Objectives: Medication errors are a major source of preventable harm. We systematically reviewed workflow integrated artificial intelligence (AI) decision support (machine learning [ML], natural language processing, computer vision, and large language models [LLMs]) aimed at detecting, intercepting, or preventing medication errors, and considered implications for Saudi Arabia.Methods: We searched MEDLINE (PubMed), Embase, Scopus, Web of Science Core Collection, IEEE Xplore, and the Cochrane Library for English language studies published from January 2010 to February 2026. Eligible studies used real clinical/pharmacy data and evaluated an AI component designed to prevent or detect medication errors across prescribing/order verification, dispensing, administration, reconciliation, or medication directions. Two reviewers screened and extracted data independently; risk of bias was assessed with PROBAST, QUADAS2, or ROBINS-I.Results: A total of 16 studies met inclusion criteria. Nine evaluated prescribing/order verification tools (including look alike/sound alike and outlier detection), 3 evaluated administration support (including computer vision verification), 2 evaluated reconciliation/triage, 1 evaluated dispensing verification, and 1 evaluated an LLM for medication directions. Performance was often high for constrained tasks, but external validation, calibration, and explicit alert threshold governance were uncommon. Deployed evaluations suggested potential reductions in errors or near misses, but most impact evidence was observational, and alert burden/overrides were inconsistently reported. Heterogeneity precluded meta-analysis; using GRADE, certainty of evidence for error reduction was generally low.Conclusion: Workflow integrated AI decision support can augment medication safety, particularly via high plus pressure variation surveillance and task specific verification, but the current evidence base provides limited causal certainty for harm reduction. For Saudi Arabia, scale up should be phased and governed through interoperability, clinical decision support governance, human factors evaluation, Saudi Food and Drug Authority aligned software as a medical device pathway, and post deployment monitoring.
Objective: To evaluate AI readiness among paramedics working within the Saudi Red Crescent Authority (SRCA) and examine demographic and professional characteristics related to their readiness.Methods: This cross-sectional study was conducted among SRCA paramedics. Data were collected through a structured online questionnaire covering demographic and professional features and the validated Medical Artificial Intelligence Readiness Scale for Medical Students (MAIRS-MS), which assesses four dimensions: Cognition, Ability, Vision, and Ethics. Comparisons between groups were performed with the Mann-Whitney U and Kruskal-Wallis tests, while correlations were assessed using Spearman's rank correlation, and a multiple linear regression model examined factors independently related to AI readiness.Results: A total of 262 paramedics participated. The mean overall AI readiness score was 72.55 (SD 21.79) on the 110-point scale, reflecting a moderate degree of readiness. The Ethics and the Cognition domains had the highest (3.57/5) and the lowest (3.15/5) mean item scores, respectively. Participants reported greater confidence in AI's ethical and practical aspects than in understanding AI concepts and terminology. Prior AI training was the only factor independently associated with higher AI readiness (B=10.13, 95% CI 3.45–16.81; P=.003).Conclusion: SRCA paramedics demonstrated moderate AI readiness, with greater confidence in AI's ethical and practical application than in its conceptual and technical foundations. Prior AI training was the only independent factor associated with higher readiness, highlighting the potential of structured AI education before the broader implementation of AI in EMS.
Objectives: To report the characteristics, severity, and treatment of drooling in cerebral palsy (CP) children.Methods: This retrospective observational study included CP children with active complaints of drooling from the Drooling Clinic, Sultan Bin Abdulaziz Humanitarian City (Riyadh, Saudi Arabia). Drooling severity was assessed via the Modified Drooling Questionnaire (MDQ). The functional motor and manual ability levels were classified via the Gross Motor Function Classification System (GMFCS) and Manual Ability Classification System (MACS).Results: In total, 122 patients were identified. Mean age was 6.28 ± 2.71 years, and 67/122 (54.9%) were boys. Gastroesophageal reflux disease was noted in 23/122 (18.9%). Drooling was anterior for most (n = 103/120, 85.8%), 82/85 (96.5%) had spastic CP, and 49/82 (59.8%) had quadriplegia. Level III (n = 36/110, 32.7%) on GMFCS and level V (n = 53/110, 48.2%) on MACS were the most common. Bilateral involvement was present in 106/110 (96.4%). For drooling severity assessed via the MDQ (mean = 17.00 ± 4.03; range = 11–34), majority experienced mild (n = 109/122, 89.3%) or moderate (n = 12/122, 9.8%) drooling. Overall, 36/122 (29.5%) patients received pharmacological treatment, among whom glycopyrrolate and scopolamine patches were the most frequently used, each of which was administered to a similar proportion of patients (n = 21/36, 58.3%).Conclusion: Drooling remains a common problem in CP children, particularly those with greater motor and manual impairment (GMFCS III, MACS V). It is most often anterior with mild-to-moderate severity, although one patient presented with severe drooling.
Objective: To explore peer-reviewed research on the application of Explainable Artificial Intelligence (XAI) techniques for histopathology-driven cancer risk evaluation, and to address the limitations delaying its transition into routine clinical pathology practice (such as model reliability, clinical interpretability, and governance oversight).Methods: We conducted a systematic review with quantitative meta-analysis of studies evaluating explainable artificial intelligence (XAI) in histopathology. Searches were performed in Web of Science, Scopus, PubMed, and Google Scholar through November 2025. Eligible studies were assessed for diagnostic performance, explainability approaches, and methodological quality. Of the 47 included studies, 9 provided sufficient quantitative data for meta-analysis. Post-hoc visual attribution approaches were the most frequently reported explanation methods in the included literature, with Class Activation Map (CAM) and Gradient-weighted (Grad) -CAM being the most prevalent.Results: Nine studies provided sufficient quantitative data for meta-analysis. The combined area beneath the curve was 0.962 (95% confidence intervals (CI): 0.909–0.985), despite significant inconsistency among the selected studies (evolution of heterogenicity (I2) = 97.1%). The investigation of subgroup revealed that models tested at the slide level performed better and more consistently than those examined at the patient level, suggesting that the unit of analysis has a significant effect on reported accuracy and heterogeneity.Conclusion: Future clinical translation of XAI in histopathology must prioritize explanation methods that are meaningful to pathologists, conduct usability testing in real diagnostic workflows, ensure prospective validation, and establish implementation within quality-management and governance frameworks.
Alzheimer's disease (AD) is a progressive neurodegenerative disorder requiring timely diagnosis. This structured narrative review examines artificial intelligence (AI) and deep-learning applications (DL) for MRI-based diagnosis of AD and mild cognitive impairment (MCI). PubMed and Web of Science were searched between April-June 2026; Scopus and ScienceDirect supported supplementary discovery. Studies were purposively selected using predefined eligibility criteria, with 42 primary studies summarized. Convolutional neural networks, transformers, transfer learning, and multimodal approaches frequently reported promising classification performance. However, findings varied across datasets, tasks, validation procedures, and populations, while high accuracies were often derived from curated datasets and internal validation. Limited external and prospective validation, potential over fitting or data leakage, class imbalance, and insufficient demographic and scanner diversity restrict clinical generalizability. Independent multicenter and real-world assessments are required prior to routine clinical adoption.
