Hierarchical Classification Scheme Research Articles

Mapping Annual Tidal Flat Loss and Gain in the Micro-Tidal Area Integrating Dual Full-Time Series Spectral Indices

Tracking long-term tidal flat dynamics is crucial for coastal restoration decision making. Accurately capturing the loss and gain of tidal flats due to human-induced disturbances is challenging in the micro-tidal areas. In this study, we developed an automated method for mapping the annual tidal flat changes in the micro-tidal areas under intense human activities, by integrating spectral harmonization, time series segmentation from dual spectral indices, and the tide-independent hierarchical classification strategy. Our method has two key novelties. First, we adopt flexible temporal segments for each pixel based on the dual full-time series spectral indices, instead of solely using a fixed period window, to help obtain more reliable inundation frequency features. Second, a tide-independent hierarchical classification strategy based on the inundation features and the Otsu algorithm capture the tidal flat changes well. Our method performed well in Guangdong, Hong Kong, and Macao (GHKM), a typical area with micro-tidal range and intense human activities, with overall accuracies of 89% and 92% for conversion types and turning years, respectively. The tidal flats in GHKM decreased by 24% from 1986 to 2021, resulting from the loss of 504.45 km2, partially offset by an accretion of 179.88 km2. Further, 70.9% of the total loss was in the Great Bay Area, concentrated in 1991–1998 and 2001–2016. The historical trajectories of tidal flat loss were driven by various policies implemented by the national, provincial, and local governments. Our method is promising for extension to other micro-tidal areas to provide more scientific support for coastal resource management and restoration.

Comparing Hierarchical Approaches to Enhance Supervised Emotive Text Classification

The performance of emotive text classification using affective hierarchical schemes (e.g., WordNet-Affect) is often evaluated using the same traditional measures used to evaluate the performance of when a finite set of isolated classes are used. However, applying such measures means the full characteristics and structure of the emotive hierarchical scheme are not considered. Thus, the overall performance of emotive text classification using emotion hierarchical schemes is often inaccurately reported and may lead to ineffective information retrieval and decision making. This paper provides a comparative investigation into how methods used in hierarchical classification problems in other domains, which extend traditional evaluation metrics to consider the characteristics of the hierarchical classification scheme, can be applied and subsequently improve the classification of emotive texts. This study investigates the classification performance of three widely used classifiers, Naive Bayes, J48 Decision Tree, and SVM, following the application of the aforementioned methods. The results demonstrated that all the methods improved the emotion classification. However, the most notable improvement was recorded when a depth-based method was applied to both the testing and validation data, where the precision, recall, and F1-score were significantly improved by around 70 percentage points for each classifier.

Hierarchical Classification of Transversal Skills in Job Advertisements Based on Sentence Embeddings

This paper proposes a classification methodology aimed at identifying correlations between job ad requirements and transversal skill sets, with a focus on predicting the necessary skills for individual job descriptions using a deep learning model. The approach involves data collection, preprocessing, and labeling using ESCO (European Skills, Competences, and Occupations) taxonomy. Hierarchical classification and multi-label strategies are used for skill identification, while augmentation techniques address data imbalance, enhancing model robustness. A comparison between results obtained with English-specific and multi-language sentence embedding models reveals close accuracy. The experimental case studies detail neural network configurations, hyperparameters, and cross-validation results, highlighting the efficacy of the hierarchical approach and the suitability of the multi-language model for the diverse European job market. Thus, a new approach is proposed for the hierarchical classification of transversal skills from job ads.

The cost of crossing gender boundaries: Trans women of color and the racialized workplace gender order

AbstractHow does persistent and cumulative gender regulation produce economic insecurity? Trans people face markedly high levels of workplace discrimination, unemployment, and poverty, and therefore offer unique insight into this question. Prior research theorizes how trans workers get repositioned in a binary, patriarchal gender order, but we lack a conceptual model to explain the labor market experience of people who are systematically sanctioned as gender deviants. By analyzing work history interviews from 23 trans women of color based across the United States, this article argues that crossing gender boundaries is a racialized experience that can come with an economic cost. After transitioning, trans women of color face three forms of economic sanctioning: exclusion, a racially gendered glass ceiling, and constrained employment options within a segmented labor market. Thus, work organizations premised on a hierarchical classification scheme have the option, not only to reposition people on the basis of a classification change, but to deem them unassimilable.

Developing a Quality Improvement Implementation Taxonomy for Organizational Employee Wellness Initiatives.

Standardized taxonomies (STs) facilitate knowledge representation and semantic interoperability within health care provision and research. However, a gap exists in capturing knowledge representation to classify, quantify, qualify, and codify the intersection of evidence and quality improvement (QI) implementation. This interprofessional case report leverages a novel semantic and ontological approach to bridge this gap. This report had two objectives. First, it aimed to synthesize implementation barrier and facilitator data from employee wellness QI initiatives across Veteran Affairs health care systems through a semantic and ontological approach. Second, it introduced an original framework of this use-case-based taxonomy on implementation barriers and facilitators within a QI process. We synthesized terms from combined datasets of all-site implementation barriers and facilitators through QI cause-and-effect analysis and qualitative thematic analysis. We developed the Quality Improvement and Implementation Taxonomy (QIIT) classification scheme to categorize synthesized terms and structure. This framework employed a semantic and ontological approach. It was built upon existing terms and models from the QI Plan, Do, Study, Act phases, the Consolidated Framework for Implementation Research domains, and the fishbone cause-and-effect categories. The QIIT followed a hierarchical and relational classification scheme. Its taxonomy was linked to four QI Phases, five Implementing Domains, and six Conceptual Determinants modified by customizable Descriptors and Binary or Likert Attribute Scales. This case report introduces a novel approach to standardize the process and taxonomy to describe evidence translation to QI implementation barriers and facilitators. This classification scheme reduces redundancy and allows semantic agreements on concepts and ontological knowledge representation. Integrating existing taxonomies and models enhances the efficiency of reusing well-developed taxonomies and relationship modeling among constructs. Ultimately, employing STs helps generate comparable and sharable QI evaluations for forecast, leading to sustainable implementation with clinically informed innovative solutions.

Using multimodal, low‐cost data to efficiently identify clinical subgroups and classify Alzheimer’s diagnosis

AbstractBackgroundWith increasingly greater prevalence of Alzheimer’s disease (AD) worldwide, cost‐ and time‐efficient methods for identifying at‐risk individuals are needed. Survey, simple neuropsychological, and demographic information are low‐cost, but their lower levels of sensitivity or specificity when considered individually are problematic. However, past work has demonstrated the usefulness of statistical machine learning tools for predicting clinical diagnosis using multimodal data. To better understand and quantify Alzheimer’s risk and progression at follow‐up using lower‐cost data, we apply statistical and machine learning tools to screen, cluster, and classify individuals from the National Alzheimer’s Coordinating Center (NACC) dataset.MethodPreviously, we developed hierarchical classification strategies to use lower‐cost data for 3‐class prediction (unimpaired, AD‐MCI, and AD‐dementia). Building upon this idea, here we leverage a Gaussian mixture model to infer subgroup identities of subjects and use subgroup identity to further aid classification. Individuals belonging to homogeneous subgroups consisting primarily of a single diagnosis are more likely to also be given that diagnosis. This approach enables confidence in screening given a threshold for identifying the primary diagnosis within each cluster. Additional classification is performed on subjects that belong to heterogeneous subgroups. The NACC database is funded by NIA/NIH Grant U24 AG072122. NACC data are contributed by the NIA‐funded ADRCs.ResultThe proposed method unmasks the clinical subgroup trends within the sample and allows for detailed longitudinal risk tracking. We identified five subgroups within the sample, representing five clinical clusters. These subgroups include predominantly unimpaired individuals, mixtures of unimpaired and MCI, MCI and AD‐dementia, and individuals predominantly with AD‐dementia. A large proportion of individuals can be confidently screened based upon subgroup identity. Using only baseline visit information, predictions of clinical diagnosis post‐clustering is highly reflective of conversion to a worse diagnosis within 18 months. Changes in the conversion rate to a more severe diagnosis are dependent on the cluster identity.ConclusionUsing cost‐ and time‐efficient data enables more effective screening and tracking of disease progression. The proposed approach of clustering, screening, and prediction yields a measure of confidence in predictions of clinical diagnosis and allows for greater understanding of clinical profiles and progression.

A New Method for Bare Permafrost Extraction on the Tibetan Plateau by Integrating Machine Learning and Multi-Source Information

Bare permafrost refers to permafrost with almost no vegetation on the surface, which is an essential part of the ecosystem of the Tibetan Plateau. An accurate extraction of the boundaries of bare permafrost is vital for studying how it is being impacted by climate change. The accuracy of permafrost and bare land distribution maps is inadequate, and the spatial and temporal resolution is low. This is due to the challenges associated with obtaining significant amounts of data in high-altitude and alpine regions and the limitations of current mapping techniques in effectively integrating multiple factors. This study introduces a novel approach to extracting information about the distribution of bare permafrost. The approach introduced here involves amalgamating a sample extraction method, the fusion of multi-source remote sensing information, and a hierarchical classification strategy. Initially, the available multi-source permafrost data, expert knowledge, and refinement rules for training samples are integrated to produce extensive and consistent permafrost training samples. Using the random forest method, these samples are then utilized to create features and classify permafrost. Subsequently, a methodology utilizing a hierarchical classification approach in conjunction with machine learning techniques is implemented to identify an appropriate threshold for fractional vegetation cover, thereby facilitating the extraction of bare land. The bare permafrost boundary is ultimately derived through layer overlay analysis. The permafrost classification exhibits an overall accuracy of 90.79% and a Kappa coefficient of 0.806. The overall accuracies of the two stratified extractions in bare land were 97.47% and 96.99%, with Kappa coefficients of 0.954 and 0.911. The proposed approach exhibits superiority over the extant bare land and permafrost distribution maps. It is well-suited for retrieving vast bare permafrost regions and is valuable for acquiring bare permafrost distribution data across a vast expanse. It offers technical assistance in acquiring extended-term data on the distribution of exposed permafrost on the Tibetan Plateau. Furthermore, it facilitates the elucidation of the impact of climate change on exposed permafrost.

Atolls of the world: A reappraisal from an optical remote sensing and global mapping perspective

The Millennium Coral Reef Mapping Project (MCRMP) aimed to map coral reefs worldwide at geomorphological thematic scales using Landsat satellite images at 30 m spatial resolution. The 5-level hierarchical classification scheme implemented by MCRMP identified at Level 2 ‘Atolls’ as one of the main types of coral reef complexes. In this review, the qualitative criteria used by MCRMP to identify atolls are presented. Then, we report on the global census of atolls, from which a consistent geomorphologic GIS database is provided. A total of 598 atolls are identified and mapped. The quantitative database provides surface areas for all geomorphologic units at Level 4 and 5 of the MCRMP classification scheme. It allows further work on atoll and coral reef classifications, regionally and globally, in order to identify geomorphic trends and outliers. It is also a convenient database to use in multivariate analyses with ancillary biodiversity, fishery, socio-economic or climate data.

Hierarchical Classification Strategy for Mitigating the Impact of The Presence of Pain in fNIRS-based BCIs.

Brain computer interfaces (BCIs) can find applications in assistive systems for patients who experience conditions that impede their motor abilities. A BCI uses signals acquired from the brain to control external devices. As physical pain influences cortical signals, the presence of pain can negatively impact the performance of the BCI. In this work, we propose a strategy to mitigate this negative impact. Cortical signals are acquired from test subjects while they performed two mental arithmetic tasks, in the presence and the absence of painful stimuli. The task of the BCI is to reliably classify the two mental arithmetic tasks from the cortical recordings, irrespective of the presence or the absence of pain. We propose to do this classification, hierarchically, in two levels. In the first level, the data is classified into those captured in the presence and the absence of pain. Depending on the results of the classification from the first level, in the second level, the BCI performs the classification of tasks using a classifier trained either in the presence or the absence of pain. A 1-dimensional convolutional neural network (1D-CNN) is used for classification at both levels. It is observed that using this hierarchical strategy, the BCI is able to classify the tasks with an accuracy greater than 90%, irrespective of the presence or the absence of pain. Given that the presence of physical pain has shown previously to reduce the classification accuracy of a BCI to almost chance levels, this mitigation strategy will be a significant step towards enhancing the performance of BCIs when they are used in assistive systems for patients.

Hierarchical confusion matrix for classification performance evaluation

Abstract This study proposes the novel concept of hierarchical confusion matrix, opening the door for popular confusion-matrix-based (flat) evaluation measures from binary classification problems, while considering the peculiarities of hierarchical classification problems. The concept is developed to a generalised form and proven its applicability to all types of hierarchical classification problems including directed acyclic graphs, multi-path labelling, and non-mandatory leaf-node prediction. Finally, measures based on the novel confusion matrix are used for three real-world hierarchical classification applications and compared to established evaluation measures. The results, the conformity with important attributes of hierarchical classification schemes and its broad applicability justify its recommendation.

Triplet-Net Classification of Contiguous Stem Cell Microscopy Images.

Cellular microscopy imaging is a common form of data acquisition for biological experimentation. Observation of gray-level morphological features allows for the inference of useful biological information such as cellular health and growth status. Cellular colonies can contain multiple cell types, making colony level classification very difficult. Additionally, cell types growing in a hierarchical, downstream fashion, can often look visually similar, although biologically distinct. In this paper, it is determined empirically that traditional deep Convolutional Neural Networks (CNN) and classical object recognition techniques are not sufficient to distinguish between these subtle visual differences, resulting in misclassifications. Instead, Triplet-net CNN learning is employed in a hierarchical classification scheme to improve the ability of the model to discern distinct, fine-grain features of two commonly confused morphological image-patch classes, namely Dense and Spread colonies. The Triplet-net method improves classification accuracy over a four-class deep neural network by ∼ 3 %, a value that was determined to be statistically significant, as well as existing state-of-the-art image patch classification approaches and standard template matching. These findings allow for the accurate classification of multi-class cell colonies with contiguous boundaries, and increased reliability and efficiency of automated, high-throughput experimental quantification using non-invasive microscopy.

Applying the phylogeographic concept to identify franciscana dolphin stocks: implications to meet management objectives

High numbers of franciscanas (Pontoporia blainvillei) have been bycaught in gillnets for at least four decades. The impact is strong but not homogeneous along the species distribution range, and there is evidence that at least one local population is declining. Reliable stock determination is important to evaluate how bycatch affects stocks so that they can be preserved through local management actions. Following the phylogeographic approach of Dizon et al. (1992), which applies a hierarchical classification scheme to stock designations, available information relevant for franciscana stock discreteness was reviewed, including data on distribution, population response, phenotype and genotype. Data on mtDNA, morphometrics and population parameters all together provide evidence for splitting the species into four management stocks: two inhabiting coastal waters of Brazil; the third occurring in Rio Grande do Sul State (southern Brazil) and Uruguay, and the fourth inhabiting coastal Argentine waters. The areas where these stocks occur are herein called Franciscana Management Areas or FMA I to IV. Although in some cases the evidence for such sub-division into four stocks is weak due to lack of data, the use of the proposed classification into four FMAs is recommended to warrant effective management on a local scale.

Hierarchical strategy for sEMG classification of the hand/wrist gestures and forces of transradial amputees.

The myoelectric control strategy, based on surface electromyographic signals, has long been used for controlling a prosthetic system with multiple degrees of freedom. Several methods classify gestures and force levels but the simultaneous real-time control of hand/wrist gestures and force levels did not yet reach a satisfactory level of effectiveness. In this work, the hierarchical classification approach, already validated on 31 healthy subjects, was adapted for the real-time control of a multi-DoFs prosthetic system on 15 trans-radial amputees. The effectiveness of the hierarchical classification approach was assessed by evaluating both offline and real-time performance using three algorithms: Logistic Regression (LR), Non-linear Logistic Regression (NLR), and Linear Discriminant Analysis (LDA). The results of this study showed the offline performance of amputees was promising and comparable to healthy subjects, with mean F1 scores of over 90% for the "Hand/wrist gestures classifier" and 95% for the force classifiers, implemented with the three algorithms with features extraction (FE). Another significant finding of this study was the feasibility of using the hierarchical classification strategy for real-time applications, due to its ability to provide a response time of 100 ms while maintaining an average online accuracy of above 90%. A possible solution for real-time control of both hand/wrist gestures and force levels is the combined use of the LR algorithm with FE for the "Hand/wrist gestures classifier", and the NLR with FE for the Spherical and Tip force classifiers.

Introducing multi-dimensional hierarchical classification: Characterization, solving strategies and performance measures

Classification problems where there exist multiple class variables that need to be jointly predicted are known as Multi-dimensional classification problems. If the labels of these class variables are organized as hierarchies, we can take advantage of specific strategies designed for the Hierarchical classification paradigm. In this paper we present the Multi-dimensional hierarchical classification (MDHC) paradigm, a result of the combination of Multi-dimensional and Hierarchical classification paradigms. We propose four MDHC learning strategies which are designed to exploit the particularities of this new paradigm, combining characteristics of Multi-dimensional and Hierarchical classification strategies. Along with these strategies, we present a framework for classifier comparison in which we use a set of performance measures specifically designed for MDHC, and a procedure to create MDHC synthetic scenarios. Using this framework and the performance measures presented, we study how characteristics of the MDHC problems influence the performance of the different MDHC strategies proposed, and compare them to other non-MDHC strategies.

Digital Medieval Studies: Practice and Preservation ed. by Laura K. Morreale and Sean Gilsdorf

Reviewed by: Digital Medieval Studies: Practice and Preservation ed. by Laura K. Morreale and Sean Gilsdorf Paul Evans Digital Medieval Studies: Practice and Preservation. By Laura K. Morreale and Sean Gilsdorf, eds. Leeds: Arc Humanities Press, 2022. Digital Medieval Studies is a short, 121-page, edited collection that makes an argument for using a specific methodology, known as the digital documentation process (DDP), to document digital scholarly projects in medieval studies. Although it is now common for medievalists to preserve public GitHub repositories associated with their digital projects, DDP is a more structured approach aimed at documenting projects narratively in a form accessible to scholars who are not versed in digital methodologies, including those who are involved in hiring and tenure decisions. Medievalists who have done at least some digital work are the book's intended audience, and its primary agenda is to encourage them to preserve their projects in what, the contributors hope, will become a comprehensive, persistent, corpus of digital medieval scholarship. Except for an introduction by editors Laura K. Morreale and Sean Gilsdorf and a concluding retrospective by Lisa Fagin Davis, the contributed chapters focus on projects that have been archived according to DDP. Although the chapter authors do not go out of their way to foreground the process, their individual articles form an expanded commentary on each project's archiving dossier narrative (ADN), the historical component of DDP. Readers interested in archiving their own projects according to DDP will benefit from consulting the project information in the appendix to connect the content of the essays with the corresponding digital artifacts. In the book's five contributed chapters, the editors offer a commendable range of perspectives, spanning public humanities, research, teaching, and technical approaches. In chapter 1, Deborah Everhart and Martin Irvine narrate the history of the Labyrinth, a 1990s-era portal that collected and presented a curated set of links to resources for medieval studies. The authors take justifiable pride in this early project, which established medieval studies as a first mover in using the web as a vehicle for scholarly communication and for what today is called public humanities. They note, however, that the demise of the Labyrinth project offers a cautionary lesson on the vulnerability of still-valuable [End Page 147] project information to the vagaries of platform evolution. Today there are only two existing archives of the Labyrinth contents: the original hand-edited HTML from 2007 and an Adobe ColdFusion format from 2014. ColdFusion was a web platform that enjoyed some vogue among university IT departments until about 2010 but is rarely used now. This means that the 2014 archive is effectively unrecoverable for most users. In chapter 2, John McEwan describes the digital sigillography project (DIGISIG), which makes the standard print references, upon which the subdiscipline of sigillography has long depended, available in an online form. In doing so, DIGISIG shows these references to be more than the sum of their parts because it invites both extensive cross-referencing and sophisticated hierarchical classification schemes. McEwan also indirectly highlights a more general problem: that of recording, organizing, and searching for information not easily represented in standard character formats such as Unicode text. In this respect sigillography has much in common with the only partially solved problem of digital paleography. In chapter 3, Rowan Doran presents a clear narrative of the life-cycle of the Corpus Synodalium project and highlights the importance of a good set of explicit fundamental decisions in guiding digital humanities projects—noting, in this case, that the project should be "designed for permanence" and "driven by research questions'' (55) and should not "replicate the existing unevenness in availability of sources" (56). In the case of the Corpus Synodalium project, design for permanence entailed a clear separation between format and platform commitments: plain-text data and comma-separated metadata, on the one hand; the ARTFL Project's PhilLogic4 on the other. (Doran's contention that capturing information in the simplest formats possible, such as .txt and .csv, is a key to long-term project data survivability is also a theme of Davis's closing chapter.) Doran is forthright in acknowledging that his project enjoyed an unusual level of institutional support, so...

An artificial intelligence platform provides an accurate interpretation of esophageal motility from Functional Lumen Imaging Probe Panometry studies.

Functional lumen imaging probe (FLIP) Panometry is performed at the time of sedated endoscopy and evaluates esophageal motility in response to distension. This study aimed to develop and test an automated artificial intelligence (AI) platform that could interpret FLIP Panometry studies. The study cohort included 678 consecutive patients and 35 asymptomatic controls that completed FLIP Panometry during endoscopy and high-resolution manometry (HRM). "True" study labels for model training and testing were assigned by experienced esophagologists per a hierarchical classification scheme. The supervised, deep learning, AI model generated FLIP Panometry heatmaps from raw FLIP data and based on convolutional neural networks assigned esophageal motility labels using a two-stage prediction model. Model performance was tested on a 15% held-out test set (n=103); the remainder of the studies were utilized for model training (n=610). "True" FLIP labels across the entire cohort included 190 (27%) "normal," 265 (37%) "not normal/not achalasia," and 258 (36%) "achalasia." On the test set, both the Normal/Not normal and the achalasia/not achalasia models achieved an accuracy of 89% (with 89%/88% recall, 90%/89% precision, respectively). Of 28 patients with achalasia (per HRM) in the test set, 0 were predicted as "normal" and 93% as "achalasia" by the AI model. An AI platform provided accurate interpretation of FLIP Panometry esophageal motility studies from a single center compared with the impression of experienced FLIP Panometry interpreters. This platform may provide useful clinical decision support for esophageal motility diagnosis from FLIP Panometry studies performed at the time of endoscopy.

Simultaneous Fault Diagnosis Based on Hierarchical Multi-Label Classification and Sparse Bayesian Extreme Learning Machine

This paper proposes an intelligent simultaneous fault diagnosis model based on a hierarchical multi-label classification strategy and sparse Bayesian extreme learning machine. The intelligent diagnosis model compares the similarity between an unknown sample to be diagnosed and each single fault mode, then outputs the probability of each fault mode occurring. First, multiple two-class sub-classifiers based on SBELM are trained by using single-fault samples to extract the correlation between various pairs of single-fault, and the sub-classifiers are integrated with the proposed hierarchical multi-label classification (HMLC) strategy to form the diagnostic model based on HMLC-SBELM. Then, samples of single faults and simultaneous faults are used to generate the optimal discriminative thresholds by using optimization algorithms. Finally, the probabilistic output generated by the HMLC-SBELM-based model is transformed into the final fault modes by using the optimal discriminative threshold. The model performance is evaluated by using actual vibration signals of the main reducer and is compared with several classical models. The contrastive results indicate that the proposed model is more accurate, efficient, and stable.

Forensic Classification of Subjects Counteracting Investigation of High-Tech Crimes

The paper defines the concept of counteraction to the investigation of high-tech crimes employing the author’s definition of high-tech crimes. A forensic classification of the subjects of this counteraction is proposed, in which the procedural status of these subjects is used as a formal classification criterion. A hierarchical classification scheme is described, which includes groups and subgroups of counteraction subjects of varying degrees of similarity. The subjects of «internal» opposition include participants in criminal proceedings, divided into four main groups: on the part of the defense; by the prosecution; administering justice; other subjects. For the subjects of «external» opposition, the following groups are identified: law enforcement officers who are not subjects of criminal proceedings in a specific criminal case; persons associated with the accused (defendants); heads of government bodies; media workers. For each group of subjects, forensically significant features of counteraction to the investigation of high-tech crimes are analyzed, typical motives and methods of counteraction are given. It is noted that the proposed classification scheme can be used not only for high-tech crimes, but also for other types of criminal acts.

SCovNet: A skip connection-based feature union deep learning technique with statistical approach analysis for the detection of COVID-19

Background and ObjectiveThe global population has been heavily impacted by the COVID-19 pandemic of coronavirus. Infections are spreading quickly around the world, and new spikes (Delta, Delta Plus, and Omicron) are still being made. The real-time reverse transcription-polymerase chain reaction (RT-PCR) is the method most often used to find viral RNA in a nasopharyngeal swab. However, these diagnostic approaches require human involvement and consume more time per prediction. Moreover, the existing conventional test mainly suffers from false negatives, so there is a chance for the virus to spread quickly. Therefore, a rapid and early diagnosis of COVID-19 patients is needed to overcome these problems. MethodsExisting approaches based on deep learning for COVID detection are suffering from unbalanced datasets, poor performance, and gradient vanishing problems. A customized skip connection-based network with a feature union approach has been developed in this work to overcome some of the issues mentioned above. Gradient information from chest X-ray (CXR) images to subsequent layers is bypassed through skip connections. In the script’s title, “SCovNet” refers to a skip-connection-based feature union network for detecting COVID-19 in a short notation. The performance of the proposed model was tested with two publicly available CXR image databases, including balanced and unbalanced datasets. ResultsA modified skip connection-based CNN model was suggested for a small unbalanced dataset (Kaggle) and achieved remarkable performance. In addition, the proposed model was also tested with a large GitHub database of CXR images and obtained an overall best accuracy of 98.67% with an impressive low false-negative rate of 0.0074. ConclusionsThe results of the experiments show that the proposed method works better than current methods at finding early signs of COVID-19. As an additional point of interest, we must mention the innovative hierarchical classification strategy provided for this work, which considered both balanced and unbalanced datasets to get the best COVID-19 identification rate.

Hierarchical, multiclass prediction of Alzheimer’s clinical diagnosis using imputed, multimodal NACC data

AbstractBackgroundEarly detection and diagnosis are critical for effective treatment of Alzheimer’s disease (AD). Statistical machine learning holds promise for disease prediction using large AD repositories, such as the National Alzheimer’s Coordinating Center (NACC) dataset. Past work has extensively explored binary classifications of AD diagnosis. Here, we extended this to address two major, remaining challenges: data imputation and multiclass classification (e.g., AD vs Mild Cognitive Impairment, MCI vs. Healthy Cognition, HC).MethodFor imputation, we tested single‐value (baseline), k‐NN, and Bayesian imputation on a subset of complete training data with artificially removed values (completely at random). Per modality of interest, a systematic search was conducted for the best input feature set to impute the missing values. The mean absolute deviation and mismatch rate between the imputed and true values were used to evaluate the methods on a held‐out test set. For classification, we then implemented both “flat” multiclass and hierarchical classification, an extension of multiclass classification that organizes classes into a tree, to predict clinical diagnosis using multimodal inputs. Several pipelines with combinations of imputation methods, classifiers, and input modalities were tested on different hierarchical classification strategies, e.g., HC vs [AD vs MCI], compared with the flat, multiclass baseline (AD vs MCI vs HC).ResultCompared to the baseline methods for both imputation and classification, the proposed alternatives performed better, as measured by the predetermined imputation and classification metrics evaluated on the test set. For imputation, behavioral and neuropsychiatric features were more easily and accurately imputed while ApoE genotype was more difficult to impute. For classification, using a combination of classifiers and input features in the hierarchy outperformed the baseline method and support the notion that select biomarkers are better able to characterize AD subpopulations within the hierarchy.ConclusionEstablishing a systematic imputation method tailored to AD biomarker modalities fills a critical gap in achieving more robust disease prediction. Building a hierarchical classifier parallels working through a differential diagnosis for AD, which offers greater translational value than binary classification of AD. In combination, these results form an important step toward developing reliable, holistic classifiers of AD disease status.