Pacific States University

Hospital pharmacies require short-horizon demand forecasts that remain useful across thousands of medicines with widely different issue volumes. This study evaluates an artificial intelligence (AI)-enabled feature-engineering framework for adaptive decision support using 72 corrected finalised Secondary Care Medicines Data (SCMD) files published by the NHS Business Services Authority (NHSBSA) for April 2019 to March 2025. The analysis processes 22.67 million records into 522,813 product-month observations and 488,924 exact next-calendar-month positive-issue forecasting rows. Demand volatility is treated as an operational signal for disruption forecasting because unstable medicine flows can increase the risk of review delays, emergency purchasing or stock imbalance even when shortage labels are unavailable. Target construction, lag variables and expanding product summaries are calendar-indexed and use only information available at the forecast origin. Descriptive statistics, unit-root and trend diagnostics, temporal holdout validation, seven full rolling-origin refits, trend-aware baselines and paired Wilcoxon tests are used. The raw holdout target is extremely skewed (skewness 38.05), whereas log(1 + quantity) is approximately symmetric (skewness 0.04). Trust-network-informed Light Gradient Boosting Machine (LightGBM) achieves the lowest holdout root mean squared logarithmic error (RMSLE = 0.4832) and mean rolling-origin RMSLE (0.4910), while rolling means achieve the lowest weighted absolute percentage error (WAPE). A transparent monitoring score identifies 10.0

Large language models can generate fluent clinical summaries, yet their reliability is constrained by hallucinations, limited provenance, and safety risks when used in conjunction with automated image-based predictions. This study presents an end-to-end screening-oriented decision support pipeline that integrates a dual-path vision model with a domain-constrained retrieval-augmented generation framework to produce evidence-grounded explanations. The vision component, Attention-Guided Dual-Path FusionNet, combines a ConvNeXtV2 backbone for global semantic representation with a Swin Transformer branch for hierarchical local feature modeling, followed by attention-guided feature fusion for four-class CT image classification. The language component, ClinicGraphRAG, performs allowlisted retrieval over curated clinical sources with metadata-preserving provenance and generates constrained outputs using a grounded-only input interface. The dataset comprises 1,535 CT images expanded to 3,600 samples through class-specific, training-only augmentation under a leakage-safe stratified 5-fold cross-validation protocol, which may introduce bias due to reliance on synthetic data and limited sample diversity. The proposed model achieves a mean cross-validation accuracy of 98.74 ± 0.19, outperforming selected baseline architectures; however, these results are derived from a single curated dataset and may not generalize to heterogeneous clinical settings. The system is designed strictly for screening-oriented decision support and education, with explicit constraints to prevent diagnostic or treatment use, and requires external validation and broader cohort evaluation before real-world deployment.

Background and motivation: Drug shortages disrupt patient care and hospital operations, while public records lack inventory and procurement data. Problem statement: This study separately evaluates whether U.S. Food and Drug Administration (FDA) shortage histories can rank 90-day recurrence risk and how a status-derived risk proxy changes inventory-policy trade-offs in a synthetic simulator; it does not imply autonomous procurement. Methodology: The analysis used 1688 FDA records. Snapshot classification served only as a diagnostic. The main empirical task used a 90-day recurrence panel with historical-only features, temporal splits, leakage controls, group holdouts, and machine-learning models. Aggregate counts were compared with naive, rolling-mean, exponential-smoothing, and autoregressive integrated moving average (ARIMA) baselines. Separately, tabular Q-learning was evaluated in a disclosed synthetic inventory environment against economic order quantity (EOQ), min-max, safety-stock, threshold, and human-rule policies using stress tests, sensitivity analysis, effect sizes, and governance proxies. Findings: With 15 positive test windows, gradient boosting achieved receiver-operating-characteristic area under the curve (ROC-AUC) 0.831, precision-recall area under the curve (PR-AUC) 0.0465, and recall 0.667; logistic regression captured 40

A brain tumour is an abnormal or mass growth of cells in the brain. It can be benign or malignant. This problem, if not identified in its early stages, can be fatal to the patient. Detection of Brain tumours in their early stages has become a major challenge in the realm of healthcare. This research uses about 4600 images of brain tumour. Initially, the data labels are encoded and the images are normalized. Pre-existing Neural Network models such as ResNet-50 and AlexNet are trained to predict the presence of Brain tumour. The respective architectures have been used instead of the pre-trained model. In this study, the batch sizes of the defined models are varied and the performance is compared based on the model loss and accuracy. The AlexNet model resulted in a testing accuracy of 0.96, 0.95 and 0.96 for the batch sizes 16,32 and 64 respectively whereas a testing accuracy of only 0.56 resulted for a batch size of 8. But ResNet-50 model, performed well for all batch sizes of 8, 16, 32 and 64 yielding an accuracy of 0.94,0.69,0.97 and 0.96 respectively depicting itself to be a suitable model for brain tumour classification.

Background: Inclusion systems of special education have long struggled to provide individualized, transparent and equitable learning conditions. Traditional reactive models are not usually able to support learner variability, which is one of the roots of a global educational divide whereby millions of children with disabilities have inadequate access to quality education. As promising as an adaptive learning and intelligent tutoring tool may be, the black-box nature of Artificial Intelligence (AI) tends to erode trust and cast doubt on ethical aspects of automated decision-making. Explainable Artificial Intelligence (XAI) overcomes these drawbacks by improving the interpretability of algorithm processes. At the same time, Universal Design for Learning (UDL) offers a proactive pedagogical model that stresses variability of learners, their agency, and elimination of systemic barriers as further supported in the UDL 3.0 guidelines. Purpose: This review synthesizes the literature on XAI and UDL, demonstrating their complementary strengths and proposing an integrated XAI-UDL system to support the inclusive special education system. It also analyzes the applicability and implication of educational and research ecosystem of the United States in the promotion of this integrated system. Procedure: A systematic review of the literature was performed using Scopus, Web of science, ERIC, Pubmed and Google scholar to retrieve 2019-2026 publications. Search terms were: Explainable AI, Universal Design for Learning, special education, inclusive learning. The peer-reviewed journal articles, conference proceedings, and empirical studies were subjected to rigorous quality appraisal criteria, which resulted in a final dataset of 165 sources. Results: The synthesis demonstrates strong synergistic benefits from integrating XAI and UDL. XAI offers transparency and interpretability, builds trust, accountability, and learner agency, and UDL provides the pedagogical framework that is required to transform AI-driven insights into meaningful instructional practices. Empirical data show that digital literacy, learner self-efficacy, and system usability among students with disabilities greatly improved. Moreover, the educational and research ecosystem in the United States, which is characterized by strong policy frameworks, interdisciplinary research funding, technological innovation, and leadership on inclusive education, provides a highly conducive environment to the development, validation and large-scale application of the XAI-UDL framework. Federal requirements in favor of inclusive education, effective academia-industry partnership, and investment in the study of AI helps the ethical implementation of explainable systems in educational practice. Conclusion: The proposed XAI–UDL framework offers a viable pathway toward transparent, adaptive, and inclusive learning environments. Its congruency with the developed educational system and research capabilities of the United States further underpins its scalability and universal applicability. This synthesis offers crucial advice to researchers, educators, and policymakers to utilize AI in a responsible manner and enhance equity and inclusion in special education systems.