy videos and subjectivity of manual grading. To overcome such limitations, we propose an explainable machine learning (ML)-based solution for efficient URT assessment. Specifically, a cascaded YOLOv8 architecture is utilized to segment the key semantic regions and landmarks per frame. Several spatiotemporal features are then extracted from key landmarks points and fed to a decision tree (DT) model to classify LH as Grade 1,2,3 or 4 denoting absence of LH, mild, moderate, and severe LH, respectively. The proposed method, validated through 5-fold cross-validation on 107 videos, showed promising performance in classifying different LH grades with 100%, 91.18%, 94.74% and 100% sensitivity values for Grade 1 to 4, respectively. Further validation on an external dataset of 72 cases confirmed its generalization capability with 90%, 80.95%, 100%, and 100% sensitivity values for Grade 1 to 4, respectively. We introduced several explainability related assessment functions, including: (i) visualization of YOLOv8 output to detect landmark estimation errors which can affect the final classification, (ii) time-series visualization to assess video quality, and (iii) backtracking of the DT output to identify borderline cases. We incorporated domain knowledge (e.g., veterinarian diagnostic procedures) into the proposed ML framework. This provides an assistive tool with clinical-relevance and explainability that can ease and speed up the URT assessment by veterinarians.

A supervised machine learning procedure for EPMA classification and plotting of mineral groups

An analytical method to automatically characterize rock samples for geological or petrological purposes is here proposed, by applying machine learning approach (ML) as a protocol for saving experimental times and costs.Proper machine learning algorithms, applied to automatically acquired microanalytical data (i.e., Electron Probe Micro Analysis, EPMA), carried out with a SEM-EDS microprobe on randomly selected areas from a petrographic polished thin section, are trained, used, tested, and reported.Learning and Validation phases are developed with literature mineral databases of electron microprobe analyses on 15 main rock-forming mineral groups. The Prediction phase is tested using an eclogite rock from the Western Alps, considered as an unknown sample: randomly selected areas are acquired as backscattered images whose intervals of gray levels, appropriately set in the gray level histogram, allow the automated particle mineral separation: automated separating Oxford Instruments Aztec Feature ® packages and a mineral plotting software are applied for mineral particle separation, crystal chemical formula calculation and plotting.Finally, a microanalytical analysis is performed on each separated mineral particle. The crystal chemical formula is calculated, and the final classification plots are automatically produced for any determined mineral. The final results show good accuracy and analytical ease and assess the proper nature of the unknown eclogite rock sample. Therefore, the proposed analytical protocol is especially recommended in those scenarios where a large flow of microanalytical data is automatically acquired and needs to be processed.

DES-AS: Dynamic ensemble selection based on algorithm Shapley

Dynamic ensemble selection (DES) effectively improves aggregation performance by dynamically finding the most appropriate subset of classifiers for each query sample. The key component of this strategy is to define an appropriate criterion for evaluating the competence of the candidate classifiers. In this study, we present a new group-based criterion, synergy competence, to comprehensively measure the magnitude of the intricate synergy effect among classifiers. In addition, we introduce the algorithm Shapley, a variant of the Shapley value (SV) in the game theory, to measure the synergy competence of each classifier, which helps in the construction of a synergy-based DES system, called DES-AS. We argue that candidate classifiers with positive algorithm Shapley should be selected as they contribute to the ensemble. After selecting a subset of classifiers, the normalized algorithm Shapley is employed to calculate the voting weights for a cooperative final classification decision. The experimental analyses conducted on forty real-world datasets not only indicate that the synergy competence estimated by the algorithm Shapley is more applicable than classic group-based metrics, but also show that DES-AS is substantially more effective and robust than state-of-the-art models. The statistical test results also indicate that the estimated synergy competence reflects the real competence of each classifier better than the baseline models.

A multi-level collaborative self-distillation learning for improving adaptive inference efficiency

A multi-exit network is an important technique for achieving adaptive inference by dynamically allocating computational resources based on different input samples. The existing works mainly treat the final classifier as the teacher, enhancing the classification accuracy by transferring knowledge to the intermediate classifiers. However, this traditional self-distillation training strategy only utilizes the knowledge contained in the final classifier, neglecting potentially distinctive knowledge in the other classifiers. To address this limitation, we propose a novel multi-level collaborative self-distillation learning strategy (MLCSD) that extracts knowledge from all the classifiers. MLCSD dynamically determines the weight coefficients for each classifier’s contribution through a learning process, thus constructing more comprehensive and effective teachers tailored to each classifier. These new teachers transfer the knowledge back to each classifier through a distillation technique, thereby further improving the network’s inference efficiency. We conduct experiments on three datasets, CIFAR10, CIFAR100, and Tiny-ImageNet. Compared with the baseline network that employs traditional self-distillation, our MLCSD-Net based on ResNet18 enhances the average classification accuracy by 1.18%. The experimental results demonstrate that MLCSD-Net improves the inference efficiency of adaptive inference applications, such as anytime prediction and budgeted batch classification. Code is available at https://github.com/deepzlk/MLCSD-Net.

Deep Learning-Assisted Automatic Diagnosis of Anterior Cruciate Ligament Tear in Knee Magnetic Resonance Images.

Anterior cruciate ligament (ACL) tears are prevalent knee injures, particularly among active individuals. Accurate and timely diagnosis is essential for determining the optimal treatment strategy and assessing patient prognosis. Various previous studies have demonstrated the successful application of deep learning techniques in the field of medical image analysis. This study aimed to develop a deep learning model for detecting ACL tears in knee magnetic resonance Imaging (MRI) to enhance diagnostic accuracy and efficiency. The proposed model consists of three main modules: a Dual-Scale Data Augmentation module (DDA) to enrich the training data on both the spatial and layer scales; a selective group attention module (SG) to capture relationships across the layer, channel, and space scales; and a fusion module to explore the inter-relationships among various perspectives to achieve the final classification. To ensure a fair comparison, the study utilized a public dataset from MRNet, comprising knee MRI scans from 1250 exams, with a focus on three distinct views: axial, coronal, and sagittal. The experimental results demonstrate the superior performance of the proposed model, termed SGNET, in ACL tear detection compared with other comparison models, achieving an accuracy of 0.9250, a sensitivity of 0.9259, a specificity of 0.9242, and an AUC of 0.9747.

MMHFND: Fusing Modalities for Multimodal Multiclass Hindi Fake News Detection via Contrastive Learning

Multimodal content contains more deception than unimodal information, causing significant social and economic impacts. Current techniques often focus on single modality, neglecting knowledge fusion. While most studies have concentrated on English fake news detection, this study explores multi-modality for low-resource languages like Hindi. This work introduces the MMHFND model, based on M-CLIP, which uses late fusion for coarse (Fake vs Real) and fine-grained (World vs India vs Politics vs News vs Fact-Check) configurations. We extract deep representations from image and text using image transformer ResNet-50, a BERT-based L3cube-HindRoberta text transformer handling headlines, content, OCR text, and image captions, paired M-CLIP transformers, and an ELA (Error Level Analysis) image forensic method incorporating EfficientNet B0 to analyse multimodal news in Hindi language based on Devanagari script. M-CLIP integrates crossmodal similarity mapping of images and texts with retrieved multimodal features. The extracted features undergo redundancy reduction before being channelled into the final classifier. The MAM (Modality Attention Mechanism) is introduced, which generates weights for each modality individually. The MMHFND model uses a computed modality divergence score to identify dissonance between modalities and a modified contrastive loss on the score. We thoroughly analyse HinFakeNews dataset in a multimodal context, achieving accuracy in coarse and fine-grained configurations. We also undertake an ablation study to evaluate outcomes and explore alternative fusion processes on three different setups. The results show that the MMHFND model effectively detects fake news in Hindi with an accuracy of 0.986, outperforming other existing multimodal approaches.

Prediction of Super-enhancers Based on Mean-shift Undersampling

Background: Super-enhancers are clusters of enhancers defined based on the binding occupancy of master transcription factors, chromatin regulators, or chromatin marks. It has been reported that super-enhancers are transcriptionally more active and cell-type-specific than regular enhancers. Therefore, it is necessary to identify super-enhancers from regular enhancers. A variety of computational methods have been proposed to identify super-enhancers as auxiliary tools. However, most methods use ChIP-seq data, and the lack of this part of the data will make the predictor unable to execute or fail to achieve satisfactory performance. Objective: The aim of this study is to propose a stacking computational model based on the fusion of multiple features to identify super-enhancers in both human and mouse species. Methods: This work adopted mean-shift to cluster majority class samples and selected four sets of balanced datasets for mouse and three sets of balanced datasets for human to train the stacking model. Five types of sequence information are used as input to the XGBoost classifier, and the average value of the probability outputs from each classifier is designed as the final classification result. Results: The results of 10-fold cross-validation and cross-cell-line validation prove that our method has superior performance compared to other existing methods. The source code and datasets are available at https://github.com/Cheng-Han-max/SE_voting. Conclusion: The analysis of feature importance indicates that Mismatch accounts for the highest proportion among the top 20 important features.

Otitis media sequelae and hearing in adolescence after administration of an 11-valent conjugate pneumococcal vaccine in infancy: a prospective cohort study with long-term follow-up of the ARIVAC trial

Pneumococcal conjugate vaccines (PCVs) have been shown in randomised controlled trials and epidemiological studies to prevent acute otitis media caused by vaccine serotype pneumococci, although their role in preventing complications of acute otitis media is less clear. We hypothesised that the 11-valent PCV would reduce the long-term sequelae of acute otitis media, including moderate-to-severe ear disease and hearing loss. This prospective cohort study, referred to as 11PCV study, included follow-up after 16-20 years of children previously enrolled in 2000-04, at age 6 weeks to 6 months, in the randomised, placebo-controlled, ARIVAC trial of 11-valent PCV for the prevention of radiographical pneumonia. The ARIVAC trial and this 11PCV study were conducted at six study centres in Bohol, Philippines. Ear disease was classified using video-otoscopy review and observations derived from the ear exam. The final classification of the worst ear disease was mild (ie, acute otitis media, otitis media with effusion, healed perforation, or tympanosclerosis), moderate (ie, dry perforation or adhesive otitis media), or severe (chronic suppurative otitis media). Hearing loss was assessed following a standard schema and classified according to the worst ear as mild (>15 to 30 dB puretone average) or moderate-to-profound (>30 dB pure tone average). We calculated the relative and absolute risk reduction in the primary outcome of moderate-to-severe ear disease and the secondary outcomes of mild or moderate-to-profound hearing loss in adolescents who previously received the 11-valent PCV compared with those who received placebo during infancy in ARIVAC. Of the 15 593 children assessed for eligibility in ARIVAC, 12 194 were randomly assigned and 8926 were alive and could be located for enrolment in this 11PCV study between Sept 19, 2016, and Dec 13, 2019. 8321 (4188 in the vaccine group and 4133 in the placebo group) completed follow-up of the 11PCV study by March 30, 2020, and had sufficient data to classify ear disease and be included in the primary outcome analysis. The primary outcome of the absolute risk reduction in moderate-to-severe ear disease in the vaccine group (310 [7·4%] of 4188) versus those in the placebo group (356 [8·6%] of 4133) was 1·2% (95% CI 0·0-2·4; p=0·046) and the relative risk reduction was 14·1% (0·0 to 26·0). There were no differences in secondary outcomes of mild hearing loss or moderate-to-profound hearing loss between the vaccine and placebo groups. The absolute risk reduction for moderate-to-severe ear disease in adolescence of 1·2% (12 per 1000 children) was almost three times higher than the 0·45% reduction (4·5 per 1000 children) in radiographical pneumonia in the first 2 years of life shown in ARIVAC. Administration of 11-valent PCV in infancy was associated with absolute and relative risk reductions in the sequelae of acute otitis media 16-20 years after the original ARIVAC trial. Bill & Melinda Gates Foundation.

Authenticity Verification of Alcoholic Beverages in IoT Environment through Fluorescence Hyperspectral Technology and Machine Learning

In the Internet of Things (IoT) environment, verifying the authenticity of beverages is an important issue. Baijiu, a unique Chinese distilled liquor, has gained popularity among people due to its distinctive flavor. In recent years, there has been increasing concern about the adulteration of baijiu. Many studies in this field have attempted to combine chemometrics, electronic nose, Raman spectroscopy, and other detection technologies with machine learning for baijiu detection. However, no research has yet applied the mature detection technology of fluorescence hyperspectral technology to this field. Additionally, existing studies have mainly focused on branded baijiu, without addressing the problem of adulteration in bulk baijiu, which remains a long-standing issue. In this study, a combination of fluorescence hyperspectral and machine learning was used for rapid and nondestructive adulteration detection of bulk liquor in Chengdu. The fluorescence hyperspectral data of three types of samples, namely, original liquor, adulterated liquor with water, and adulterated liquor with industrial alcohol, were obtained, and the raw data were preprocessed using Savitzky–Golay (SG), baseline correction (BL), multivariate scatter correction (MSC), and the combination of Savitzky–Golay and baseline correction (SG-BL). Finally, PCA and MDS are also used to extract spectral data characteristics, respectively. For model selection, three models were chosen: back-propagation artificial neural network (BP-ANN), support vector machine (SVM), and one-dimensional convolutional neural network (1D-CNN). For model evaluation, we choose accuracy, recall, precision, F1-score, memory occupied, and time consumed by the model to evaluate. The experiments show that the combination of fluorescence hyperspectral technology and machine learning can provide a new method, MSC-PCA-SVM, for bulk liquor detection, with a final classification accuracy, recall, precision, and F1-score all reaching 100%, The effectiveness and application potential of this method in the IoT environment have been demonstrated. The results of this study can not only be applied to liquor detection but also have the value of reference for the adulthood and detection of the entire food industry. It is of great significance to standardize the market and protect consumer rights.

Learning end-to-end respiratory rate prediction of dairy cows from red, green, and blue videos

Respiratory rate (RR) is an important indicator of the health and welfare status of dairy cows. In recent years, progress has been made in monitoring the RR of dairy cows using video data and learning methods. However, existing approaches often involve multiple processing modules, such as region of interest (ROI) detection and tracking, which can introduce errors that propagate through successive steps. The objective of this study was to develop an end-to-end computer vision method to predict RR of dairy cows continuously and automatically. The method leverages the capabilities of a state-of-the-art Transformer model, VideoMAE, which divides video frames into patches as input tokens, enabling the automated selection and featurization of relevant regions, such as a cow's abdomen, for predicting RR. The original encoder of VideoMAE was retained, and a classification head was added on top of it. Further, the weights of the first 11 layers of the pre-trained model were kept, whereas the weights of the final layer and classifier were fine-tuned using video data collected in a tiestall barn from 6 dairy cows. Respiratory rates measured using a respiratory belt for individual cows were serving as the ground truth (GT). The evaluation of the developed model was conducted using multiple metrics, including mean absolute error (MAE) of 2.58 breaths per minute (bpm), root mean squared error (RMSE) of 3.52 bpm, root mean squared prediction error (RMSPE; as a proportion of observed mean) of 15.03%, and Pearson r of 0.86. Compared with a conventional method involving multiple processing modules, the end-to-end approach performed better in terms of MAE, RMSE, and RMSPE. These results suggest the potential to implement the developed computer vision method for an end-to-end solution, for monitoring RR of dairy cows automatically in a tiestall setting. Future research on integrating this method with other behavioral detection and animal identification algorithms for animal monitoring in a freestall dairy barn can be beneficial for a broader application.

Representation of coxeter group and orthogonal group

The paper is primarily divided into two parts. The main focus of the first part is the construction of a representation of Coxeter groups. This begins with the definition of the Coxeter system and connected components, followed by the introduction of the length function and subsequent theorems. The faithfulness of this representation is then proven, allowing for the identification of isomorphisms that enable the final classification of finite Coxeter groups. This classification is achieved by leveraging the established relationship between irreducible representations of Coxeter groups and positive definite quadratic forms. Given the strong connection between Coxeter groups and orthogonal groups, the primary objective of the second part is to create a specific representation of orthogonal groups. This is accomplished through an examination of the decomposition of harmonic polynomials into subspaces of homogeneous harmonic polynomials, using the action of O(2) on these subspaces. The paper concludes by drawing connections to results in Invariant Theory, demonstrating the applicability of the presented concepts in a more general duality context.

Breast Cancer Detection and Analytics Using Hybrid CNN and Extreme Learning Machine.

Early detection of breast cancer is essential for increasing survival rates, as it is one of the primary causes of death for women globally. Mammograms are extensively used by physicians for diagnosis, but selecting appropriate algorithms for image enhancement, segmentation, feature extraction, and classification remains a significant research challenge. This paper presents a computer-aided diagnosis (CAD)-based hybrid model combining convolutional neural networks (CNN) with a pruned ensembled extreme learning machine (HCPELM) to enhance breast cancer detection, segmentation, feature extraction, and classification. The model employs the rectified linear unit (ReLU) activation function to enhance data analytics after removing artifacts and pectoral muscles, and the HCPELM hybridized with the CNN model improves feature extraction. The hybrid elements are convolutional and fully connected layers. Convolutional layers extract spatial features like edges, textures, and more complex features in deeper layers. The fully connected layers take these features and combine them in a non-linear manner to perform the final classification. ELM performs classification and recognition tasks, aiming for state-of-the-art performance. This hybrid classifier is used for transfer learning by freezing certain layers and modifying the architecture to reduce parameters, easing cancer detection. The HCPELM classifier was trained using the MIAS database and evaluated against benchmark methods. It achieved a breast image recognition accuracy of 86%, outperforming benchmark deep learning models. HCPELM is demonstrating superior performance in early detection and diagnosis, thus aiding healthcare practitioners in breast cancer diagnosis.

SFPL: Sample-specific fine-grained prototype learning for imbalanced medical image classification

Imbalanced classification is a common and difficult task in many medical image analysis applications. However, most existing approaches focus on balancing feature distribution and classifier weights between classes, while ignoring the inner-class heterogeneity and the individuality of each sample. In this paper, we proposed a sample-specific fine-grained prototype learning (SFPL) method to learn the fine-grained representation of the majority class and learn a cosine classifier specifically for each sample such that the classification model is highly tuned to the individual’s characteristic. SFPL first builds multiple prototypes to represent the majority class, and then updates the prototypes through a mixture weighting strategy. Moreover, we proposed a uniform loss based on set representations to make the fine-grained prototypes distribute uniformly. To establish associations between fine-grained prototypes and cosine classifier, we propose a selective attention aggregation module to select the effective fine-grained prototypes for final classification. Extensive experiments on three different tasks demonstrate that SFPL outperforms the state-of-the-art (SOTA) methods. Importantly, as the imbalance ratio increases from 10 to 100, the improvement of SFPL over SOTA methods increases from 2.2% to 2.4%; as the training data decreases from 800 to 100, the improvement of SFPL over SOTA methods increases from 2.2% to 3.8%.

A robust ensemble deep learning framework for accurate diagnoses of tuberculosis from chest radiographs.

Tuberculosis (TB) stands as a paramount global health concern, contributing significantly to worldwide mortality rates. Effective containment of TB requires deployment of cost-efficient screening method with limited resources. To enhance the precision of resource allocation in the global fight against TB, this research proposed chest X-ray radiography (CXR) based machine learning screening algorithms with optimization, benchmarking and tuning for the best TB subclassification tasks for clinical application. This investigation delves into the development and evaluation of a robust ensemble deep learning framework, comprising 43 distinct models, tailored for the identification of active TB cases and the categorization of their clinical subtypes. The proposed framework is essentially an ensemble model with multiple feature extractors and one of three fusion strategies-voting, attention-based, or concatenation methods-in the fusion stage before a final classification. The comprised de-identified dataset contains records of 915 active TB patients alongside 1,276 healthy controls with subtype-specific information. Thus, the realizations of our framework are capable for diagnosis with subclass identification. The subclass tags include: secondary tuberculosis/tuberculous pleurisy; non-cavity/cavity; secondary tuberculosis only/secondary tuberculosis and tuberculous pleurisy; tuberculous pleurisy only/secondary tuberculosis and tuberculous pleurisy. Based on the dataset and model selection and tuning, ensemble models show their capability with self-correction capability of subclass identification with rendering robust clinical predictions. The best double-CNN-extractor model with concatenation/attention fusion strategies may potentially be the successful model for subclass tasks in real application. With visualization techniques, in-depth analysis of the ensemble model's performance across different fusion strategies are verified. The findings underscore the potential of such ensemble approaches in augmenting TB diagnostics with subclassification. Even with limited dataset, the self-correction within the ensemble models still guarantees the accuracies to some level for potential clinical decision-making processes in TB management. Ultimately, this study shows a direction for better TB screening in the future TB response strategy.

An adaptive and late multifusion framework in contextual representation based on evidential deep learning and Dempster–Shafer theory

AbstractThere is a growing interest in multidisciplinary research in multimodal synthesis technology to stimulate diversity of modal interpretation in different application contexts. The real requirement for modality diversity across multiple contextual representation fields is due to the conflicting nature of data in multitarget sensors, which introduces other obstacles including ambiguity, uncertainty, imbalance, and redundancy in multiobject classification. This paper proposes a new adaptive and late multimodal fusion framework using evidence-enhanced deep learning guided by Dempster–Shafer theory and concatenation strategy to interpret multiple modalities and contextual representations that achieves a bigger number of features for interpreting unstructured multimodality types based on late fusion. Furthermore, it is designed based on a multifusion learning solution to solve the modality and context-based fusion that leads to improving decisions. It creates a fully automated selective deep neural network and constructs an adaptive fusion model for all modalities based on the input type. The proposed framework is implemented based on five layers which are a software-defined fusion layer, a preprocessing layer, a dynamic classification layer, an adaptive fusion layer, and an evaluation layer. The framework is formalizing the modality/context-based problem into an adaptive multifusion framework based on a late fusion level. The particle swarm optimization was used in multiple smart context systems to improve the final classification layer with the best optimal parameters that tracing 30 changes in hyperparameters of deep learning training models. This paper applies multiple experimental with multimodalities inputs in multicontext to show the behaviors the proposed multifusion framework. Experimental results on four challenging datasets including military, agricultural, COIVD-19, and food health data provide impressive results compared to other state-of-the-art multiple fusion models. The main strengths of proposed adaptive fusion framework can classify multiobjects with reduced features automatically and solves the fused data ambiguity and inconsistent data. In addition, it can increase the certainty and reduce the redundancy data with improving the unbalancing data. The experimental results of multimodalities experiment in multicontext using the proposed multimodal fusion framework achieve 98.45% of accuracy.