This study aimed to investigate the effectiveness of the Cogmed computer-assisted cognitive rehabilitation program and training for reading in improving reading skills in children with linguistic dyslexia. A randomized, controlled clinical trial was used. Thirty children with linguistic dyslexia were assigned to experimental group (n = 15) or control group (n = 15). Two participants (one from each group) dropped out during the intervention, resulting in a final sample of 28 participants (n = 14 per group. The experimental group received 25 sessions of reading training and the Cogmed cognitive rehabilitation intervention, while the control group only received reading training. Data were collected using the NEMA Reading and Dyslexia Test, the Specific Learning Disorder Questionnaire, Raven's Progressive Matrices Test, and the Assessment of Persian Reading Ability. The findings indicated that the combined Cogmed and reading training intervention significantly improved multiple reading components, including low-frequency word reading (F = 56.34, p < 0.001, η2 = 0.68), high-frequency word reading (F = 29.62, p < 0.001, η2 = 0.53), word chains (F = 39.76, p = 0.001, η2 = 0.60), and word comprehension (F = 49.19, p = 0.001, η2 = 0.65). The combined Cogmed and reading training intervention produced statistically significant improvements across multiple reading subskills in children with linguistic dyslexia. These improvements remained stable over a one-month follow-up period, indicating short-to-medium-term durable cognitive and linguistic gains. Future research should examine long-term maintenance, test efficacy in other dyslexia subtypes, and utilize executive function (EFs) assessments along with neuroimaging techniques to explore the underlying mechanisms of change.
Visual attention is essential for processing not only visual objects and scenes but also written words and continuous text. Yet, its precise contribution to different levels of reading remains unclear. Using a meta-analytic approach across 140 fMRI experiments (n = 1,541), we examined the interplay between bottom-up and top-down attention, word and text reading. Our results reveal a layered functional organization within the middle precentral gyrus: the superior part shows convergence between top-down attention and text reading, the inferior part links top-down attention with word reading, and the central area, which overlaps with area 55b, shows selective engagement in reading. In addition, the fusiform gyrus, including the VWFA, emerges as a region of spatial convergence between visual attention and word reading. These findings extend the neuronal recycling framework by showing that reading builds on visual and attentional circuits, while also revealing linguistic-related regions that are partly distinct from domain-general attentional systems.
The home literacy environment, particularly shared reading, plays a critical role in preschool children's cognitive and socioemotional development. However, its associations with emotional and behavioral problems remain underexplored in large-scale studies. This study examined the relationship between shared reading and emotional/behavioral problems as well as prosocial behavior in preschool children. A cross-sectional study was conducted using stratified cluster sampling across 189 kindergartens in a major city in western China. A total of 21,366 parent-child pairs were included. Shared reading was assessed with the reading subscale of the StimQ-P (score range 0-22), which evaluates quantity, diversity of concepts and content, and interactivity quality. Emotional and behavioral problems were measured using the parent-reported Strengths and Difficulties Questionnaire (SDQ). Multivariate logistic regression and generalized additive models were employed to examine associations, adjusting for child age, gender, parental socioeconomic factors, lifestyle variables, and parental mental health (CES-D). Higher shared reading scores were significantly associated with lower odds of emotional/behavioral problems (adjusted OR = 0.96 per point increase, 95% CI: 0.95-0.97, P < 0.0001) and higher odds of adequate prosocial behavior (adjusted OR = 1.09, 95% CI: 1.08-1.10, P < 0.0001) in fully adjusted models. All four dimensions of shared reading showed independent associations. Nonlinear analyses revealed threshold effects, with associations becoming stronger above approximately 18 points for total difficulties and 15 points for prosocial behavior. These associations were largely consistent across subgroups after correction for multiple testing. In this large cross-sectional study conducted in western China, higher levels of shared reading were associated with lower odds of emotional/behavioral problems and higher odds of prosocial behaviors among preschool children. The results suggest possible threshold patterns in these associations. However, given the cross-sectional nature of the study, causality cannot be established.
Reading has been proposed as a protective factor in mental health; however, evaluating this is challenging due to a lack of trials and the possibility of confounding in observational studies. We used the complementary approaches of covariate balancing propensity score weighting and random intercepts cross-lagged panel models in the large longitudinal Zurich Project on Social Development from Childhood to Adulthood study. Outcomes were measured from age 13 to 20 for anxiety and depression and at age 20 for psychosis-like symptoms. After accounting for potential confounding there were no significant effects of reading frequency on adolescent mental health across either approach. Increasing reading frequency without considering motivation, content and style of reading engagement is not a promising protective factor in adolescent mental health.
Neurotypical people are generally quite adept at interpreting social signals from dynamic bodies and faces. This ability prevents one from incurring high costs associated with ineffective and maladaptive social interactions. Individuals with mental disorders, such as schizophrenia (SZ), often exhibit deficits in nonverbal social cognition. It remains unclear whether reading the language of bodies and faces is affected by SZ, and if this is indeed the case, how these potential deficits are related to one another. In the present study, participants (28 males with SZ and 28 typically developing, TD, matched controls) were administered face-to-face computer tasks on inferring emotions from dynamic point-light body motion and faces. The outcome indicates that SZ patients exhibit global impairments in both body and face reading, albeit patients demonstrate a similar emotion recognition profile as TD controls. In SZ patients only, a positive link was found between accuracy of recognizing emotions expressed through faces and bodies, whereas processing speed of emotions conveyed through bodies and faces was tied to each other in both SZ and TD individuals. For SZ patients, inferring social signals from dynamic faces and bodies may be rather challenging in terms of neurocognitive mechanisms, which is reflected in the tight link in recognition accuracy. Along with previous data on inferring emotions in the eyes collected in the same cohort, this work provides novel insights into the specific global aberrations in social cognition in SZ and offers a blueprint for the development of strategies for the targeted treatment of gender-specific mental disorders.
Multimodal large language models (MLLMs) demonstrate significant limitations in visual Theory of Mind (ToM) abilities compared to humans. Investigating the neural mechanisms underlying human mind-reading not only addresses critical gaps in visual ToM research but also provides valuable insights for MLLM optimization. Using fMRI data from 83 participants, we systematically compared neural processing patterns between two conditions in visual ToM tasks: MLLM-incorrect/human-correct (MLLMI) and mutually correct (MLLMC). The questionnaire results indicated that, compared with the MLLMC condition, human confidence was significantly lower under the MLLMI condition, and expectations for MLLM performance were also lower. Natural language analysis revealed that, relative to MLLM responses, human responses were more closely aligned with the question context, more concise, and exhibited greater certainty. Neuroimaging results indicated significantly stronger activation in bilateral precuneus and middle temporal gyrus in MLLMI. Furthermore, we observed enhanced functional connectivity in networks associated with task coordination and attention allocation. Leveraging these neural signatures, we constructed multiple prediction models for decoding the two conditions (MLLMI vs. MLLMC), among which the 2-layer Transformer model achieved the highest classification accuracy of 78.6%. Extending these findings, we propose the Knowledge-Thinking-Adaptation (KTA) framework, which integrates memory retrieval, divergent thinking, and multi-level attention mechanisms to provide a potential roadmap for future work in developing AI systems with human-like visual ToM capabilities.
A new AI agent mines both text and figures from lithium-metal-battery papers, pointing toward a future where the literature itself helps guide discovery.
暂无摘要（点击查看详情）
Tokenization is both a prerequisite and a central challenge for genomic language models, due to the inherent difficulty of delineating meaningful segments within continuous DNA sequences. We present GenART, a genomic language model framework that dynamically segments DNA into variable-length words directly from raw sequences without manual annotation. Its key adaptive tokenization module infers biologically meaningful boundaries from multi-scale contextual signals during pretraining. GenART demonstrates competitive performance across diverse tasks, outperforming leading approaches. Word boundaries derived from unsupervised tokenization show strong correspondence with base-resolution regulatory signals, such as DNA methylation, and with multi-nucleotide functional elements annotated in GENCODE v.49. These findings highlight adaptive tokenization as a powerful strategy for building interpretable and biologically responsive genomic language models.
This longitudinal study investigated subtypes of dyslexia based on the distinction between word reading accuracy and word reading rate. Building on prior cross-sectional research, we inquired whether the cognitive profiles of these subtypes can be identified before children learn to read. A large sample of Hebrew-speaking children were assessed on phonological awareness (PA), morphological awareness (MA), and rapid automatized naming (RAN) in preschool and then followed into first grade to evaluate their reading accuracy and rate. In preschool, findings revealed two distinct subgroups: one with selective deficits in PA and MA but intact RAN (PA + MA-disabled) and another with impaired RAN but preserved PA and MA (RAN-only disabled). In first grade, as predicted, the PA + MA-disabled group demonstrated significantly lower reading accuracy compared to both the RAN-only disabled and control groups. The RAN-only subgroup exhibited slow reading but intact accuracy. However, contrary to predictions, the PA + MA-disabled group also exhibited slower reading rates. These results suggest that at the beginning of reading development, low levels of reading accuracy limit reading rate. Our study (i) supports accuracy-rate subtyping of dyslexia, (ii) reinforces the role of both PA and MA in achieving early word reading accuracy, and (iii) highlights the existence of a highly specific RAN-rate dyslexia subtype. These findings have significant implications for early diagnosis and intervention as well as the definition of dyslexia.
This study describes the development and validation of ROAR Palabra, a novel Spanish lexical decision task designed for use with both Spanish-speaking children and Spanish-English bilinguals. This self-administered task requires students to decide whether a string of letters presented on the screen is a real word in Spanish. While there is evidence that scores on English lexical decision tasks are highly predictive of performance on conventional (time- and resource-intensive) word reading assessments in English, we explore whether this holds in Spanish, which has a much more transparent orthography. The specific goals are (i) to create a linguistically fair task using item-response theory and (ii) to evaluate whether such task can serve as a reliable proxy for conventional word reading measures, offering a quick and easy-to-administer tool for assessing reading skills across linguistic and cultural contexts. Results demonstrated strong correlations between performance on ROAR Palabra and standardized word reading assessments such as the Woodcock-Muñoz Batería IV, suggesting its effectiveness as a substitute measure. Notably, the task was sensitive to differences in language proficiency across both monolingual and multilingual groups, reflecting expected developmental and environmental influences. While not specifically designed for the comparisons between monolingual and multilingual populations, the findings underscore the potential of this task as a versatile and culturally adaptable tool for reading assessments in different Spanish-speaking and bilingual contexts.
Families often report searching the internet for guidance on how best to support children when a significant adult has cancer. This study aimed to identify and evaluate the quality, reliability, readability and content of websites, videos, and artificial intelligence (AI) resources available to adults with cancer who have caregiving responsibilities for children. Online platforms were searched using 10 phrases across Google web, YouTube, TikTok and four AI platforms. The mDISCERN instrument assessed reliability and quality, GQS assessed overall quality, and the NHS Medical Document Readability Tool assessed readability. Quantitative differences between sources were determined using pairwise analysis. Google web had significantly higher quality and reliability compared with AI and TikTok sources, with mean mDISCERN and GQS scores of 3.74 and 3.72, respectively. AI-generated resources showed lower mean mDISCERN and GQS scores of 2.77 (P < .05) and 2.32 (P < .05), respectively. TikTok videos had lower scores of 2.73 (P < .05) for mDISCERN and 2.49 (P < .05) for GQS. Estimated reading time was significantly longer (P < .05) for Google web (11:45mins) compared to AI (02:09mins). However, reading age did not differ (P = .31) at 15.09 years and 15.17 years respectively. There was a lack of accessible and inclusive resources for non-nuclear families, adults with neurodivergent children, culturally and ethnically diverse populations and families at end of life. Although Google web resources demonstrated higher overall quality and reliability, written resources across platforms often exceeded recommended reading levels, which may represent a significant health equity concern for individuals with lower health literacy and families experiencing deprivation. AI presents an opportunity whereby a single high-quality and evidence-based resource can be rapidly adapted into multiple formats, reading levels, languages and be culturally relevant. Future resources may benefit from co-production using a trusted, regulated, and centralised information hub, with supportive collaboration between health and social care professionals and technology providers.
Ophthalmological reports are often written at a complexity level that exceeds the reading ability of many patients. Large language models (LLMs) may help simplify these texts, but their performance depends on prompt design and must preserve clinical fidelity. This study evaluated whether different prompting strategies improve the readability and safety of simplified ophthalmological reports. We analyzed 443 de-identified reports from a tele-ophthalmology platform, including 280 retinal fundus and 163 ocular ultrasound reports. Each report was simplified by four LLMs (GPT-3.5, GPT-4.0, Gemini, and Copilot) using three prompts: a basic command, a patient-oriented prompt, and a targeted 7th-grade reading level prompt. Readability was measured with the average Readability Grade Level (aRGL). A representative eligible subset of the generated texts was independently evaluated by two third-year ophthalmology residents, and each selected response underwent independent review by both evaluators, with disagreements resolved by a board-certified retina ophthalmologist. Outcomes included factual accuracy, information completeness, and potential for harm. The original reports were highly complex, with a median aRGL of 13.8 overall, 11.6 for fundus reports, and 15.4 for ocular ultrasound reports. All LLMs improved readability scores to varying degrees. Prompt engineering was a major determinant of performance, and the targeted 7th-grade prompt produced the best results across models. Copilot, Gemini, and GPT-4.0 achieved median aRGL values closest to the recommended patient-facing reading level, while GPT-3.5 showed weaker performance in some comparisons. Clinical validation showed substantial agreement between raters (kappa range, 0.70-0.97). After adjudication, 91% of simplified texts were factually accurate, 86% retained all critical information, and 95% showed no potential for harm. LLMs can simplify ophthalmological reports while preserving clinical fidelity, but performance depends strongly on prompt specificity. These tools show promise for patient-facing communication, although human oversight remains essential.
Professional society patient education materials frequently exceed recommended literacy levels, limiting equitable health information access. This study aimed to compare the readability, information quality, understandability, and actionability of artificial intelligence (AI)-generated patient education materials versus professional society materials across gastroenterology, surgery, ophthalmology, and anesthesiology. We conducted a cross-sectional comparative analysis of 100 paired topics (25 per specialty), comparing professional society materials with the responses generated by ChatGPT (OpenAI, San Francisco, California, United States) under standardized conditions. Readability was assessed using the Flesch-Kincaid grade level, information quality with DISCERN, and understandability and actionability with the Patient Education Materials Assessment Tool (PEMAT). Paired two-sided t-tests assessed within-specialty differences. In surgery, AI-generated materials had lower reading levels and higher quality, understandability, and actionability (all p<0.001). In anesthesiology, AI materials were more readable (p<0.001) with no differences in other measures. In ophthalmology, AI improved readability (p<0.001), while professional society materials had higher quality and understandability (p<0.01) with no difference in actionability. In gastroenterology, AI materials had higher reading levels (p<0.001) with no differences in quality or usability. The performance of AI-generated patient education materials varied by specialty and appeared to depend on the structure and complexity of clinical content. AI improved readability in several domains, but these gains were not uniform across specialties, particularly in areas requiring more complex or longitudinal explanations.
Titanium and its alloys are widely used in modern medical devices because of their favorable biocompatibility and mechanical properties. Allergic reactions to titanium-based implants are considered rare, and their diagnosis is hampered by the poor dermal penetration of metallic allergens. A 29-year-old female patient with seronegative myasthenia gravis requiring parenteral nutrition presented with persistent inflammatory dermatitis overlying a titanium port catheter (X-Port BARD Titan low-profile) 4 weeks after implantation in July 2023. Despite initial diagnostic uncertainty and sequential management modifications, her condition continued to deteriorate. Similar eruptions developed after re-implantation of the same titanium port model on the contralateral thorax. Rigorous patch testing incorporating standardized tape-stripping and extended reading intervals revealed positive sensitization to the titanium port body, while all other port components and an alternative plastic port system (BARD SlimPort M.R.I. Ultra Low-Profile) tested negative. Energy-dispersive X-ray spectroscopy confirmed the titanium body composition as a Ti-Al6-V4 alloy. Substitution with the plastic port catheter in February 2025 resulted in complete symptom resolution, with sustained clinical remission through July 2025. Although very uncommon, contact sensitization to titanium medical devices can occur and may present significant diagnostic and therapeutic challenges. Rigorous patch testing methodology, incorporating standardized tape-stripping and extended 168-hour readings, appears essential for detecting delayed-type hypersensitivity to metallic implants. Clinical resolution following biocompatible device substitution supports an allergic etiology in this case.
Radiology reports are often written for clinicians and contain complex medical terminology, which patients struggle to understand. This comprehension gap may lead to anxiety and misinformed decisions. This is especially important in cancer care. Large language models (LLMs) provide an opportunity to bridge this gap. Direct use of LLM by patients may potentially be misleading and may not ensure clinical fidelity. We developed an LLM- based tool to automatically translate radiology findings into clear, patient-friendly language which could improve patient-centered care. To develop and validate a bilingual LLM driven tool that simplifies oncology radiology reports into patient-understandable English and Hindi, while maintaining diagnostic fidelity and emotional tone. This study was approved by the Institute's ethics committee. A retrospective corpus of 100 computed tomography (CT) reports (April 2025-July 2025) of patients with colo-rectal cancers were used for the development of the pipeline. Five large language models-GPT-4o, Gemini 2.5 Pro, Claude Opus, LLaMA-3.1-8B and Phi-3.5-mini-were tested. Five iterative prompt versions guided LLMs through successive refinements to ensure medical accuracy, clarity, and inclusion of a standard disclaimer. A custom pipeline; Vernacular Language Coverter(VLC), based on Gemini 2.5 Pro that takes a radiologist's report as input and generates two patient-facing outputs: (1) a simplified English and (2) a Hindi explanation was developed. The tool thus developed was then prospectively validated using 100 de-identified reports. Simplified outputs were reviewed by two radiologists, assessing accuracy, language clarity/Terminology and readability/tone. Completeness was assessed in terms of core diagnostic completeness and minor incompleteness. Flesch Reading Ease (FRE) was calculated for a fraction of reports. Mean rubric scores were high: Accuracy 4.77 ± 0.62, language clarity/Terminology 4.78 ± 0.56 and readability/tone 4.9 ± 0.32. Core diagnostic completeness was attained in all patients. 92% percent of reports were released 'as is. Readability improved markedly (Flesch Reading Ease 49.2→73). Empathy phrases in English and Hindi were appropriate. An AI-driven bilingual framework significantly enhances the clarity, tone, and readability of oncology radiology reports while retaining diagnostic precision. This tool demonstrates the feasibility of safe, patient-centered communication aligned with the goals of personalized medicine.
This study aimed to estimate the air kerma-area product of a mobile X-ray fluoroscopy system by using a real-time dosimeter in situations where an area dosimeter is not available. A real-time dosimeter was mounted on the C-arm of a mobile X-ray fluoroscopy system. Using an ionization dosimeter as a reference, we evaluated the correlation between the air kerma-area product measured by the fluoroscopy system and the readings obtained from the mounted real-time dosimeter. Tube voltage was varied from 60 to 110 kV in 10-kV increments, and phantom thickness varied from 5 to 20 cm in 5-cm increments. In addition, the irradiation field area was adjusted across five levels, ranging from 168.1 to 75.5 cm2. A strong linear relationship was observed between the real-time dosimeter readings x-axis and the air kerma-area product values y-axis, expressed as the regression equation y=36.181x+0.2188. The correlation coefficient was r=0.99. This study suggests that the air kerma-area product can be estimated by placing a real-time dosimeter on a mobile X-ray fluoroscopy system without an area dosimeter.
Tackle football is the most participated youth sport in the U.S. with leagues beginning as early as age 5. Exposure to cumulative repetitive head impacts (RHI) over years of play is increasingly viewed as a major contributor to chronic traumatic encephalopathy (CTE), a progressive neurodegenerative disease documented in contact sport athletes. Amid growing awareness of CTE, parents may turn to online information to guide decisions about youth tackle football participation. This cross‑sectional study examined the readability of online CTE information. Using the search term, 'CTE,' 68 URLs providing non‑technical information were identified after applying exclusion criteria. Online software was used to generate metrics from six widely-used readability formulas. Grade-level readability scores were categorized as ≤ Grade 8, 9-12, and ≥ 13 and summarized using descriptive statistics; distributions were compared by URL designation using chi-square tests (P < 0.05). Web page publication/revision date and presence of references were recorded. Median readability scores ranged from high school to early college with few pages meeting the recommended ≤ Grade 8 reading level for the general population. Levels were similarly high across non-commercial (.org,.gov,.edu) and commercial (.com) domains. Nearly 40% lacked clear publication or revision dates; fewer than half (47.1%) included references. Commonly accessed online CTE resources exceed recommended reading levels. This digital barrier impairs parents' functional health literacy and capacity for informed decision-making. As research on CTE and tackle football participation evolves, there is a need for plain‑language, clearly-sourced, updated online resources tailored to this decisional context.
This preliminary study aimed to evaluate the feasibility of thermal microchips as a non-invasive alternative for body temperature monitoring in dairy donkeys. Temperature was recorded in 11 jennies using thermal microchips implanted subcutaneously at the neck and read transcutaneous with an RFID scanner, a digital rectal thermometer, and an infrared thermographic camera directed at the neck over the microchip implantation site. Mean temperatures were 35.5 ± 0.82°C (microchip), 36.5 ± 0.59°C (rectal), and 36.0 ± 1.9°C (thermography). Bland-Altman analysis revealed that microchip readings were systematically lower than rectal temperatures by a mean of 1.01°C (95% LoA: -2.68 to 0.64°C), exceeding the clinically acceptable limit of 0.5°C (p = 0.03). Differences between microchip and thermography (bias: -0.49°C; 95% LoA: -5.20 to 4.22°C) and between rectal and thermography readings (bias: 0.52°C; 95% LoA: -3.35 to 4.41°C) were not statistically significant but exceeded acceptable agreement limits. The three methods are not interchangeable.
Differentiating between changes after radiotherapy and radionecrosis during the follow-up of brain metastasis remains challenging. Hybrid positron-emission tomography/magnetic resonance imaging is a relatively novel imaging modality that integrates advanced magnetic resonance imaging sequences with metabolic data from positron-emission tomography. This study ailed to evaluate retrospectively the contribution of hybrid (18F)-fluorodeoxyglucose positron-emission tomography/magnetic resonance imaging in distinguishing radionecrosis from recurrent cerebral metastatis, comparing the diagnosis performance of (18F)-fluorodeoxyglucose positron-emission tomography and magnetic resonance imaging independently, without relying clinical or prior imaging data. Over a one-year period, we included 72 patients with brain metastases undergoing evaluation for differential diagnosis between recurrence and radionecrosis. Two senior nuclear medicine physicians and two senior radiologists independently analysed positron-emission tomography and magnetic resonance images. For semi-quantitative positron-emission tomography, analysis maximum standardized uptake value was measured at 30min and 4h postinjection and retention indices were calculated. Reading correlations between physicians was assessed using Cohen's Kappa coefficients. Clinical and imaging follow-up, considered the gold standard, was used to classify the lesion. For discordant cases, a joint interdisciplinary reading was conducted. A total of 142 treated brain metastases from 90 hybrid (18F)-fluorodeoxyglucose positron-emission tomography/magnetic resonance imaging scans of 72 patients were analysed. Based on the gold standard for diagnosis, the 142 lesions were categorized into two groups: 47 recurrences and 95 cases of radionecrosis. The sensibility and sensitivity of hybrid (18F)-fluorodeoxyglucose positron-emission tomography/magnetic resonance imaging were 97.9 and 100 %, respectively, outperforming both magnetic resonance imaging and positron-emission tomography alone. Dual time-point (18F)-fluorodeoxyglucose positron-emission tomography, with retention index calculation, shows significant in differentiating radionecrosis from recurrence. Technological advancements, such as hybrid positron-emission tomography/magnetic resonance imaging, enhance diagnostic accuracy in substantial patients' cohorts. Further studies incorporating histological confirmation as the gold standard is warranted to validate these findings.