Journals
2026 EN
Tourkmani Ayla M. · AlHarbi Turki J. · Alghamdi Ahmad Abdullah
+3 more
Abstract Introduction Diabetic kidney disease (DKD) and diabetic nephropathy (DN) affect around 40% of diabetic patients but lack accurate risk prediction tools that include social determinants and demographic complexity. We developed and validated an ensemble machine learning model for three‐year DKD/DN risk prediction with deployment readiness. Methods We analysed 18 742 eligible adult type 2 diabetic patients from Prince Sultan Military Medical City (PSMMC) registry between 2019 and 2024 in Riyadh, Saudi Arabia. Using temporal patient‐level splitting, we developed a stacked ensemble model (LightGBM + CoxBoost) with several features including multiple literature‐informed imputed variables including family history, non‐steroidal anti‐inflammatory drug (NSAID) use, socioeconomic deprivation, diabetic retinopathy severity, and antihypertensive medications, imputed via Bayesian multiple imputation by chained equations (MICE) with external study priors. Primary outcome was incident/progressive DKD/DN within 3 years' timeframe. We assessed discrimination, calibration, model utilisation, and algorithmic fairness. Results The final model achieved excellent discrimination (receiver operating characteristic [AUROC] of 0.852, 95% CI 0.847–0.857) and near‐perfect calibration (slope 0.98, intercept −0.012) on multi‐trial validation. Decision curve evaluation demonstrated superior net benefit (+22 events prevented per 1000 patients at 10% threshold) compared to treat‐all strategies. Bootstrap validation showed minimal optimism in discrimination ( C ‐statistic optimism = 0.005). No algorithmic bias was detected across demographic subgroups (maximum |Δ‐AUROC| = 0.010). Prior sensitivity analysis confirmed validity and significance (AUROC variation ≤0.008). The model was engineered and deployed as an interactive web‐based application ( https://nephrarisk.streamlit.app/ ). Conclusions Our developed and demonstrated model provided accurate and well‐fair DKD/DN risk prediction with excellent calibration, allowing for better decision making with deployment as a web‐based research tool and framework for future prospective clinical validation. Further validation and testing are warranted from different centres and healthcare systems to increase confidence and dissemination of our model findings for better utilisation purposes in the future.
Journals
2026 EN
Hammad Mohammed Ali Mohammed · Elbarky Ismael · Biomy Reda
+3 more
ABSTRACT Background Heart failure (HF) with reduced ejection fraction (HFrEF) is characterized by progressive left ventricular (LV) remodeling and is often complicated by secondary (functional) mitral regurgitation (MR). Sacubitril–valsartan (S/V), an angiotensin receptor–neprilysin inhibitor (ARNI), has demonstrated superiority over enalapril in reducing morbidity and mortality in HFrEF. Its mechanism of action includes promoting LV reverse remodeling (LVRR), but the comprehensive effect on both LVRR and secondary MR, particularly using advanced three‐dimensional (3D) echocardiography, requires further elucidation. Objectives To evaluate the effect of 1‐year S/V therapy on LVRR and the severity of secondary MR in patients with HFrEF, utilizing both two‐dimensional (2D) and 3D echocardiographic parameters. Methods This prospective, single‐center study included 80 patients with HFrEF (LVEF ≤ 40%) initiated on S/V. Comprehensive 2D and 3D transthoracic echocardiography was performed at baseline 6 months, and 1 year of treatment. LVRR was assessed by changes in LV volumes (EDVI, ESVI), LVEF, and LV mass index (LVMI). MR severity was quantified using the effective regurgitant orifice area (EROA). Paired statistical tests ( t ‐test or Wilcoxon signed‐rank test) were used for pre‐ and posttreatment comparisons. Correlation and multivariable linear regression analyses were performed to identify predictors of LVRR. Results After 1 year of S/V therapy, patients with HFrEF demonstrated marked and consistent improvements in both MR and (LV) structure and function. EROA decreased from 0.36 ± 0.08 cm 2 to 0.24 ± 0.09 cm 2 (−34.6%, p < 0.001), accompanied by reductions in regurgitant volume (RV) (−32.6%, p < 0.001) and regurgitant fraction (−30.8%, p < 0.001). LVRR was evidenced by significant increases in LVEF (3D: +42.1%, p < 0.001, Cohen's d = 2.63) and reductions in LVEDV (−18.0%, p < 0.001) and LVESV (−31.8%, p < 0.001), as well as a 14.7% decrease in LVMI ( p < 0.001). Correlation analysis revealed that MR improvement closely paralleled LV remodeling, with ΔEROA strongly associated with reductions in LVESVI ( r = 0.774, p < 0.001), LVEDVI ( r = 0.558, p < 0.001), and LVEF improvement ( r = −0.781, p < 0.001). Multivariable regression identified lower baseline LVEF (3D) as the strongest independent predictor of LVRR ( β = −2.304, p < 0.001) Conclusions S/V therapy induces significant and comprehensive LVRR and a marked reduction in secondary MR severity in HFrEF patients. The strong correlation between these two effects suggests that MR improvement is primarily driven by LVRR. Lower baseline LVEF is a powerful predictor of the magnitude of LVRR.
Journals
2026 EN
Schmitt Alexander · Akin Ibrahim · Reinhardt Marielen
+9 more
Abstract Background This study investigates the prognostic impact of albumin, the urea‐to‐albumin ratio ( UAR ), and albumin‐to‐creatinine ratio ( ACR ) in patients with heart failure with mildly reduced ejection fraction ( HFmrEF ), since hypoalbuminemia, renal disease and malnutrition often coincide with heart failure ( HF ). Methods Consecutive patients hospitalized with HFmrEF at one university medical centre were retrospectively included from 2016 to 2022. Patients were stratified into quartiles based on albumin, the UAR , and ACR . The primary endpoint was all‐cause mortality at 30 months (median follow‐up), key secondary endpoint was long‐term HF ‐related rehospitalization. Results The study cohort comprised 2,061 patients with HFmrEF with a median albumin level of 32.4 g/L. Albumin levels, the UAR and ACR were predictive for the risk of long‐term all‐cause mortality, which was still observed after multivariable adjustment (albumin Q1 vs. Q4 : HR = 2.260; 95% CI 1.623–3.148; p = .001 / UAR Q4 vs. Q1 : HR = 1.507; 95% CI 1.071–2.119; p = .019/ ACR Q1 vs. Q4 : HR = 2.208; 95% CI 1.528–3.190; p = .001). However, neither albumin nor the UAR or ACR predicted the risk of HF ‐related rehospitalization (albumin Q1 vs. Q4 : HR = 1.117; 95% CI .678–1.842; p = .664 / UAR Q4 vs. Q1 : HR = 1.589; 95% CI .922–2.738; p = .095 / ACR Q1 vs. Q4 : HR = 1.112; 95% CI .624–1.981; p = .720). Conclusions Hypoalbuminemia is common in hospitalized HFmrEF patients. Low albumin levels, ACRs , and elevated UARs independently predicted long‐term all‐cause mortality, but not HF ‐related rehospitalization. The UAR and ACR did not provide a clinically significant predictive advantage over albumin levels alone. Trial Registration ClinicalTrials.gov Identifier: NCT05603390 (date of registration: 10.10.2020)
Journals
2026 EN
Sönmez Muhammet · Öksüz Halil İbrahim · Gök Seçkin
+1 more
ABSTRACT This study aimed to examine the effectiveness of ChatGPT‐generated read‐aloud plans compared to teacher‐prepared read‐aloud plans for third‐grade elementary school students, specifically focusing on their vocabulary and inference skills. A total of 93 students from a public elementary school participated in the study. The treatment group engaged in ChatGPT‐generated read‐aloud plans, while the control group participated in teacher‐prepared ones. A MANCOVA analysis was conducted to assess the impact of these activities on students' vocabulary and inference skills, controlling for potential covariates. Despite the adequate sample size and statistical power, the results revealed no significant differences between the treatment (ChatGPT‐generated) and comparison (teacher‐prepared) groups, and the observed effect size was small. These findings suggest that there was no significant difference between the read‐aloud plans generated by ChatGPT and those prepared by the teacher, indicating that both types of plans had a similar impact on students' vocabulary and inference skills under the current conditions. The implications of these results for educational practice and the use of AI tools in the classroom are discussed.
Journals
2026 EN
Ali Ahmed S. · Alhirsan Saleh M. · Elshazly Mahmoud
+10 more
ABSTRACT Artificial Intelligence (AI) is reshaping healthcare education, yet structured AI training within physiotherapy programmes remains uneven across the Middle East. We conducted a cross‐sectional, multi‐country online survey of 3195 undergraduate and internship‐level physiotherapy students from nine Middle Eastern countries (Egypt, Jordan, Lebanon, Libya, Palestine, Saudi Arabia, Sudan, Tunisia, and the United Arab Emirates). Using Technology Acceptance Model (TAM) constructs—perceived usefulness (PU) and perceived ease of use (PEOU)—we examined factors associated with AI acceptance and students' perceived barriers to AI integration in physiotherapy education. AI acceptance differed significantly by country (highest in the UAE and lowest in Tunisia), gender (male > female), academic level, GPA, and income ( p < 0.05). Prior AI workshop participation, use of specific AI tools (e.g., DeepSeek), perceived time‐management benefits, and trust in AI were associated with higher acceptance. In multivariable regression, these sociodemographic and AI‐exposure variables explained 24.4% of the variance in AI acceptance. These findings indicate substantial regional disparities in AI preparedness and access, supporting the need for equitable, competency‐based AI education and governance policies tailored to physiotherapy training across the Middle East.
Journals
2026 EN
Eissa Ahmed · Fredrik Kismul Jan · Pentz Atle Bråthen
+9 more
ABSTRACT The human auditory system rapidly distinguishes between novel and familiar sounds, a process reflected in mismatch negativity (MMN), an electroencephalogram (EEG)‐based biomarker of auditory novelty detection. MMN is impaired in psychiatric conditions, most notably schizophrenia (SCZ), yet the neuronal mechanisms underlying this deficit remain unclear. Here, we combined computational modelling and genetic analyses to investigate how SCZ‐associated cellular abnormalities affect auditory novelty detection. We developed an integrate‐and‐fire spiking network model capable of detecting four types of auditory novelty, including stimulus omission. Based on assumptions of short‐term depressing synapses between the subpopulations of the network and the existence of neuronal inputs that are phase‐locked to the rhythm of the recently experienced stimulus sequence, we showed that the model reliably reproduced MMN‐like novelty detection and allowed systematic testing of SCZ‐related cellular alterations. We also demonstrated that the required phase locking can theoretically be achieved in a synfire chain network exhibiting spike‐timing dependent plasticity (STDP) in its feedback synapses that becomes entrained to the rhythmic stimulus. Simulations of our novelty‐detecting network revealed that both reduced pyramidal cell excitability, linked to ion channel dysfunction, and decreased spine density impaired novelty detection, with the latter producing stronger deficits. Our work provides a flexible spiking network model of auditory novelty detection that can link cellular‐level abnormalities to measurable MMN deficits, improving their mechanistic interpretation and helping to explain the heterogeneity of SCZ.
Journals
2026 EN
Roueinfar Mina · Mohammadzadeh Pardis · Schwerdtfeger Luke A.
+2 more
ABSTRACT Stress activates the hypothalamic–pituitary–adrenal (HPA) axis, a neuroendocrine system that regulates responses related to feeding, reproduction, and aggression, among other homeostatic functions. Stressors significantly impact gene expression along the HPA axis and in hypothalamic nuclei that drive it, including the paraventricular nucleus (PVN). To identify genetic regulators of stress responses in the PVN, adult mice underwent 2 h of multi‐modal stress before gene expression profiles were analyzed using bulk RNA sequencing. A transcription factor zinc finger and BTB domain containing 16 (Zbtb16), also known as PLZF, was identified as a stress responsive, glucocorticoid receptor (GR) target in the PVN. Zbtb16 mRNA expression was increased by two‐fold in male and female mice within 2 h of restraint stress or injection of a synthetic glucocorticoid, dexamethasone (DEX). Immunohistochemistry (IHC) confirmed Zbtb16 protein expression and localization in the PVN following 20 min of restraint stress and 4 h of recovery. Cellular analyses revealed that Zbtb16 was highly expressed in CRH neurons in the PVN, neurons routinely activated post‐stress as indicated by colocalization with c‐FOS. Adult mice were also exposed to an immune stress by injection of tumor necrosis factor alpha (TNFα) to assess Zbtb16 regulation. Expanded analyses indicated that the cell specificity of Zbtb16 expression was region‐specific, colocalizing with CRH neurons in the mid‐PVN but more in astrocytes surrounding the PVN. These findings identify Zbtb16 as a glucocorticoid‐ and cytokine‐inducible transcriptional regulator with region‐ and cell type‐specific roles in PVN stress circuitry.
Journals
2026 EN
Mithani Karim · Ibrahim George M.
Abstract The rapid expansion of neuromodulation therapies for drug resistant epilepsy has introduced a growing risk of normalization of deviance (NoD). This occurs when incremental deviations from established procedures are iteratively reinterpreted as the care standard in the absence of immediate adverse outcomes. While motivated by an urgent need to develop novel therapies to improve the quality‐of‐life of persons with epilepsy, NoD risks harm to the very individuals these neurotechnologies are designed to help. In this commentary, we borrow from organizational safety research to highlight systemic, institutional, and cultural factors that enable NoD in the pursuit of neuromodulation for epilepsy. We highlight the dual imperative of driving forward innovations in neurotechnologies while prioritizing patient safety. Finally, we propose safeguards against NoD including institutional review, centralized data collection, and cultural shifts that prioritize safety and mitigate risk.
Journals
2026 EN
Dinger Thiemo F. · Mithani Karim · AlHasan Hosni Abu
+42 more
Abstract Objective Although vagus nerve stimulation (VNS) is a well‐established neuromodulation therapy for drug‐resistant epilepsy, treatment outcomes remain heterogeneous. One possible source of variability lies in differing interpretations of seizure frequency ratings (SFRs). This study examined interrater reliability (IRR) in SFRs between (1) retrospective clinician–clinician chart reviews and (2) prospective caregiver–clinician reports, and explored sources of disagreement. Methods Data were collected from the CONNECTiVOS database. In the retrospective cohort ( n = 254), two clinicians independently reviewed medical records and rated seizure frequency across multiple timepoints. In the prospective cohort ( n = 214), caregivers and clinicians independently reported SFR in children treated with VNS. IRR was assessed across different measurement thresholds, and potential causes of disagreement were analyzed. Results Clinician–clinician agreement in retrospective chart reviews was excellent (intraclass correlation coefficient [ICC] > .90, Cohen κ > .80), with 18.8% divergent ratings and 4.8% exceeding the reliable change index. Disagreement was significantly associated with higher mean seizure frequency at baseline ( p = .004) and at postoperative timepoints ( p < .001). In the prospective caregiver–clinician comparison, agreement for absolute seizure frequency was poor (ICC < .50), with discrepancies in 86.5% of cases, although only 1.8% were statistically significant. When rating pairs diverged, clinicians more often reported lower absolute seizure frequencies ( p = .002) and greater relative seizure reductions ( p = .023) and were more likely to classify patients as achieving a 90% reduction ( p = .043). Significance This study highlights interrater variability in both retrospective and prospective SFR assessments, a finding systematically related to baseline seizure frequency. Coarser classifications (e.g., 50% or 90% seizure reduction) may improve agreement but reduce clinical nuance. Future efforts should focus on structured, patient‐centered documentation and the development of objective outcome measures in VNS evaluation, particularly for children with high seizure burden.
Journals
2026 EN
Abbas Muhammad John · Khan Muhammad Attique · Dillshad Veena
+5 more
ABSTRACT Alzheimer's Disease (AD) is a progressive neurodegenerative disease diagnosed through cognitive impairment, and an early diagnosis is essential to improve treatment and care options. Current diagnostic approaches of AD, such as neuroimaging, cognitive assessments and biomarker research, are lengthy, vague and not sufficient to assess the early stages of AD. To address these problems, we introduced a novel deep learning model, ‘NeuroMixFormer’, which is based on a mixture‐of‐experts architecture for AD classification from MRI. The proposed multistage architecture employs a dynamic routing mechanism and four expert blocks per stage, each integrating dense connectivity with a spatial and channel attention module for feature extraction. To improve early feature learning, auxiliary classifiers are incorporated at intermediate stages of training. Evaluations on three datasets (ADNI, Mendeley and Kaggle Augmented Alzheimer's MRI) demonstrated the proposed model's superior performance over existing deep learning architectures and state‐of‐the‐art methods, achieving up to 99.48% accuracy on the Kaggle dataset, 90.28% on the Mendeley dataset and 99.86% on the ADNI dataset, respectively. Ablation studies confirmed the importance of dual‐attention mechanisms, and expert routing analysis showed clear specialisation patterns across AD stages, improving both classification accuracy and interpretability. These results underscore the effectiveness and generalisability of NeuroMixFormer in automated dementia detection, highlighting its potential to support early and precise AD diagnosis. However, the high computational cost and inference time associated with this high accuracy limit the practicality of the proposed approach in clinical settings.