Category: Statistics

Neural Netw. 2025 Nov 14;196:108325. doi: 10.1016/j.neunet.2025.108325. Online ahead of print.

ABSTRACT

We conduct a comprehensive study on the quality inspection of lithium battery materials, which evaluates material conformity by analyzing particle sizes in scanning electron microscope (SEM) images. Currently, enterprises rely heavily on manual annotation to complete this task. However, manual annotation is labor-intensive and prone to subjective errors. To address these challenges, we reformulate the quality inspection task as the lithium battery materials SEM image segmentation (LBMS) task and aim to resolve it using artificial intelligence technology. To this end, we collect and construct a dedicated SEM image dataset for the LBMS dataset, called LBMS dataset. Then we propose a specialised model for the LBMS task, named LBMS-SAM. Specifically, we design an edge feature extraction module based on Sobel and Gabor convolutions (GSEFE), which aims to accurately extract and enhance image edge information. Additionally, We design a multi-layer denoised features fusion module (MDFF) that uses wavelet transform to denoise the output features of each global attention layer in the ViT model. The denoised features from different layers are then fused, enabling efficient extraction of global contextual information and suppressing noise introduced by the ViT architecture. The proposed model introduces minimal additional parameters, and extensive experiments on the LBMS dataset demonstrate that LBMS-SAM outperforms state-of-the-art (SOTA) methods across all relevant evaluation metrics.

PMID:41289643 | DOI:10.1016/j.neunet.2025.108325

Intensive Crit Care Nurs. 2025 Nov 24;93:104197. doi: 10.1016/j.iccn.2025.104197. Online ahead of print.

NO ABSTRACT

PMID:41289628 | DOI:10.1016/j.iccn.2025.104197

Eur J Radiol. 2025 Nov 17;194:112545. doi: 10.1016/j.ejrad.2025.112545. Online ahead of print.

ABSTRACT

OBJECTIVE: To characterize craniofacial, temporal-bone, vertebral, and systemic anomalies in oculo-auriculo-vertebral (OAV) spectrum using high-resolution computed tomography (HRCT) and to examine associations with clinical severity by the Tasse Objective Scoring System.

METHODS: We performed a retrospective study (2015-2024) at a national tertiary center including 223 clinically diagnosed OAV patients; 217 had bilateral temporal-bone HRCT suitable for analysis. HRCT assessed external auditory canal (EAC), ossicular, and intratemporal facial-nerve anatomy; inner-ear/vestibulocochlear-nerve abnormalities were evaluated in a subset. Vertebral anomalies were CT-confirmed when coverage was available; renal and cardiac findings were extracted from clinical records. Statistics included chi-square or Fisher tests with Cramér’s V, Cochran-Armitage trend tests across Tasse grades, and Spearman correlation for vertebral anomaly counts (two-sided α = 0.05).

RESULTS: Mean age was 7.6 ± 4.2 years; 55.2 % were male. In the HRCT subset, EAC stenosis/atresia and ossicular abnormalities were frequent and increased with Tasse severity (EAC: 48.4 %→59.8 %→82.8 %, p = 0.0078; ossicles: 40.3 %→49.6 %→82.8 %, p < 0.001), as did aberrant intratemporal facial-nerve course (24.2 %/27.4 %/53.3 %, p = 0.010). Inner-ear malformations were identified in 14.3 % and vestibular/vestibulocochlear-nerve anomalies in 42.9 % of those specifically evaluated. CT-confirmed vertebral anomalies occurred in 29.1 % overall; segmentation defects showed a strong grade-wise increase (p < 0.001) and the cumulative vertebral anomaly count correlated with Tasse severity (Spearman ρ = 0.41, p < 0.001). Renal anomalies were present in 16.6 % and rose across grades (p = 0.044; trend p < 0.001), whereas cardiac anomalies occurred in 14.8 % with no significant between-grade difference (p = 0.19).

CONCLUSION: Pairing HRCT phenotyping with Tasse severity stratification provides clinically actionable information for operative planning (canaloplasty/ossiculoplasty/device candidacy) and prioritizes systemic surveillance (spine and renal screening) in OAV spectrum. This integrated approach supports coordinated multidisciplinary care and offers a framework for future standardized screening and outcome-oriented research.

PMID:41289623 | DOI:10.1016/j.ejrad.2025.112545

J Chromatogr B Analyt Technol Biomed Life Sci. 2025 Nov 21;1269:124867. doi: 10.1016/j.jchromb.2025.124867. Online ahead of print.

ABSTRACT

Chronic inflammation is a significant contributor to various diseases but its assessment via blood sampling presents challenges, particularly in children. The evaluation of urinary biomarkers, including 3-bromotyrosine (Bty), 3-chlorotyrosine (Cty) and leukotriene E4 (LTE4), offers a non-invasive alternative. This study presents the optimization and validation of a sensitive liquid chromatography-tandem mass spectrometry (LC-MS/MS) method for the simultaneous quantification of Bty, Cty and LTE4 in urine. Under optimized conditions, sample preparation was based on SPE using Oasis MAX cartridges, followed by LC-MS/MS analysis. Method performance was validated using the ICH 10 guidelines, resulting in satisfactory results for all analytes in terms of recovery, linearity, limits of quantification, precision and accuracy. Recovery rates ranged from 82 % to 97 %, while matrix effects were observed within the range of -11 % to 26 %. Linear range spanned from 0.08 to 20 ng/mL for the three analytes. Application to 332 urine samples from the ENVIRONAGE birth cohort (Belgium), comprising of children aged 4-11 years, revealed detection frequencies of 18 % for LTE4, 19 % for Cty and 50 % for Bty. Notably, creatinine-corrected Cty and LTE4 exhibited statistically significant Spearman correlations with established systemic inflammation markers. Specifically, Cty was positively correlated with absolute monocyte count (ρ = 0.53, p < 0.05), while LTE4 showed a positive correlation with relative eosinophil levels (ρ = 0.46, p < 0.05) and a negative correlation with the relative neutrophil levels (ρ = -0.56, p < 0.01). These results highlight the validated method as a valuable tool for investigating distinct inflammatory pathways in epidemiological settings and clinical research.

PMID:41289620 | DOI:10.1016/j.jchromb.2025.124867

Neural Netw. 2025 Nov 8;195:108311. doi: 10.1016/j.neunet.2025.108311. Online ahead of print.

ABSTRACT

Identifying appropriate structures for generative or world models is essential for both biological organisms and machines. This work shows that synaptic pruning facilitates efficient statistical structure learning. We extend previously established canonical neural networks to derive a synaptic pruning scheme that is formally equivalent to an online Bayesian model selection. The proposed scheme, termed Bayesian synaptic model pruning (BSyMP), utilizes connectivity parameters to switch between the presence (ON) and absence (OFF) of synaptic connections. Mathematical analyses reveal that these parameters converge to zero for uninformative connections, thus providing reliable and efficient model reduction. This enables the identification of a plausible structure for the environmental model, particularly when the environment is characterized by sparse likelihood and transition matrices. Through causal inference and rule learning simulations, we demonstrate that BSyMP achieves model reduction more efficiently than the conventional Bayesian model reduction scheme. These findings indicate that synaptic pruning could be a neuronal substrate underlying structure learning and generalizability in the brain.

PMID:41289617 | DOI:10.1016/j.neunet.2025.108311

J Radiol Prot. 2025 Nov 25. doi: 10.1088/1361-6498/ae23d9. Online ahead of print.

ABSTRACT

Breast tissue is highly sensitive to ionizing radiation, making dose management in mammography crucial to reducing the risk of radiation-induced cancer and hereditary effects. Dose optimisation, guided by the ALARA principle, aims to minimise exposure while maintaining diagnostic quality. This study focuses on establishing National Diagnostic Reference Levels (NDRLs) for digital mammography in Nepal to support dose optimisation efforts. A retrospective analysis was conducted using data from 786 patients across six hospitals equipped with digital mammography systems. Both symptomatic and screening mammograms in cranial-caudal (CC) and mediolateral oblique (MLO) views were included for both breasts. Mean glandular dose (MGD) and entrance skin dose (ESD) were extracted from DICOM headers. For each mammogram view, data from a minimum of 50 patients were analysed. Technical parameters such as tube voltage (kVp), tube current (mAs), compression force (CF), and compressed breast thickness (CBT) were also documented. The established NDRLs for digital mammography are 1.03 mGy (RCC), 1.02 mGy (LCC), 1.18 mGy (RMLO), and 1.15 mGy (LMLO). The mean CBT and CF are 56±13 mm and 122±29 N, respectively. The overall NDRLs for CC and MLO views are 1.03 mGy and 1.17 mGy. Comparisons with other countries highlight the potential for further dose optimization to maintain diagnostically adequate images at lower exposure levels. Implementing such strategies can reduce patient radiation dose in digital mammography without compromising diagnostic performance.

PMID:41289610 | DOI:10.1088/1361-6498/ae23d9

JMIR Med Educ. 2025 Nov 25;11:e78363. doi: 10.2196/78363.

ABSTRACT

BACKGROUND: Learner autonomy-the ability to self-direct and regulate learning-is a key determinant of success in online education, yet its quantifiable impact in voluntary noncredit courses remains unclear. Understanding how autonomy translates into measurable behaviors and outcomes in clinical skills training may inform more effective online learning design and learning outcomes.

OBJECTIVE: This study aims to quantify the association between behavioral indicators of learner autonomy and performance in a voluntary noncredit online cardiac auscultation course.

METHODS: We conducted a prospective, self‑controlled, single‑center study. A total of 199 registrants (n=122 physicians and n=77 medical students) were recruited via WeChat and attended four weekly 2‑hour synchronous sessions using authentic patient heart sound recordings with imaging‑based explanations. The primary outcome was the final posttraining quiz score (0-100); training effectiveness was assessed by the pre‑ to posttraining score change. The autonomy indicators were full participation (attendance at all four sessions), in‑class engagement (number of responses to brief content‑aligned prompts posed approximately every 10-15 minutes; responses recorded for participation monitoring only), and postclass review (frequency/duration of reviewing recordings and materials). Analyses included Wilcoxon signed rank tests, χ² tests, multivariable linear regression, and receiver operating characteristic profiling of “excellent learners” (top 10% improvement).

RESULTS: Of the 199 registrants, 146 (73.4%) attended ≥1 session and 46 (23.1%) completed all sessions. Median test scores improved from 40 (IQR 20-50) to 70 (IQR 50-83; P<.001). Intrinsic motivation was associated with full participation (χ²₁=4.03; P=.045). In multivariable models, full participation (unstandardized B=41.55, 95% CI 24.43-58.67; standardized β=0.60; P<.001) and in‑class engagement (B=4.79 per additional response, 95% CI 3.05-6.45; β=0.70; P<.001) independently predicted higher final scores (adjusted R²=0.48). Receiver operating characteristic profiling indicated that greater postclass review (recordings/materials) led to learners achieving excellent performance.

CONCLUSIONS: In this voluntary online clinical skills course, showing up consistently, engaging during class, and reviewing after class-practical expressions of learner autonomy-were key correlates of short-term performance. These behaviors may be encouraged through simple, feasible course designs such as clear expectation setting, periodic interactive prompts, and structured review opportunities, which warrant prospective evaluation in future studies.

PMID:41289585 | DOI:10.2196/78363

JMIR Aging. 2025 Nov 25;8:e77227. doi: 10.2196/77227.

ABSTRACT

BACKGROUND: While the positive effects of digital technology on cognitive function are established, the specific impacts of different types of technology activities on distinct cognitive domains remain underexplored.

OBJECTIVE: This study aimed to examine the associations between transitions into and out of various technology activities and trajectories of cognitive domains among community-dwelling older adults without dementia.

METHODS: Data were drawn from 5566 community-dwelling older adults without dementia who participated in the National Health and Aging Trends Study from 2015 to 2022. Technology activities assessed included online shopping, banking, medication refills, social media use, and checking health conditions online. The cognitive domains measured were episodic memory, executive function, and orientation. Asymmetric effects models were used to analyze the associations between technology activity transitions and cognitive outcomes, adjusting for demographic, socioeconomic, and health-related covariates. Lagged models were applied for sensitivity analysis.

RESULTS: In the asymmetric effects models, the onset of online shopping (β=.046, P=.02), medication refills (β=.073, P<.001), and social media use (β=.065, P=.01) was associated with improved episodic memory. The cessation of online shopping was associated with faster episodic memory decline (β=-.023, P=.047). In contrast, the cessation of online banking (β=-.078, P=.01) and social media use (β=-.066, P=.003) was associated with decreased episodic memory. The initiation of instrumental, social, and health-related technology activities was associated with slower cognitive decline in orientation. The lagged models further emphasized the effects of stopping online banking and starting online medication refills in relation to episodic memory, as well as the positive associations between online shopping and social media use and orientation. All significant effects were of small magnitude.

CONCLUSIONS: Combining findings from the main and sensitivity analyses, results suggest that interventions designed to support episodic memory in older adults should emphasize promoting the use of online medication refill services and sustaining engagement with online banking, particularly among those who have already established these habits. To support orientation, strategies should focus on facilitating adoption of online shopping and social media use, helping older adults become comfortable navigating these platforms. Future trials are needed to assess the clinical relevance of targeted interventions for specific cognitive domains, to promote the initiation and maintenance of digital activities to help mitigate domain-specific cognitive decline in aging populations.

PMID:41289578 | DOI:10.2196/77227

Neurol Res. 2025 Nov 25:1-14. doi: 10.1080/01616412.2025.2592911. Online ahead of print.

ABSTRACT

OBJECTIVE: To develop a predischarge model for predicting 1-year post-discharge all-cause mortality in patients with aneurysmal subarachnoid haemorrhage (aSAH) following endovascular treatment (EVT).

METHODS: We retrospectively analysed 947 patients with aSAH who were discharged alive between April 2021 and April 2023 from four neurointerventional centres in China as the training cohort. Candidate variables were selected using the least absolute shrinkage and selection operator (LASSO) combined with the Boruta algorithm. Based on these features, six models – logistic regression (LR), XGBoost, random forest (RF), AdaBoost, decision tree, and gradient boosting decision tree (GBDT) – were developed and compared. The optimal model was selected by the area under the receiver operating characteristic curve (AUC). The external validation cohort comprised 692 aSAH patients discharged alive between April 2023 and April 2024 from two additional centres. Model performance was evaluated using AUC, calibration curves, and decision curve analysis (DCA). Given the imbalanced outcome distribution, we applied the Synthetic Minority Over-sampling Technique (SMOTE) to further assess model generalisability.

RESULTS: Among 1,639 patients alive at discharge, 67 (4.1%) died within 1 year. LASSO and Boruta jointly identified five key predictors for model construction: age, modified World Federation of Neurosurgical Societies (mWFNS) grade, ICU length of stay (ICU-LOS), C-reactive protein (CRP), and monocyte-to-HDL ratio (MHR). The random forest achieved the best discrimination in training set and remained strong in external validation cohorts.Moreover, SMOTE training yielded further improvements in generalisability.

CONCLUSION: Random forest model enables individualised pre-discharge risk stratification and may guide perioperative management.

PMID:41289577 | DOI:10.1080/01616412.2025.2592911

J Med Internet Res. 2025 Nov 25;27:e73131. doi: 10.2196/73131.

ABSTRACT

BACKGROUND: Preventive strategies integrated with digital health and artificial intelligence (AI) have significant potential to mitigate the global burden of atherosclerotic cardiovascular disease (ASCVD). AI-enabled clinical decision support (CDS) systems increasingly provide patient-specific insights beyond traditional risk factors. Despite these advances, their capacity to enhance clinical decision-making in resource-constrained settings remains largely unexplored.

OBJECTIVE: We conducted a randomized controlled study to assess the effect of AI-based CDS on 10-year ASCVD risk assessment and management in primary prevention.

METHODS: In a 3-way, within-subject randomized design, doctors completed 9 clinical vignettes representative of primary care presentations in a resource-constrained outpatient setting. For each vignette, participants assessed 10-year ASCVD risk and made management decisions using a conceptual prototype of AI-based CDS, automated CDS, or no decision support. The conceptual prototype represented contemporary risk calculators based on traditional machine learning models (eg, random forest, neural networks, logistic regression) that incorporate additional predictors alongside traditional risk factors. Primary outcomes were correct risk assessment and patient management (prescription of aspirin, statins, and antihypertensives; referral for advanced examinations). Decision-making time and perceptions about AI utility were also measured.

RESULTS: In total, 102 doctors from all 7 geographical regions of Indonesia participated. Most (n=85, 83%) participants were 26-35 years of age, and 57 (56%) were male, with a median of 6 (IQR 4.75) years of clinical experience. AI-based CDS improved risk assessment by 27% (χ22 (n=102)=48.875, P<.001) when compared to unassisted risk assessment, equating to 1 additional correct risk classification for every 3.7 patients where doctors used AI (number needed to treat=3.7, 95% CI 2.9-5.2). The prescription of statins also improved by 29% (χ22 (n=102)=36.608, P<.001). In pairwise comparisons, doctors who were assisted by the AI-based CDS correctly assessed significantly more cases (z=-5.602, n=102, adjusted P<.001) and prescribed the appropriate statin more often (z=-4.936, adjusted P<.001, medium effect size r=0.35) when compared with the control. AI-assisted cases required less time (estimated marginal means 63.6 s vs 72.8 s, F2, 772.8=5.710, P=.003). However, improvements in the prescription of aspirin and antihypertensives did not reach statistical significance (P=.08 and P=.30, respectively). No improvement was observed in referral decisions. Participants generally viewed AI-based CDS positively, with 81 (79%) agreeing or strongly agreeing that they would follow its recommendations and 82 (82%) indicating they would use it if given access. They believed CDS could enhance the efficiency of risk assessment, particularly in high-volume primary care settings, while noting the need to verify AI recommendations against clinical guidelines for each patient.

CONCLUSIONS: Improvements in risk assessment and statin prescription, coupled with reduced decision-making time, highlight the potential utility of AI in ASCVD risk assessment, particularly in resource-constrained settings where efficient use of health care resources and doctors’ time is crucial. Further research is needed to ascertain whether improvements observed in this online study translate to real-world low-resource settings.

PMID:41289576 | DOI:10.2196/73131