116320 research outputs found
Sort by
A systematic review for artificial intelligence-driven assistive technologies to support children with neurodevelopmental disorders
Get PDFThis systematic review examines AI-powered assistive technologies for children with neurodevelopmental disorders, with a focus on dyslexia (DYS), attention-deficit hyperactivity disorder (ADHD), and autism spectrum disorder (ASD). Our analysis of 84 studies from 2018 to 2024 provides the first thorough cross-disorder comparison of AI implementation patterns. According to our data, each condition has different success rates and technological preferences. AI applications are expanding quickly, especially in research on ASD (56 % of studies), followed by ADHD (36 %), and DYS (8 %). In almost half of the reviewed studies, computer-assisted technologies, which have demonstrated encouraging results in terms of treatment support and diagnostic accuracy, became the main mode of intervention. Despite high accuracy in controlled settings, the implementation of these technologies in clinical practice faces significant challenges. While human oversight remains essential in clinical applications, future advancements should prioritize privacy protection and the ability to assess tools longitudinally. Notably, multimodal approaches that integrate various data types have improved diagnostic accuracy; recent research has shown that they can detect ASD with up to 99.8 % accuracy and ADHD with up to 97.4 % accuracy. A promising trend is the combination of mobile applications and wearable technology, especially for real-time monitoring and intervention. This review highlights the potential and current limitations of AI-driven tools in supporting children with neurodevelopmental disorders. Future development should focus not on replacing clinical expertise, but on augmenting it. Research efforts should aim at creating tools that enhance professional judgment while preserving the essential human components of assessment and intervention
Intersecting SARS-CoV-2 spike mutations and global vaccine efficacy against COVID-19.
Get PDFIn line with encountering the world with the emergence of vaccine-resistance variants of SARS-CoV-2, 15,669,529 samples that received COVID-19 vaccines until April 2023 were investigated as two doses in the first phase and booster vaccinations in the second phase. The analysis shows that D614G and P681 mutations occurred in both phases. The E484 and Y655 mutations significantly emerged during the second phase. The 762-889 and 254-381 regions are revealed as conserved parts and could be considered in vaccine design. The Kruskal-Wallis test revealed a significant reduction in single mutations between populations with 20%-50% and those with 70%-100% vaccination coverage (p=0.017). The Mann-Whitney U test proposes a link between vaccination and suppression of viral mutation rates. Dynamic modeling suggests that key mutations have facilitated the virus' evolution and immune escape. The study's findings are crucial for understanding virus genome mutations, especially E614 and P681 in Delta and E484 and H655 in Omicron. This highlights the need to adjust strategies and strengthen global efforts in combating the pandemic
Enhancing trust in news media: A multimodality approach to detecting fake news with social constructs.
Get PDFThe widespread dissemination of misinformation and fake news on social media platforms poses a significant challenge to the integrity of public discourse, democratic stability, and societal trust. Although misinformation detection has been widely studied, one critical but underexplored dimension is the misquotation of legitimate news articles on social media. Misquotations-deliberate or accidental distortions of original news content-can mislead audiences and diminish trust in credible news sources, thereby amplifying the reach of misinformation. Addressing this issue requires a robust dataset that captures the original news content and its associated social media representations. To this end, we construct a comprehensive dataset by performing multi-source triangulation across four established datasets: FAKENEWSNET, NELA-GT, TruthSeekers, and Twitter15/16. This process yields approximately 158,400 aligned pairs of news stories and related social media posts. Building on this dataset, we propose a multimodal binary classification framework designed to detect misquotations by jointly modeling textual, visual, and contextual features. The model integrates engineered features representing social context and event semantics within a shared latent space, enabling a nuanced understanding of content distortion. Furthermore, we analyze the relative contribution of each modality-textual, visual, and contextual-using ablation studies and performance metrics to assess their impact on detection accuracy. This study introduces a novel approach to misquotation detection, contributing to the combat of misinformation through multimodal analysis
Handwriting Intra-Variability Across Surface Transitions: Implications for Writer Identification
Get PDFHandwriting exhibits intra-variability within the same writer due to friction between different writing surfaces, such as transitioning from paper to a computer tablet. This study investigates such intra-variability in handwriting characteristics across different writing surfaces and its implications for writer identification. An empirical study is conducted to assess the performance of state-of-the-art deep architectures in identifying writers amidst such intra-variation. Additionally, a transformer-based model is proposed to capture writer identification under these intra-variable circumstances. A dataset comprising 1560 handwritten English text-line images from 130 writers is created and utilized for experimentation. The results reveal insightful outcomes regarding the utilization of deep architectures and the proposed model in handling intra-variability for writer identification. This study contributes to advancing the understanding of intra-variability in handwriting and offers practical implications for forensic analysis and document authentication in the digital age
Gaze–Hand Steering for Travel and Multitasking in Virtual Environments
Get PDFAs head-mounted displays (HMDs) with eye tracking become increasingly accessible, the need for effective gaze-based interfaces in virtual reality (VR) grows. Traditional gaze- or hand-based navigation often limits user precision or impairs free viewing, making multitasking difficult. We present a gaze–hand steering technique that combines eye tracking with hand pointing: users steer only when gaze aligns with a hand-defined target, reducing unintended actions and enabling free look. Speed is controlled via either a joystick or a waist-level speed circle. We evaluated our method in a user study (n = 20) across multitasking and single-task scenarios, comparing it to a similar technique. Results show that gaze–hand steering maintains performance and enhances user comfort and spatial awareness during multitasking. Our findings support using gaze–hand steering in gaze-dominant VR applications requiring precision and simultaneous interaction. Our method significantly improves VR navigation in gaze–dominant, multitasking-intensive applications, supporting immersion and efficient control
FIT-RAG: Black-Box RAG with Factual Information and Token Reduction
No full textDue to the extraordinarily large number of parameters, fine-tuning large language models (LLMs) to update long-tail or out-of-date knowledge is impractical in lots of applications. To avoid fine-tuning, we can alternatively treat a LLM as a black-box (i.e., freeze the parameters of the LLM) and augment it with a retrieval-augmented generation (RAG) system, namely black-box RAG. Recently, black-box RAG has achieved success in knowledge-intensive tasks and has gained much attention. Existing black-box RAG methods typically fine-tune the retriever to cater to LLMs' preferences and concatenate all the retrieved documents as the input, which suffers from two issues: (1) Ignorance of Factual Information. The LLM preferred documents may not contain the factual information for the given question, which can mislead the retriever and hurt the effectiveness of black-box RAG; (2) Waste of Tokens. Simply concatenating all the retrieved documents brings large amounts of unnecessary tokens for LLMs, which degenerates the efficiency of black-box RAG. To address these issues, this article proposes a novel black-box RAG framework which utilizes the factual information in the retrieval and reduces the number of tokens for augmentation, dubbed FIT-RAG. FIT-RAG utilizes the factual information by constructing a bi-label document scorer which takes the factual information and LLMs' preferences as labels respectively. Besides, it reduces the tokens by introducing a self-knowledge recognizer and a sub-document-level token reducer, which enables FIT-RAG to avoid unnecessary augmentation and reduce augmentation tokens as much as possible. FIT-RAG achieves both superior effectiveness and efficiency, which is validated by extensive experiments across three open-domain question-answering datasets: TriviaQA, NQ, and PopQA. FIT-RAG can improve the answering accuracy of Llama2-13B-Chat by 14.3% on TriviaQA, 19.9% on NQ and 27.5% on PopQA, respectively. Furthermore, it can save approximately half of the tokens on average across the three datasets
Illumina complete long read assay yields contiguous bacterial genomes from human gut metagenomes.
Get PDFMetagenomics enables direct investigation of the gene content and potential functions of gut bacteria without isolation and culture. However, metagenome-assembled genomes are often incomplete and have low contiguity due to challenges in assembling repeated genomic elements. Long-read sequencing approaches have successfully yielded circular bacterial genomes directly from metagenomes, but these approaches require high DNA input and can have high error rates. Illumina has recently launched the Illumina Complete Long Read (ICLR) assay, a new approach for generating kilobase-scale reads with low DNA input requirements and high accuracy. Here, we evaluate the performance of ICLR sequencing for gut metagenomics for the first time. We sequenced a microbial mock community and 10 human gut microbiome samples with standard, shotgun 2 × 150 paired-end sequencing, ICLR sequencing, and nanopore long-read sequencing and compared performance in read lengths, assembly contiguity, and bin quality. We find that ICLR human metagenomic assemblies have higher N50 (119.5 ± 24.8 kilobases) than short read assemblies (9.9 ± 4.5 kilobases; P = 0.002), and comparable N50 to nanopore assemblies (91.0 ± 43.8 kilobases; P = 0.32). Additionally, we find that ICLR draft microbial genomes are more complete (94.0% ± 20.6%) than nanopore draft genomes (85.9% ± 23.0%; P ≤ 0.001), and that nanopore draft genomes have truncated gene lengths (924.6 ± 114.7 base pairs) relative to ICLR genomes (954.6 ± 71.5 base pairs; P ≤ 0.001). Overall, we find that ICLR sequencing is a promising method for the accurate assembly of microbial genomes from gut metagenomes.IMPORTANCEMetagenomic sequencing allows scientists to directly measure the genome content and structure of microbes residing in complex microbial communities. Traditional short-read metagenomic sequencing methods often yield fragmented genomes, whereas advanced long-read sequencing methods improve genome assembly quality but often suffer from high error rates and are logistically limited due to high input requirements. A new method, the Illumina Complete Long Read (ICLR) assay, is capable of generating highly accurate kilobase-scale sequencing reads with minimal input material. To evaluate the utility of ICLR in metagenomic contexts, we applied short-read, long-read, and ICLR methods to simple and complex microbial communities. We found that ICLR outperforms short-read methods and yields comparable metagenomic assemblies to standard long-read approaches while requiring less input material. Overall, ICLR represents an additional option for assembling complete genomes from complex metagenomes
Designing an Ontology for a Smart Learning Health System Framework.
Get PDFThe concept of a "Learning Health System Framework" has been defined in various forms, making it difficult for healthcare leaders to adopt an appropriate framework for organisational transformation. There is a pressing need to standardise its components and relationships. Limitations of the existing LHS frameworks include their lack of flexibility to adapt to sociotechnical changes and inability to guide the automatic evaluation initiative on health organisational performance. Ontologies facilitate consensus by aligning stakeholders on a shared vocabulary and structure, which can reduce ambiguity in communication and foster better collaboration across different stakeholder groups. Therefore, this study designed a Smart Learning Health System Ontology, abbreviated as "SMARTLHS". We followed a four-step realist synthesis methodology to develop SMARTLHS: (1) ontology requirements specification, (2) iterative and inductive ontology conceptualisation, (3) concept and relationship comparison and development, (4) ontology evaluation and refinement. The resultant SMARTLHS ontology comprises 475 unique classes and subclasses (concepts) and 135 object properties. The object properties define the relationships between these classes. SMARTLHS addresses the critical limitations of the existing LHS frameworks. It serves as a knowledge base for healthcare professionals. It can also be integrated into Retrieval Augmented Generative (RAG) applications and feed into large language models such as GPT 4o for advanced exploration of healthcare organisations' performance in developing learning health systems
Improved atom probe specimen preparation by focused ion beam with the aid of multi-dimensional specimen control
Get PDFFocused ion beam lift-out has become an essential technique for fabricating small-scale specimens in atom probe tomography (APT). By using a rotatable micromanipulator, we developed methods that can precisely extract the regions of interest for APT samples with challenging-to-prepare geometries. Combining this function with pre-milling and pre-tilt operations, we prepare three typically challenging APT specimens: nanoparticles, nanowires, and thin films. This combination can effectively decrease the sample preparation time and increase the accuracy of focused ion beam lift-out specimen preparation
Negatively correlated ensemble against transfer adversarial attacks
No full textDeep neural networks (DNNs) have become a popular tool in various fields, but their susceptibility to adversarial attacks poses a significant threat to the security of machine learning systems. Adversarial examples, which are intentionally crafted inputs to deceive a DNN, can easily transfer between different machine learning models, further exacerbating the problem. To address the transfer adversarial vulnerability issue, ensemble methods have been developed to promote diversity among members, thereby impeding the transfer of adversarial examples among them. However, we observed that the diversity among ensemble members tends to diminish rapidly as the magnitude of adversarial perturbations increases. All ensemble members tend to be deceived by the same adversarial example, resulting in a poor performance against slightly larger transfer adversarial perturbations. To overcome this challenge, we introduce NCRE (negative correlation robust ensemble) and NCRE+(negative correlation robust ensemble on adversarial vulnerabilities) in this work. Our approach leverages the explicit maximization of negative correlation among ensemble member outputs to enhance ensemble diversity and robustness against adversarial perturbations, particularly transfer black-box attacks. Extensive experimental studies and comparisons with state-of-the-art algorithms demonstrate the effectiveness of our approach
Data sources:
Access Repository Dashboard
Do you manage OPUS? Access insider analytics, issue reports and manage access to outputs from your repository in the CORE Repository Dashboard!