Browse Articles

Artificial intelligence for early detection of diabetic retinopathy: A vision transformer-based approach

Background Early identification of diabetic retinopathy (DR), which is a primary cause of vision impairment globally, is a crucial phasis for effective intervention and treatment. Traditional screening workflows rely on manual diagnosis by ophthalmologists, which remains the gold standard but can be time-consuming and subject to variability due to human factors. To support and enhance the screening process, artificial intelligence (AI)-based tools have shown promise in automating DR detection, particularly with recent advances in deep learning. However, medical images with long-range dependencies and spatial linkages can be challenging for CNN-based algorithms to handle. Methods This paper proposes a Vision Transformer (ViT)-based model, specifically using a Compact Convolutional Transformer (CCT), for early automated detection of DR. The model uses self-attention techniques to improve feature extraction and classification performance; combining three main stages: the CCT tokenizer, transformer encoder, and sequence pooling. The proposed approach was trained on public datasets (EyePACS and APTOS 2019) and evaluated against state-of-the-art deep learning architectures. Results Our experimental findings demonstrate that ViT performs among the best in the current state of the art with an overall accuracy of 97% and F1-scores above 0.95 across all DR severity levels. Our system is primarily designed for the pre-screening stage of diabetic retinopathy workflows, enabling rapid and reliable identification of potential DR cases for further clinical evaluation. Conclusion These results highlight the potential of transformer-based designs in medical picture analysis, as well as the implications for telemedicine and e-health solutions in real-time, especially in cases of low-resource settings.

Persistence of viable opportunistic pathogens in a multi-stage natural wastewater treatment system

Nature-based wastewater treatment systems are increasingly implemented to support water reuse in arid regions, yet their effectiveness in removing viable culturable opportunistic pathogens remains insufficiently characterized. In this exploratory, culture-based study, we assessed bacterial population dynamics across a five-stage natural wastewater treatment system (Wadi Hanifa, Riyadh, Saudi Arabia), tracking culturable bacteria from secondary-treated influent to sand-filtered effluent. Water samples were collected at five sequential treatment stages and analyzed for physicochemical parameters, total culturable bacterial abundance, bacterial diversity, taxonomic composition, and antimicrobial susceptibility of persistent isolates. Total culturable bacterial counts decreased by approximately 1.1 log 10 CFU mL ‒1 across the system, accompanied by an approximately 50% reduction in observed isolate richness. The fecal indicator Escherichia coli was detected only in upstream and intermediate stages (sampling locations L3.1-L3.3) and was absent from downstream samples. In contrast, the opportunistic pathogen Klebsiella pneumoniae was recovered across all five treatment stages and accounted for 44.4% (8/18) of all morphologically distinct isolates grown on the selected culture media. Turbidity declined by 71% along the treatment train and showed a strong positive correlation with bacterial richness (Kendall’s τ = 0.84, p = 0.038), although this exploratory correlation should be interpreted with caution given the small sample size (n = 5 stages). Phenotypic antimicrobial susceptibility testing revealed multidrug resistance (MDR) in 62.5% (5/8) of K. pneumoniae isolates, although all remained susceptible to amikacin and meropenem. Under the conditions examined, multi-stage natural wastewater treatment substantially reduced overall bacterial abundance and diversity but did not eliminate viable, multidrug-resistant Klebsiella pneumoniae . The discordance between fecal-indicator removal and opportunistic pathogen recovery highlights system-specific limitations of indicator-based monitoring for assessing microbial safety in wastewater reuse systems. Given the limited isolate number (n = 8 K. pneumoniae ) and the absence of molecular resistance-gene characterization, broader claims about wastewater as a dissemination pathway for antimicrobial resistance cannot be drawn from these data. These findings are based on a single cross-sectional sampling event and should be considered hypothesis-generating rather than confirmatory.

Dual regulatory effects of medical bovine collagen sponge on macrophages in vitro

Medical bovine collagen sponges (MBCS) are extensively utilized in clinical settings for hemostasis and tissue augmentation owing to their superior biocompatibility, biodegradability and hemostatic properties. Nevertheless, their immunomodulatory influence on macrophage function remains largely unexplored. In this study, we systematically evaluated the effects of MBCS on RAW264.7 macrophages through comprehensive in vitro assays, including proliferation analysis, phagocytosis assessment, cytokine expression and angiogenesis evaluation. Our results demonstrate that MBCS significantly enhances macrophage proliferation and phagocytic capacity while potently stimulating VEGF secretion. Furthermore, MBCS-conditioned macrophage supernatants markedly promote endothelial tubule formation in vitro. Notably, MBCS moderately promotes inflammatory responses in resting macrophages, whereas pretreatment with MBCS effectively inhibits lipopolysaccharide (LPS)-induced pro-inflammatory reactions, as reflected by reduced production of TNF‑α, nitric oxide and CD86. These findings reveal the effects of MBCS on macrophages at the cellular level, providing experimental evidence and mechanistic support for its tissue-repairing functions.

The effect of squat training with different eccentric contraction tempos on lower limb muscle strength

Objective This study aims to investigate the impact of 8-week squat training with different eccentric contraction tempos on maximal squat strength, jump height, and sprint performance among university students majoring in sports training. Methods Thirty participants were randomly assigned to either the 4/0/X/0 group, the 2/0/X/0, or the self-selected eccentric, isometric and inter-repetition tempo group (V/V/X/V), performing tempo-specific loaded squat training twice weekly for 8 weeks. Maximal squat strength, 30-meter sprint performance, countermovement jump (CMJ), and squat jump (SJ) heights were assessed before and after the intervention. A two-way repeated-measures ANOVA and effect size analysis were used to evaluate the training outcomes. Results 1) For maximal squat strength, no statistically significant inter-group difference was detected (P > 0.05). All three groups exhibited significant within-group improvements relative to baseline (P < 0.05 to P < 0.01). Descriptive effect size analysis showed the magnitude of improvement ranked as follows: 2/0/X/0 group (ES = 1.45)> V/V/X/V group (ES = 0.85)> 4/0/X/0 group (ES = 0.64). 2) For CMJ height, both the 2/0/X/0 and V/V/X/V groups showed significant improvement compared to pre-intervention (P < 0.05), while the 4/0/X/0 group exhibited a significant decrease (P < 0.01). ES magnitudes followed the order: 2/0/X/0 group (ES = 1.07) > V/V/X/V group (ES = 0.30) > 4/0/X/0 group (ES = −0.90). 3) SJ height results mirrored CMJ trends, with ES values ranking:2/0/X/0 group (ES = 1.09) > V/V/X/V group (ES = 0.34) > 4/0/X/0 group (ES = −0.83). 4) In 30-meter sprint performance, the 2/0/X/0 group demonstrated significant improvement (P < 0.05), the 4/0/X/0 group showed significant decline (P < 0.01), and the V/V/X/V group exhibited no significant change. Effect sizes indicated significant improvement in the 2/0/X/0 group (ES = −0.76), no meaningful change in the V/V/X/V group (ES = −0.10), and a significant decline in the 4/0/X/0 group (ES = 0.63). Notably, despite significant gains in maximal strength, the 4/0/X/0 group displayed concurrent decrements in stretch-shortening cycle-based explosive performances. Conclusion Compared with the 4/0/X/0 and V/V/X/V tempos, the 2/0/X/0 tempo showed the most favorable improvement trend in lower-limb maximal strength, and induced significantly greater enhancements in explosive power performance.

Public parks utilization and citizen satisfaction in Bangkok Metropolitan: An integrated theoretical model for tropical urban health

Background Tropical megacities face escalating non-communicable disease burdens requiring innovative urban health solutions. Despite public parks’ critical role as health infrastructure, limited research exists on utilization patterns and satisfaction in tropical urban contexts, where extreme heat and humidity create unique challenges. The equitable distribution of park benefits across socioeconomic groups remains underexplored in Southeast Asian contexts, a gap this study directly addresses. Objective This study developed and validated an integrated theoretical model of public park utilization and citizen satisfaction in tropical environments, identifying key determinants and providing evidence-based recommendations for health-promoting urban design, with attention to socioeconomic and gender equity dimensions. Methods We conducted a sequential explanatory mixed-methods study combining the Social Ecological Model, Expectation-Disconfirmation Theory, and Place Attachment Theory. Data were collected from 1,200 park users across 30 Bangkok parks using four-stage stratified random sampling, supplemented by 30 in-depth interviews. Analysis included hierarchical regression, thematic analysis, and comprehensive validity assessment. Results Participants (58.3% female, mean age 35.9 years) visited parks 3.3 times weekly for 1.8 hours per session. Park quality emerged as the strongest satisfaction predictor (β  =  0.401–0.607, all p  <  0.001), followed by usage patterns and accessibility. The integrated model explained 61.7–68.5% of satisfaction variance, a relatively high figure that partly reflects the common-method effects inherent in self-report designs (see Limitations). All nine hypotheses were confirmed with strong correlations (r  =  0.554–0.796). Climate-specific challenges included inadequate shade coverage (M  =  2.96/5.0) and insufficient evening lighting (M  =  2.98/5.0). Qualitative analysis revealed parks function as urban oases, social ‘third places,’ and community-building spaces, with heat protection identified as the primary accessibility barrier. Significant socioeconomic gradients were observed, with lower-income users reporting systematically lower satisfaction across all domains. Conclusions This study provides a comprehensive integrated theoretical model for tropical urban park utilization, indicating that strategic quality improvements are more strongly associated with satisfaction than quantity expansion. Climate-adaptive features—particularly heat protection and extended evening access—represent essential design requirements differing fundamentally from temperate guidelines. The findings indicate that park benefits are not equitably distributed, with implications for Bangkok Metropolitan Administration (BMA) policies including the 15-minute city park initiative.

A simulation model to predict the most efficient way to utilise operational resources when vaccinating badgers against bTB

Bovine tuberculosis (bTB) is a disease carried by badgers that seriously impacts the health and productivity of cattle in the United Kingdom (UK). The UK government’s aim is to be officially TB-free in cattle by 2038 in England and badger vaccination is one approach that can be used to achieve this. We use a simulation model to compare the relative performance of different vaccination strategies (treatment every year, every second year and every third year) under different field conditions and assumptions about vaccine efficacy. Population density, disease prevalence and duration of vaccine-induced immunity substantially affected the outcome of vaccination. A vaccine strategy with lower frequency treatment was effective and offers operational efficiency as multiple areas can be treated near concurrently by a single vaccination team, potentially halving the cost per unit area. The information presented here may be used to guide operational managers and decision makers towards a more efficient strategy in a particular area and may provide an estimation of the likely success of the chosen strategy.

Dietitian-led intervention to manage constipation in Parkinson’s disease: Study protocol for a parallel-group randomized controlled trial (NUTRI-GUT-PD)

Background Constipation is a frequent non-motor symptom in Parkinson’s Disease (PD). Standard care includes dietary and lifestyle guidance or the use of laxatives. Diet represents a promising non-pharmacological approach, but its management in PD is complex and goes beyond simple dietary recommendations. This study protocol aims to examine whether nutritional counseling delivered by a dietitian is more effective than usual care in reducing constipation symptoms in PD patients. Methods This 90-day randomized controlled superiority trial uses a parallel-group design. A total of 54 outpatients with PD without dementia and fulfilling the Rome IV criteria for functional constipation will be included. Baseline assessments comprise demographics, nutritional evaluation (anthropometric measurements and sarcopenia indicators), neurological assessment, and the Constipation Scoring System (CSS) scores. Participants will perform stool sample collection for microbiota analysis and will be requested to complete three 24-hour dietary recalls over the subsequent week. After baseline, participants will be randomly allocated to either the control group (usual care) or the intervention group (nutritional counseling delivered by a dietitian). At the first study visit, participants will deliver the stool sample and dietary recalls. The intervention will last for 90 days and will include one in person consultations and bi-weekly follow-up phone calls. Nutritional counseling will address four main topics: healthy eating, food processing, fiber and fluid intake, and levodopa interactions with diet. In addition, participants in the intervention group will receive an individualized diet plan. All baseline assessments will be repeated at the 90-day endpoint. The primary outcome is the change in weekly bowel movement frequency from baseline to day 90, measured through the stool diary. Secondary outcomes include CSS, changes in fecal gut microbiota, macronutrient distribution, diet quality, body composition and sarcopenia indicators. Trial registration number: NCT07213856.

A systems biology approach to find representative genes in Acute Myeloid Leukemia

In this study, we modeled gene expression profile data from Acute Myeloid Leukemia (AML) and healthy cases. At first, the GEO-GSE9476 dataset was processed, and a total of 341 genes were identified as differentially expressed genes (DEGs) in patients, and 599 DEGs in healthy individuals. Gene Ontology and pathway analysis on DEGs led to the identification of 5 Transcription Factors for patients and 3 for healthy cases. Analysis of the respective metabolic pathways revealed a common region in the metabolic pathway between AML and Tuberculosis (TB) that confirmed the validity of our procedure due to the consistency with similar reports. Upon PPI network analysis, Hub genes and three modules containing 41 up-regulated and down-regulated genes in AML patients were identified. Survival analysis on these genes results in reducing the number of identified effective genes into 3 upregulated ( ITGAM , ITGAL and CD163 ) and 5 downregulated genes ( MCM2 , MCM3 , RFC4 , RFC5 and FEN1 ). Finally, drug sensitivity analysis was performed on these genes demonstrating complexity in drug-resistance due to the pattern of gene expression. This knowledge could potentially enable personalized treatment approaches based on individual patient responses due to the epigenetics and life style which affect gene expression pattern.

Choroidal metastasis: Impact of primary tumors and age on survival - a single center analysis

Purpose The purpose of this study was to determine whether patient’s age and their primary tumor type act as independent predictors of their survival after a diagnosis of choroidal metastasis. Methods This retrospective single-center study (August 2013 – August 2025) included 70 patients with choroidal metastases. Clinical data and multimodal imaging were extracted from medical records. Tumor volume was calculated from ultrasound measurements. Patients were grouped according to primary tumor origin (lung, breast, or other primaries). Survival was compared between these three primary tumor groups and according to age. Group differences in continuous variables were assessed using ANOVA/Kruskal-Wallis testing, and survival distributions were analyzed with Kaplan-Meier curves. Patients were categorized into younger (<59.6 years) and older (≥59.6 years) groups for Kaplan-Meier survival analysis. Univariate and multivariate Cox regression models were performed to identify independent predictors of overall survival. Results Median overall survival was 71.6 weeks. Statistically significantly longer survival was found in patients <59.6 years (median 88.3 weeks; estimated 3-year survival 33.2%) compared with those ≥59.6 years (median 41.9 weeks; 3-year survival 16.8%; p = 0.037. Primary tumor group analysis showed a median survival of 94.4 weeks for breast cancer (estimated 3-year survival 38.3%), 52.7 weeks for lung cancer (9,5%), and 40.6 weeks for other primaries (19.7%). However, this difference was not statistically significant (p = 0.269). In multivariate Cox regression analysis, age ≥ 59.6 years (HR 2.10; p = 0.016) and >1 extraocular metastases elsewhere in the body (HR 2.53; p = 0.029) were independent predictors of mortality. Conclusion Survival in choroidal metastases is strongly driven by age and the number of extraocular metastatic sites at presentation. Our findings suggest that younger age and a lower metastatic burden are most reliable indicators for a more favorable prognosis.

The economic burden of Type 2 Diabetes by social determinants of health: A systematic review

Background The unequal distribution of resources in society generates social gradients that translate into health inequalities and differential use of health care resources and their costs. Non-medical factors such as employment, income, ethnicity and education impact the prevalence and treatment outcomes of patients with type 2 diabetes mellitus (T2DM); however, there is a scarcity of articles assessing the relationship between health inequalities and the economic costs of treatment. Therefore, we conducted a systematic review of published studies examining the cost differences of treating T2DM across social determinants of health (SDH). Methods We systematically searched MEDLINE, Embase, PsycINFO, EconLit, and NHS EED for original peer-reviewed articles that provided cost differences of treating T2DM by SDH: education, income, employment, residency and ethnicity. We grouped the studies by each SDH and calculated the percentage differences where possible between the lowest and highest ends of the gradient (education, income and employment). Residency was categorised as rural vs. urban and ethnicity as white or general population vs other ethnic minorities. Results We included 19 articles retrieved internationally from varying healthcare systems. Results were contextualised given the healthcare financing model. In countries with high out-of-pocket expenses, Black and Hispanic ethnic backgrounds and rural residence were associated with lower direct health care and costs likely to be determined by ability to pay rather than clinical need. Indirect costs such as lost productivity due to absenteeism were also lower in unemployed, and lower income groups. Conclusions There are evident health disparities in the direct and indirect economic consequences of T2DM. The effect of decreased healthcare use and costs on treatment outcomes needs to be further explored to inform policies to ensure healthcare delivery is based on clinical need rather than socio-economic factors.

Prevalence, risk factors, and management practices of premenstrual syndrome among female university students in Lebanon: An observational cross-sectional study

Background Premenstrual syndrome (PMS) is common among women of reproductive age and may impair quality of life, academic performance, and social functioning. In Lebanon, data on PMS and its management among female university students remain limited. This study aimed to estimate the prevalence of PMS among female university students in Lebanon, identify risk factors, and recognize adopted management practices and their perceived effectiveness. Methods An observational cross-sectional study was conducted among female students at Beirut Arab University, Lebanon, between April 20 and May 8, 2026. Data were collected using a self-administered questionnaire assessing sociodemographic, menstrual, lifestyle, behavioral, psychosocial, and management-related factors. Premenstrual symptoms, social media addiction, and perceived stress were assessed using the Premenstrual Symptoms Screening Tool, Bergen Social Media Addiction Scale, and Perceived Stress Scale-4, respectively. Binary logistic regression was used to identify factors associated with moderate-to-severe PMS and premenstrual dysphoric disorder (PMDD). Results Among 1,062 participants, 497 participants screened positive for moderate-to-severe PMS (46.8%), and 175 had symptoms consistent with PMDD based on the PSST (16.5%). The most frequently reported moderate-to-severe symptoms were physical symptoms (69.7%), fatigue or lack of energy (68.8%), depressed mood or hopelessness (67.8%), and overeating or food cravings (66.2%). Academic absenteeism was reported by 37.2%. More than one-third of the participants with moderate-to-severe PMS and PMDD (39%) reported using pharmacological management practices, mainly non-steroidal anti-inflammatory drugs. However, less than half of medication users perceived these treatments as being very effective (45.8%). Higher odds of moderate-to-severe PMS were observed among current smokers (AOR = 1.61, P = 0.003), those with heavy menstrual bleeding (AOR = 1.58, P = 0.03), high meal skipping (AOR = 1.38, P = 0.03), high fast-food consumption (AOR = 1.34, P = 0.04), higher social media addiction scores (AOR = 1.31, P < 0.001), and higher perceived stress scores (AOR = 1.48, P < 0.001). Conclusion PMS was common among female university students in Lebanon and a considerable proportion experienced PMDD. In fact, lifestyle, behavioral, and psychosocial factors were associated with moderate-to-severe symptoms, highlighting the need for university-based awareness, screening, and counseling strategies that address modifiable risk factors and support appropriate management of premenstrual symptoms.

Vehicle routing optimization and algorithms for instant delivery under customer loss mechanism

In the field of instant delivery, the mismatch between delivery resources and customer demands has led to increasingly significant customer losses. To address this issue, this study introduces the customer loss mechanism and constructs an evaluation function to screen out resource-intensive customers, thereby clarifying the scope of delivery services. Based on this, this study establishes the vehicle routing optimization model under the customer loss mechanism with the objective of minimizing the sum of vehicle fixed costs, variable routing costs, and time window penalty costs. An improved genetic algorithm is employed to solve this model. Case study results demonstrate that the improved genetic algorithm outperforms traditional genetic algorithms and tabu search algorithms in convergence speed, optimization capabilities, and stability, reducing total delivery cost by 36.25% and 4.18%, respectively, with zero delivery violations. Regarding model performance, when proactively excluding 8.33% of customers, the total delivery cost is reduced by 17.18%, primarily driven by the reduction in fleet size. Furthermore, large-scale experiments reveal a pronounced leverage effect: excluding a mere 5% of marginal customers counter-intuitively reduces both fleet size and travel distance, while a 10% loss yields an 18.39% total delivery cost reduction with zero violations, proving that the mechanism precisely screens out inefficient nodes rather than arbitrarily rejecting them. Sensitivity analysis further confirms the model’s robustness across varying resource tightness, demonstrating that proactive customer loss is a feasible and effective strategy for improving resource utilization through precise resource focusing.

Exploring the mechanism of Shuangyu Granule in regulating immune-inflammatory responses in influenza through UPLC-Orbitrap-MS/MS, GC-MS, and network target analysis

Influenza, an acute respiratory infectious disease caused by the influenza virus, remains a significant challenge for prevention and treatment due to rapid viral mutation and high pathogenicity. Traditional Chinese Medicine (TCM), including Shuangyu Granule (SYKL), has demonstrated efficacy in managing influenza. This study aimed to systematically identify the chemical components of SYKL in vitro and its absorbed constituents in vivo, and to preliminarily explore its potential mechanism in regulating influenza-related immune inflammation. UPLC-Orbitrap-MS/MS and GC-MS were used to characterize SYKL’s chemical profile, identifying 148 in vitro components and 21 prototype absorbed blood components. Network target analysis, integrated with single-cell RNA sequencing (scRNA-seq) data from influenza patients, predicted that the absorbed components may target multiple immune-inflammatory regulatory genes across various immune cell types. Molecular docking suggested favorable predicted binding potential between these components and target proteins. Experimental validation using poly(I:C)-induced inflammatory models in both RAW264.7 macrophages and mouse bone marrow-derived macrophages (BMDMs) showed that the absorbed components—loganic acid, 8-epiloganic acid, calycosin, atractylodin, eucalyptol, secoxyloganin, and paeoniflorin—significantly reduced mRNA expression of immune-inflammatory genes (DUSP6, MAPKAPK2, NOD2) and inhibited secretion of TNF-α, IL-6, IL-8, and NO. These findings suggest that SYKL may alleviate influenza-associated inflammation through multi-component, multi-cell, and multi-target pathways, highlighting its potential in modulating excessive immune responses in influenza.

A standardized imaging and analysis workflow for quantitative evaluation of cutaneous neurofibromas in Nf1-KO mice

Neurofibromatosis type 1 (NF1) is an autosomal dominant disorder in which cutaneous neurofibromas (cNFs) represent one of the most common and burdensome manifestations. No approved pharmacological treatment exists. Preclinical studies are essential to evaluate candidate therapies, but reliable outcome and endpoint measures for cNFs in animal models remain limited. We developed and validated a standardized methodology to assess drug efficacy in the Prss56Cre Nf1-KO mouse model which recapitulates key features of cNFs. In this model, Nf1 inactivation and tdTomato (Tom) reporter expression were specifically targeted to Schwann cells (SCs) responsible for cNF development. This approach enables real-time monitoring, isolation, and manipulation of tumor SCs at any time. We defined macroscopic (tumor count, total Tom + fluorescent surface area, fluorescence intensity) and microscopic (cell-type composition defined by immunolabeling with a panel of specific markers, area quantification) endpoints, developed dedicated ImageJ scripts for automated image analysis, and compared the results with those obtained using the conventional manual method. Both automated measurements showed excellent reproducibility (ICC = 1) and strong correlation with manual analysis (Spearman’s coefficient > 0.90), while significantly reducing analysis time (up to 100-fold faster). Bland–Altman analyses confirmed the absence of systematic bias compared with manual scoring. The standardized image naming and metadata integration further facilitated data consolidation and statistical analysis. This validated approach provides a reliable, reproducible, and time-efficient framework for evaluating drug effects on cNFs in preclinical studies. It establishes a foundation for robust efficacy testing of candidate therapies, facilitates cross-study comparability, and accelerates therapeutic development and clinical translation.

An automated approach to extracting head and brain circumference from MRI datasets

Head circumference is a fundamental biometric parameter for brain growth in both the clinical pediatric setting and in developmental neuroscience. However, the gold standard for obtaining head circumference by manual tape measurement is notoriously error-prone. Further, while it is known that the growth trajectories of head and brain differ over time, a systematic comparison of these two parameters as a function of age does not yet exist. We developed a new and automated algorithm for obtaining head and brain circumference from MRI data. The algorithm mimics manual head circumference measurement by placing a convex hull around axial slices which must intersect with predefined anatomical landmarks. Several differently-tilted iterations are run and results are combined. In addition to obtaining head circumference, the approach can also be applied to gray matter only, providing “brain circumference” (gray matter hull perimeter) at the same level as head circumference. To assess validity, we used T1-weighted 3D datasets (n = 153) with available, manually measured head circumference values (age range 0–226 months [0–18.8 years]). To assess test-retest reliability, a second dataset (n = 3 with 40 scans each) was used. When compared with the current gold standard (manual measure), high validity was demonstrated for the new approach, with no systematic bias. The algorithm also showed a very high reliability across multiple measurements. Developmental trajectories of both head and brain circumference were generated and compared. In summary, the algorithm represents a valid and reliable method for the automated determination of head as well as brain circumference. It offers an objective way to assess these parameters in retrospect and prospectively, and may shed light on specific clinical situations where they differ, such as in the presence of enlarged subarachnoid spaces.

Intrathecal pump refills at home or at the hospital: Protocol for a randomized controlled crossover trial—The IMPROVE study

Background Intrathecal drug delivery (IDD) offers a therapeutic option for patients suffering from refractory pain or severe spasticity. By allowing targeted and continuous infusion directly into the intrathecal space, IDD bypasses the blood-brain barrier and enhances therapeutic effectiveness of the drug. Following the implantation of an IDD pump, the most commonly performed postoperative maintenance procedure is the pump refill (at regular intervals). This process can be burdensome for patients, affects their comfort, and carries significant risks. The current aim of this study is to evaluate whether intrathecal pump refills performed at home provide a difference in patient comfort compared to refills conducted in the hospital. Methods The IMPROVE study is a monocentric, randomized controlled crossover trial, including 82 patients. For this study, each patient will undergo four intrathecal pump refill procedures (two at home and two in the outpatient clinic) allocated in a randomized order. The primary objective of this study is to determine whether at-home refills provide a difference in patient comfort compared to hospital-based refills. Secondary objectives include assessing differences in quality of life, pain, stress, anxiety, self-efficacy, caregiver burden, patient preferences, safety, and overall cost-effectiveness between the two settings. Patients will be followed over the course of four intrathecal pump refills, which is estimated to span approximately one year. Discussion Within the IMPROVE project, pump refills will be performed through hospital at home. If at-home intrathecal pump refills prove more comfortable for patients and cost-effective for society, this would strengthen the patient-centred care model and support adopting this approach as the new standard treatment for IDD patients. A graphical abstract is provided in the supplementary materials (S1 Fig). Trial registration Details on the study site can be found at ISRCTN with identifier: ISRCTN18031921; [href: https://doi.org/10.1186/ISRCTN18031921 ] https://doi.org/10.1186/ISRCTN18031921 . The trial was registered in the ISRCTN registry on 18 November 2025.

Editorial Note: A risk-averse sustainable perishable food supply chain considering production and delivery times with real-world application

Enhancing missense variant classification in predicted intrinsically disordered regions

Classifying disease-causing missense variants in intrinsically disordered regions (IDRs) remains a significant challenge, with over 25% of known deleterious variants occurring in these regions. Existing in silico missense variant predictors that predict variant classification generally perform better in ordered regions of the protein, limiting their effectiveness. To address this, we developed a machine learning methodology that integrates global IDR conformation (gIDRc) features from ALBATROSS, phase separation (PS) features from BioPython, and 1024-dimensional protein embeddings from ProtTransBertBFD generated for both wild-type (WT) and mutant IDR sequences. IDR boundaries were defined using the AlphaFold-RSA predictions, which identifies disordered regions based on AlphaFold2 pLDDT scores and relative solvent accessibility. Using ClinVar variant classifications as ground truth, AlphaMissense, EVE, and ESM1b were the highest scoring unsupervised in silico missense predictors for IDR variants. Our baseline model, using only IDR-specific features achieved competitive performance on the hold-out test set with a PR-AUC of 0.817. Critically, when these IDR features were combined with these methods we saw significant overall improvement. The AlphaMissense-Enhanced model increased its PR-AUC from 0.807 to 0.919. Similarly, ESM1b-Enhanced improved PR-AUC from 0.679 to 0.845 and EVE increased from 0.591 to 0.910. These results demonstrate the effectiveness of our enhancements for classifying missense variants in IDRs and highlight its ability to complement existing in silico missense predictors.

Identifying cluster profiles based on barriers and facilitators to physical activity during COVID-19 confinement: A cross-sectional study using machine learning analysis

Social restrictions, such as confinement periods, tend to reduce physical activity (PA) levels. However, sociodemographic factors may influence specific barriers and facilitators to PA during such periods. This study aimed to identify cluster profiles of individuals based on barriers and facilitators to physical activity (PA) during COVID-19 confinement. Brazilian adults participated in a cross-sectional online survey. The questionnaire collected demographic data, PA levels, sedentary behavior (SB), and perceived barriers and facilitators for PA. During data preprocessing, correlated barriers and facilitators related to a similar topic were aggregated. Using machine learning analysis, the K-modes evaluated by the Silhouette Score were used for barriers and the ROCK evaluated by the Silhouette Score was used for facilitators. The barriers model produced well-defined profiles, whereas the facilitators model did not. The facilitator model generated clusters with multiple negative silhouette coefficients and exhibited a significantly less cohesive cluster structure. Therefore, only the barriers-based model was used for further analysis. The best model generated eight clusters, each named according to the most frequent barriers in the group, such as “Inactive depressive women”, “Active depressive women” and “Super active”. The depressive clusters presented more barriers to PA, three barriers each one. Significant differences in PA and SB were observed across clusters. This work highlights the novelty of using unsupervised machine learning to uncover latent subgroups based on multiple concurrent barriers. In conclusion, tailored home-based and outdoor strategies should be developed, particularly targeting individuals with depressive symptoms and those facing significant time constraints.

Age-stratified prognostic performance of hematologic inflammatory indices for 30-day mortality in emergency department patients with PCR-confirmed COVID-19: A cohort study from the pre-vaccination pandemic era

Background During the pandemic era, rapid and accessible prognostic tools were essential to support clinical decision-making for emergency department (ED) patients presenting with acute infectious symptoms. Hematologic inflammatory indices derived from complete blood count (CBC) parameters, such as the systemic immune-inflammation index (SII), systemic inflammatory response index (SIRI), and pan-immune-inflammation value (PIV), have been increasingly investigated for risk stratification. This retrospective cohort study evaluated the age-stratified prognostic performance of these indices for 30-day mortality in ED patients during the pandemic period. Methods This retrospective cohort study included adults presenting to a tertiary-care ED between March 1 and May 31, 2020. All included patients were retrospectively confirmed to have SARS-CoV-2 infection by RT-PCR. CBC-derived inflammatory markers (SII, SIRI, and PIV) were calculated at admission. The primary outcome was 30-day mortality; the secondary outcome was ICU admission. Age-stratified analyses (<65 and ≥65 years) were performed. Receiver operating characteristic (ROC) analyses, area under the curve (AUC) values, optimal cut-offs, and negative predictive values (NPVs) were determined; logistic regression models assessed independent associations with mortality. Results A total of 2,778 PCR-confirmed patients were included (mean age 47.8 ± 16.2; 58.7% male). Thirty-day mortality was 6.2%. In the overall cohort, SII, SIRI, and PIV demonstrated modest prognostic performance for mortality (AUCs: 0.663, 0.659, and 0.649, respectively). In patients <65 years, performance improved particularly for SII (AUC 0.727), with SIRI and PIV yielding AUCs of 0.676 and 0.677, respectively. Among patients ≥65 years, discrimination was lower (SII: 0.570; SIRI: 0.604; PIV: 0.588). Formal DeLong testing confirmed statistically significant age-related attenuation for SII (ΔAUC = 0.159; P = 0.0055), with non-significant trends for SIRI and PIV. As an exploratory secondary outcome, direct ED-to-ICU admission occurred in 2.9% of patients; this endpoint primarily reflects the institutional pandemic-era pathway of low-threshold ward admission with subsequent ICU escalation upon clinical deterioration. All indices demonstrated high negative predictive values, particularly in younger patients, indicating potential utility for identifying lower-risk individuals during high-volume pandemic ED operations. Conclusions Hematologic inflammatory indices obtained at ED presentation demonstrated age-dependent prognostic performance for 30-day mortality, with SII showing good discrimination and high negative predictive value (98.9%) in patients younger than 65 years and reduced discriminatory performance in elderly patients. These readily available and cost-effective parameters may support rule-out decisions for younger adults in emergency settings, while in elderly patients clinical assessment and comorbidity profiling should be prioritized over inflammatory marker interpretation.