搜索结果：talked

Inspired by the auditory system's capacity to process spatiotemporal sound patterns, voiceprint recognition plays a vital role in identity authentication and security. However, current platforms often face challenges of speech frequency and amplitude variability, hindering accurate feature extraction in noisy environments. To address these issues, a large-scale hybrid metal-halide dynamic memristor (MHDM) featuring an engineered gradient-distributed architecture is developed for adaptive voiceprint recognition. The spontaneously graded metal-halide functional layer allows for precise modulation of Schottky barriers and redistribution of interface charges. This design achieves µs-scale response, enhances noise tolerance (over 20% improvement in signal-to-noise ratio), and enables kHz-scale dynamic signal processing. Experimental results demonstrate that the MHDM achieves a voiceprint recognition accuracy of 99.3%, maintaining high performance at 93.2% even in realistic background noise. These findings demonstrate the system's potential for secure and efficient voiceprint recognition, combining scalability with robust performance in noisy environments.

Flexible endoscopic evaluation of swallowing (FEES) is internationally recognised as an important instrumental assessment for paediatric dysphagia. Evidence supports its feasibility and safety in infants and children, but UK specific data remain limited. No published UK report describes a jointly delivered speech and language therapist (SLT) and paediatric otolaryngologist (ENT) FEES model applied across both neonatal and paediatric cohorts. To describe the implementation of a neonatal and paediatric FEES service delivered jointly by SLTs and ENT surgeons, and to evaluate clinical characteristics, FEES findings and subsequent changes to management. A retrospective audit included all children undergoing FEES between January 2024 and January 2025. Extracted data included demographics, comorbidities, referral indications, secretion management, penetration and aspiration events, swallow physiology, procedural tolerance and pre and post FEES feeding plans. Descriptive statistics were used to describe results. Governance approval was obtained through the hospital's information governance office. (IG 2024-890). Thirty-three children aged 4 days to 8 years underwent FEES. Comorbidities associated with dysphagia were present in 31/33 participants. Penetration or aspiration occurred in 21/33 participants, and secretion management difficulties occurred in 14/33 participants. Following FEES, 30/33 participants had a change to their feeding management plan compared with the plan developed based on a clinical feeding evaluation. No major adverse events occurred. One brief episode of mild epistaxis resolved spontaneously. The joint SLT and ENT model supported efficient scope passage and likely contributed to high procedural tolerance. A neonatal and paediatric FEES service can be safely implemented in a UK tertiary hospital using a joint SLT and ENT model. Simultaneous upper airway examination alongside FEES provided clinically meaningful information that frequently changed management. Findings support further multicentre work to establish national paediatric FEES pathways. What is already known on this subject FEES is a well-established instrumental assessment for paediatric dysphagia, providing objective insights into swallow function and airway protection. While widely used in countries like the U.S., Australia, and Canada, there is limited research describing paediatric or neonatal FEES in the United Kingdom or broader European context. Existing literature focuses primarily on feasibility and diagnostic accuracy, but does not address its implementation in UK public health systems or its integration with ENT services. What this study adds to the existing knowledge This is the first known study to describe the implementation of a joint ENT-SLT led paediatric FEES service within an NHS hospital. It identifies a high incidence of laryngomalacia during FEES, suggesting a potential association with oropharyngeal dysphagia even in the absence of classic airway symptoms. The study also demonstrates substantial variation in oral feeding plans post-FEES compared to bedside assessment, raising concerns about the standalone reliability of non-instrumental evaluations in neonates and infants. What are the potential or actual clinical implications of this work? The study supports routine inclusion of ENT in paediatric FEES, particularly for neonates, due to the frequency of structural airway anomalies such as mild laryngomalacia. It highlights critical limitations of bedside swallow evaluations as sole diagnostic tools and underscores the need for expanded access to FEES within the NHS. These findings reinforce the urgency for standardized UK-specific protocols and training pathways for neonatal and paediatric FEES to ensure accurate diagnosis and safe feeding management.

To evaluate the validity, clinical utility, barriers and facilitators of the manikin-based audiometric simulator, Clinical Assistant for Research and Learning. The validity, clinical utility, barriers, facilitators and simulation accuracy were evaluated via an online questionnaire and in-person pure-tone audiometric assessments. A total of 38 participants (age range 21-63) completed the study. All participants had formal training in pure-tone audiometric assessments. Participants were audiology students, registered practicing audiologists, hearing instrument dispensers, and undergraduate students. The programmed and measured audiograms had an absolute agreement within 5 decibel hearing level. Ease of use, clarity of workflow, realism, and feasibility in clinical practice were the reported facilitators, whereas cost and availability of technical support were the identified barriers to implementation. The manikin-based simulator was shown to be an accurate and valid tool in practicing pure-tone audiometric procedures. Clinical applications and future directions were also discussed.

Animals detect changes in the surrounding situation and predict what is likely to happen next by evaluating the current sensory information to that of a prior associative memory. Using working memory with inner speech, humans are able to flexibly predict the future situation and devise appropriate strategies to avoid dangers or achieve goals. This prospective thinking is supported not only by feed-forward sensory information with associative memory but also by working memory that maintains attended information for tens of seconds. Working memory is a type of short-term memory essential for providing the temporal and spatial continuity of attention during the transition from the current behavior to the next. We hypothesize that repeated associative learning of sensory signals with multisensory object imagery may form the reverberatory circuits between the sensory cortices and higher cognitive areas. The cortico-cortical reverberatory circuits may generate the cognitive scenes of objects. We also hypothesize that the cortico-thalamo-cortical loops may maintain the cognitive object-scenes as working memory to internally search for related memories of action planning and emotional self-state. Such internal search for relevant memory engrams may be a major role of working memory in thinking. In this perspective article, we discuss self-cognition and the circuit mechanism for it in the human brain.

Single-nuclei transcriptomics enables investigating ligand-receptor mediated cell-cell communication between different cell-types. However, current tools do not allow for non-programmatic means to access this analysis. Additionally, most methods can not account for molecular pathways involved in downstream receptor signaling in cell-cell communication. We developed scVizComm, a ShinyApp based web-portal that can be used to interactively visualize ligand-receptor networks across cell-types and different conditions. scVizComm can be used to host pathway centric pre-calculated cell-cell communication results. Using human kidney and heart single-nuclei transcriptomics data, we showcase the utility of scVizComm in understanding ligand-receptor interactions involved in fibrosis related pathways. Additionally using mouse anterior brain data, we demonstrate the major ligand-receptor expression exploration in space. scVizComm allows interactive visualization and pathway centric prioritization of cell-cell communication analyses.

In recent years, there has been an increase in studies reporting on effective child language interventions for people with or at risk for (Developmental) Language Disorder ((D)LD). However, the translation of this evidence into practice has been impeded by under-specified intervention reporting, specifically on what the active ingredients of therapy are and how they are defined. This systematic review forms part of a larger research programme conducted by the Intervention Consensus for Language Disorder group. To identify, summarise, and synthesize how the active ingredients of oral language interventions for children with or at risk for (D)LD have been labelled, defined, described, and classified in empirical and clinical literature. This registered review (PROSPERO ID CRD42024541407) adhered to PRISMA guidelines. Search terms were included in seven electronic databases. Included literature comprised peer-reviewed oral language intervention studies (published in English, German, Portuguese, Croatian, Italian, or Finnish; between January 2019 and May 2024; and reporting on participants who were ≤ 18 years, with/at risk for (D)LD); intervention-focused taxonomies; intervention manuals (published in the last 10 years); and textbooks used to teach child language interventions in pre-registration Speech and Language Therapy/Pathology courses (identified through the Whatworks database and a social media survey, respectively). Data extraction was guided by the TIDieR checklist with additional items deemed relevant for this review. 9576 articles were identified and screened; 619 were included for full text screening; and 243 articles were included in the review. Significant reporting inconsistencies were evident, including identical labels masking different mechanisms, as well as similar mechanisms operating under different labels; conflation of discrete techniques and procedures; the application of common labels in multiple ways; the use of unfamiliar terminology to refer to familiar techniques; and the use of broad terms masking multiple active ingredients. These inconsistencies do not align with what is considered well-specified active ingredients and consequently, significantly impede intervention replication and cross study comparisons. This review highlights the need to develop consensus on (1) how active ingredients of intervention are labelled and defined so that they are consistent, precise and non-overlapping and (2) a comprehensive integrative taxonomy for ease of understanding and use, when reporting on oral language interventions for children with or at risk for (D)LD. What is already known on this subject Oral language interventions contain numerous proposed 'active ingredients', but these are reported with substantial inconsistency across studies, textbooks, and intervention manuals. In addition, different taxonomies classify similar elements in divergent ways. What this paper adds to the existing knowledge We have provided the first comprehensive analysis demonstrating the extent of inconsistency, ambiguity, and conceptual overlap in how active ingredients are labelled and described in the oral language intervention literature. There is misalignment across current taxonomies and we lack an internationally agreed framework that integrates discrete active ingredients along with dosage and contextual active support ingredients. What are the potential or actual clinical implications of this work? Without clear, shared definitions of active ingredients, clinicians cannot reliably interpret research, replicate interventions, or implement evidence-based practices with fidelity. A unified internationally agreed lexicon and taxonomy would enable clinicians to understand precisely what constitutes a given technique, how it should be delivered, and what child behaviours it is intended to influence, strengthening the quality and consistency of intervention delivery.

Online or other child sexual exploitation or abuse (OCSEA) perpetration prevention programs are designed to provide services to individuals prior to offending. There is limited research on OCSEA perpetration prevention programs in jurisdictions with mandatory reporting. The present study is a mixed-method feasibility evaluation of the OCSEA perpetration prevention therapy program, Talking for Change (TFC). The feasibility study broadly addressed whether adult clients could be recruited to the program, the clinical profile of clients, and their satisfaction with the program. In total, there were 162 distinct individuals referred with the number of referrals increasing on an annual basis. Clients in the evaluation (n = 40) presented with risk relevant factors related to paraphilic interests and sexual self regulation problems with interpersonal challenges as well as several protective factors (e.g., empathy). The population had a high level of comorbid mental health problems, and many had engaged in prior CSEM use. Program satisfaction interviews resulted in five themes that highlighted the value of the group (e.g., promoting acceptance and accountability) and challenges (e.g., problems with a one size fits all approach). The results indicated that OCSEA perpetration prevention programs are feasible in mandatory reporting jurisdictions. Implications for program delivery are discussed.

To explore mechanisms that lead mothers with perinatal mental health difficulties to take up an offer of face-to-face peer support from trained volunteers, and contextual factors triggering those mechanisms. Realist evaluation methods were used to identify context-mechanism-outcome configurations for take-up in a community-based, third sector programme offering one-to-one and group peer support in the UK. Data were from semi-structured interviews with mothers (n = 20), peer support volunteers (n = 27), and staff (n = 9), and baseline quantitative and open-text data about mothers at referral (n = 260). 182/260 mothers (70%) took up peer support, varying between sites with different populations. Twenty context-mechanism-outcome configurations explained take-up, which depended on complex contextual factors. Mechanisms clustered around mothers' expectations of psychological safety; desire for a meaningful reference group for social comparison; beliefs about what they may gain; expectations about how it might compare to professional mental health support; and practical issues such as mode of delivery. The purpose of peer support can be misunderstood. Not all communities see value in talking about emotional distress with peers, and not all mothers trust peers to be non-judgemental. One-to-one and group support may feel safe to different mothers. Programmes should work with mothers, third sector organisations, and community leaders, to identify how perinatal mental health peer support can be locally offered most appealingly. Accessibility may be increased by offering a choice of group or one-to-one support, and a blend of face-to-face, telephone, and video-conferencing. Publicity material should clearly explain what peer support is and how it can help.

暂无摘要（点击查看详情）

Many autistic children have spoken vocabulary delays that impact communication abilities during critical developmental periods. Behavioral and brain predictors of spoken language development in autism have been independently identified but rarely evaluated simultaneously. Here, we examined how fine motor skills, joint attention, and white matter microstructure of speech-related tracts are associated with spoken vocabulary development in autistic children with a broad range of speech abilities. Spoken vocabulary was assessed longitudinally in 122 autistic children (38% female) first between 2-4 years (Time 1) and again approximately two years later. Children were divided into groups based on their Time 1 spoken vocabularies: larger, average, or smaller. Fifty-eight autistic children (47% of the sample) were in the smaller vocabulary group, which represented both emergent speakers and non-speaking individuals. Within this smaller vocabulary group, faster rates of spoken vocabulary development were associated with better Time 1 fine motor and joint attention skills, and lower Time 1 fractional anisotropy (FA) in the occipital inferior longitudinal fasciculus, central arcuate fasciculus, and inferior corticospinal tract. Rates of spoken vocabulary development were best predicted by models that included both brain and behavior metrics, which each individually explained significant variance and together accounted for roughly half of the model variability. Results suggest that together early childhood fine motor skills, joint attention skills, and white matter microstructure contribute to a developmental cascade supporting spoken language. Findings demonstrate the potential clinical utility of early brain scanning, indicating that microstructure is predictive of language development above and beyond salient behavioral predictors.

Effective risk management is central to patient safety, regulatory compliance, and organizational sustainability in healthcare. In South Africa, audiology and speech therapy practitioners operate within a system characterized by resource constraints, regulatory complexity, and socio-economic inequality, resulting in differentiated risk exposure across practice settings. This study explored risks encountered by practitioners and the strategies used to mitigate these risks. A qualitative design was employed in 2025 with 92 practitioners. Data were analyzed thematically, supported by descriptive and inferential analyses. Four interrelated risk domains emerged, with sectoral differences reflected across thematic and quantitative findings. Public sector practitioners reported significantly higher clinical risk, including diagnostic errors (65%) and delayed treatment (70%), particularly in rural settings (&#x3c7;2 = 18.21, p < 0.01). Additionally, operational risks such as staff shortages (75%) and equipment failure (70%) were also more prevalent. In contrast, private practitioners reported greater financial risk related to income volatility and overhead costs (70%; t = 3.12, p < 0.01). Legal and regulatory risks were evident across both sectors, while mitigation strategies included telehealth adoption and community partnerships. These findings demonstrate structurally produced risk, highlighting the need for integrated, context-sensitive risk governance frameworks in resource-constrained health systems.

With the growing use of AI-powered conversational technology, people with dementia may benefit from these tools. Finetuning and adapting such systems require an understanding of the language use of people with dementia and communication patterns. Daily conversations with 120 Korean older adults at various stages of dementia were collected over a two-year period and transcribed. Text mining techniques were applied to identify language patterns and latent meanings, including Term Frequency (TF), Term Frequency-Inverse Document Frequency (TF-IDF), co-occurrence network and concordance analyses, and topic modeling using both Latent Dirichlet Allocation (LDA) and BERTopic. TF/TF-IDF analyses of 6,989 speech segments by people with dementia revealed frequent use of pronouns and vague demonstratives. Word pairs and in-context usage of frequently used words were closely tied to their lived experiences. Topic modeling identified five themes via LDA and nine key topics from 29 BERTopic subcategories, covering hardship, family and relationships, emotions, reminiscence, and daily pleasures. BERTopic also captured health-related needs, identity, and agency, offering insight into the inner world of people with dementia. Despite declining cognitive abilities, people with dementia experience a wide range of life events and emotions. This study highlights the need for language-based support to help them retain agency, rather than be viewed solely as dependent or in need of care. Furthermore, the words and themes from this study could serve as valuable input for developing AI-powered communication algorithms, enabling more dementia-friendly and engaging conversations with people with dementia.

There is an acute unmet need for specialist palliative care services in Residential Aged Care Homes (RACHs). Palliative care needs rounds (Needs Rounds) are a promising evidence-based outreach approach to improve access to specialist care for this population. This study aimed to evaluate the feasibility and cost effectiveness of implementing needs rounds across a metropolitan Australian health service. A retrospective pre-post cohort study was conducted comparing outcomes from May-October 2021 with the same period in 2019. Primary outcomes were emergency department (ED) presentations and hospital admissions among RACH residents. Secondary outcomes included estimated cost savings and released hospital capacity. Analyses were conducted from the perspective of the public health service provider. Findings are descriptive, and no formal statistical hypothesis testing was undertaken. Needs rounds were implemented in 39 of 56 RACHs (70%) within the first nine months. Facilities not participating were primarily affected by operational constraints, including staffing pressures and pandemic-related disruptions. Compared with 2019, ED presentations declined by 20% and hospital admissions by 64%. Estimated cost savings associated with reduced hospital utilization were AU$498,315 per month. These findings should be interpreted cautiously given concurrent system-wide changes associated with the COVID-19 pandemic. Needs rounds appear feasible to implement at scale and may reduce hospital utilization among RACH residents. Preliminary economic findings suggest potential cost savings and released capacity within the health system. However, causal attribution remains uncertain, and full implementation costs were not captured. Further controlled studies incorporating patient-centred outcomes are required.

Cough hypersensitivity is increasingly recognized as a key mechanism underlying chronic cough across diverse etiologies. The Cough Hypersensitivity Questionnaire (CHQ) was developed to systematically assess characteristic triggers and abnormal sensory perceptions; however, its structural and clinical relevance across respiratory diseases remains unclear. We prospectively enrolled 300 adults with chronic cough from multiple respiratory centers. Participants completed four cough-related questionnaires: the Numeric Rating Scale (NRS), Leicester Cough Questionnaire (LCQ), Cough Assessment Test (COAT), and CHQ. Spearman correlation and network analyses were used to evaluate associations between CHQ items and other cough-related measures, including subgroup analyses by disease etiology. The CHQ demonstrated variable associations with conventional instruments and differing patterns across respiratory diseases. Triggers such as perfume and smoke correlated with LCQ hypersensitivity items, while the sensation item "urge to cough" exhibited broad correlations across physical, psychological, and social domains. Socially mediated triggers (e.g., talking) aligned with LCQ social interference items. In contrast, items such as hot air, dampness, laughter, and eating/drinking had minimal external associations. Internal CHQ correlations were sparse, and no significant associations were observed between CHQ and NRS, indicating a dissociation between hypersensitivity features and perceived severity. Disease-specific analyses suggested differing patterns, with broader associations observed in asthma and bronchiectasis compared with chronic obstructive pulmonary disease and idiopathic pulmonary fibrosis. The CHQ captures distinct and heterogeneous dimensions of cough hypersensitivity that are not adequately reflected by conventional cough measures, suggesting its potential utility for characterizing clinically relevant multidimensional aspects of chronic cough.

Brain magnetic resonance imaging (MRI) abnormalities are an important finding in the evaluation of patients with suspected autoimmune encephalitis (AE). There have been few studies evaluating the frequency and prognostic significance of MRI abnormalities, especially hippocampal swelling, in anti-N-methyl-D-aspartate receptor (NMDAR) and anti-leucine-rich glioma-inactivated 1 (LGI1) Ab-mediated encephalitides. We conducted a multi-centre, retrospective study involving adult patients with confirmed antibody-mediated encephalitis and at least one MRI scan from 10 Australian hospitals (n = 139). MRI scans were evaluated by a neuroradiologist blinded to the specific autoimmune encephalitis diagnosis. We evaluated associations between acute MRI abnormalities (e.g. hippocampal swelling) with 12-month function (modified Rankin scale, mRS &#x2265;2 = worse outcome) and radiological findings. In patients with anti-LGI1 Ab-mediated encephalitis, we identified hippocampal swelling on initial MRI to be associated with worse function at 12 months (OR 0.03; 95% CI 0.003, 0.34; P = 0.005). We found initial AE-associated T2/fluid attenuated inversion recovery (FLAIR) hyperintensities were not associated with 12-month mRS in either the anti-NMDAR (OR 0.39; 95% CI 0.04, 3.97; P = 0.42) or anti-LGI1 Ab-mediated encephalitis groups (OR 0.34; 95% CI 0.07, 1.57; P = 0.17). In anti-LGI1 Ab-mediated encephalitis, both hippocampal swelling (OR 5.76; 95% CI 1.14, 29.02; P = 0.03) and T2/FLAIR hyperintensity (OR 6.81; 95% CI 1.28, 36.22; P = 0.03) were related to the development of mesial temporal atrophy and hippocampal sclerosis. Acute hippocampal swelling is associated with worse outcomes in anti-LGI1 Ab-mediated encephalitis and, alongside initial T2/FLAIR hyperintensity, is associated with the development of both mesial temporal atrophy and hippocampal sclerosis.

Catatonia is increasingly diagnosed in neurodevelopmental disorders, yet most data focus on individuals with autism spectrum disorder. To our knowledge, catatonia has not been reported in association with chromosome 18 deletion syndrome. We describe a 19-year-old woman with a mosaic terminal 18q deletion (18q23) who developed severe catatonia following SARS-CoV-2 infection. She presented with mutism, stupor, negativism, posturing, rigidity, waxy flexibility, and withdrawal. Extensive neurological, infectious, and autoimmune workup, including cerebrospinal fluid analyses, revealed no abnormalities. Electroencephalography (EEG) showed intermittent frontotemporal right-dominant dysfunction, characterized by short clusters of high-amplitude polymorphic theta paroxysms without epileptiform discharges. Symptomatic treatment with lorazepam led to rapid initiation of speech within 24&#xa0;h and full remission of catatonic features after four days. Follow-up EEG recordings at three and ten months were normal, and no recurrence occurred. This first reported case raises the possibility that individuals with the rare chromosome 18q deletion, including mosaic forms, may be vulnerable to catatonia particularly in the presence of systemic or environmental stressors such as viral infection or environmental change. Awareness of this potential association and early benzodiazepine treatment are critical to prevent complications and facilitate recovery.

Gestures and silent pauses are fundamental in comprehension and communication of emotional speech. While previous research has examined gesture inhibition primarily in neutral or fictional speech tasks with mixed findings, an empirical study on the effect of gesture inhibition on silent pause behaviour during personal emotional autobiographical narratives is yet not examined. The present study addresses this gap. Thirty participants (15 gesture-restricted, 15 non-restricted; aged 18-23) narrated personal negative emotional memories and neutral daily routine in Hindi-English bilingual speech. Silent pauses were categorized by duration into short (250-500&#xa0;ms), medium (500-1000&#xa0;ms) and long (more than 1000&#xa0;ms) pauses, based on distinct cognitive and affective functions. Pause count, duration, and proportion were analysed using linear and generalized linear mixed-effects models on 4242 individual pauses. Emotional narratives consistently produced more frequent, longer, and proportionately larger long pauses than neutral narratives across both gesture conditions, with no comparable effects on shorter pauses. Gesture-restriction selectively increased long-pause counts in both narrative contexts, with directionally consistent but smaller effects on long-pause duration and proportion. Short and medium pauses showed no significant contrast effects across groups and narratives. The findings indicate that emotional content and gesture restriction selectively modulate the long-pause range associated with higher-order cognitive and affective processing, while leaving shorter segmenting pauses unaffected. Results are discussed within gesture-for-conceptualization and embodied emotion frameworks.

Swallowing involves the precise coordination of muscles and brain areas and can be disrupted in a variety of neurological conditions. Current methods to visualize swallowing cannot examine both the biomechanics and brain activity associated with specific swallowing events. An updated version of a pulse sequence that simultaneously samples BOLD-based fMRI and dynamic imaging (called SimulScan) is introduced that provides higher quality and faster dynamic imaging, enabling data-driven analysis of swallowing function through a partial least squares (PLS) analysis. Integrating updated dynamic imaging approaches, SimulScan achieved dynamic MRI at 23.75 frames per second with a 30&#x2009;cm field of view with BOLD fMRI at a 1.6&#x2009;s TR. Five subjects were scanned with SimulScan twice and with videofluoroscopy to compare the preliminary reliability of measuring swallowing biomechanics using computational analysis of swallowing mechanics (CASM) and the test-retest relationship in correlated functional and dynamic components of PLS. High reliability of biomechanical measures of swallowing was achieved across the two SimulScan runs with CASM (r&#x2009;=&#x2009;0.891; p&#x2009;<&#x2009;0.0001) and between SimulScan and videofluoroscopy (r&#x2009;=&#x2009;0.686; p&#x2009;<&#x2009;0.0001). Correlations between dynamic and functional imaging across runs also showed high reliability (mean correlation of first 3 latent variable timeseries was 0.49 (p&#x2009;<&#x2009;0.001) within a run and 0.17 (p&#x2009;<&#x2009;0.001) across runs), indicating that SimulScan with PLS can extract reliable maps of linked correlations between the brain and the oropharyngeal dynamics. The updated SimulScan with PLS analysis enables the study of central control of swallowing, providing simultaneous biomechanical visualization of the swallow along with brain functional signals.

Identifying autism traits and detecting the spectrum of disorder early can substantially improve quality of life. We present a text-driven approach for early autism risk detection that analyzes caregiver-reported behavioral descriptions using advanced natural language processing techniques. Synthetic free-text generated from validated screening items is used to train multiple language models, which are then evaluated on an external benchmark dataset (TASD) to assess generalization under domain shift. Fine-tuned transformer models achieve the highest performance, reaching 90% accuracy, outperforming GPT and Gemini models, and conventional NLP baselines. Augmenting training datasets with noisy, realistic text further improves model performance, specifically recall in traditional pipelines, demonstrating the potential of noise-aware data augmentation for free-text screening. This methodology enables translational and low-cost early assessment without requiring structured questionnaires or speech samples. This approach provides early cues that may assist specialist evaluation, promote accessible and proactive developmental health monitoring.