Saudi Cultural Missions Theses & Dissertations
Permanent URI for this communityhttps://drepo.sdl.edu.sa/handle/20.500.14154/10
Browse
63 results
Search Results
Item Restricted The Accuracy of Diagnosing Salivary Gland Diseases by Artificial Intelligence: Systematic Review(Saudi Digital Library, 2025) Aljohani, Wejdan; Seoudi, Noha1.1 Purpose Artificial intelligence (AI) is increasingly applied in the diagnosis of salivary gland diseases, particularly Sjögren’s syndrome (SS) and salivary gland tumours (SGTs). This review aimed to evaluate the diagnostic performance of AI models in these two disease categories and identify converging patterns, limitations, and research gaps. 1.2 Method A systematic literature search was conducted in PubMed, Scopus, and Google Scholar over the past two decades (2005-2025) using predefined inclusion and exclusion criteria. Data extraction captured study design, input modality, AI model type, performance metrics (sensitivity, specificity, accuracy, AUC). Quality analysis was performed using JBI tool. Results were stratified by disease group (SS vs SGTs) and AI model type (Machine learning vs Deep learning). 1.3 Results A total of 19 studies were included from the 221 initially retrieved. Most of the included studies were assessed as moderate risk of bias, with only three low-risk and one high-risk. In SS studies , ML models showed excellent performance when applied to structured data. Logistic Regression emerged as the best-performing architecture, achieving accuracies up to 94% with AUC values ranging from 0.88 to 0.96. DL models on histopathology ranged from weak performance in baseline Residual CNNs (ResNet) (50% accuracy) to excellent outcomes with custom architectures such as CTG-PAM (100% across sensitivity, specificity, and accuracy). In SGTs, ML models on imaging inputs showed moderate ability, with Logistic Regression achieving 78–84% accuracy (AUC up to 0.91) and ultrasound reporting lower sensitivity but good specificity. DL approaches outperformed ML, particularly hybrid CNN–Transformers on MRI (85% accuracy, AUC 0.96; Liu et al., 2023) and Vision Transformers on ultrasound (87% accuracy, AUC 0.93; He et al., 2025). CNNs were more variable: Inception showed consistent results (73–85% accuracy, AUC up to 0.91), while ResNet and Densely Connected CNN (DenseNet) performance fluctuated widely even within the same input modality. 1.4 Conclusion AI demonstrates high potential in salivary gland disease diagnosis, with structured data input and custom-made models and advanced DL architectures yielding the most promising results. However, heterogeneity in input modalities and model design limits comparability, underscoring the need for standardised, multicentre validation.2 0Item Restricted Employee Readiness for AI Adoption in Riyadh’s Healthcare Sector: Perceptions and Organizational Support(Saudi Digital Library, 2025) Almutairi, Hadeel; Cui, QinquanArtificial intelligence (AI) is widely recognized as a significant driver of digital transformation across several domains, with the healthcare sector identified as one of the most influenced sectors. This research assesses employee readiness for AI among healthcare professionals in Riyadh, Saudi Arabia, with particular attention paid to perceptions (perceived usefulness and ease of use) and organizational support, including training and management support. This study employed a quantitative, cross-sectional, and correlational design. A survey was administered to evaluate employee readiness levels and potential predictors of AI readiness. A total of 120 employees participated with overall readiness (M = 4.20, Var=0.64). The regression explained 39.4% of the variance in readiness, with perceived usefulness (B = 0.44, p < 0.001) and training (B = 0.40, p < 0.001) contributing positively to readiness, while management support contributed negatively (B = -0.17, p = 0.011), and ease of use was not significant (B = 0.05, p = 0.574). Independent t-tests and ANOVA confirmed no significant differences in readiness by gender (p = 0.40), job type (p = 0.44), or years of experience (p = 0.56). The results showed that perceived usefulness and training were the strongest predictors of employee readiness for AI. While ease of use was not significant, organizational support had a negative effect. This study contributes to the literature on AI readiness in Saudi healthcare, highlighting perceived usefulness and training as key drivers for AI adoption, while questioning assumptions about the management support role in AI adoption. Healthcare leaders and policymakers should prioritize training, communicate the practical benefits of AI, and ensure that managerial commitment is supported by resources.13 0Item Restricted Enhancing Demand Forecasting Accuracy, Inventory Performance, and Supply Chain Efficiency in Saudi Arabia’s Public Pharmaceutical Sector through Artificial Intelligence.(Saudi Digital Library, 2025) Alattas, Rawan Omar; Meriton, RoystonThis study examined the role of Artificial Intelligence (AI) in improving demand forecasting and inventory performance in pharmaceutical supply chains in Saudi Arabia’s public healthcare sector. A cross-sectional survey of 155 professionals was conducted, and data were analysed using descriptive statistics, correlation, regression, and mediation/moderation tests. Results showed that AI integration explained 71% of the variability in demand forecasting accuracy (β = 0.82, p < .001). AI adoption also predicted 69% of the variability in inventory performance (β=0.82, p < .001), with significant effects on stock turnover (β=0.83, p < .001), lead time reduction (β=0.81, p < .001), and waste minimisation (β=0.83, p < .001). Organisational capabilities mediated the link between AI adoption and supply chain performance, confirming the importance of digital infrastructure and analytics competency. Barriers such as resistance, regulatory issues, and data quality challenges were reported, but did not significantly moderate the relationship between AI integration and demand forecasting accuracy. These findings confirm that AI improves efficiency, reduces waste, and strengthens resilience in pharmaceutical supply chains. Therefore, AI adoption aligns with Saudi Arabia’s Vision 2030 healthcare reforms.15 0Item Restricted A Study on the Perceptions of Criminology and Criminal Justice Students in Scottish Universities Regarding Using Artificial Intelligence (Al) in the Criminal Justice System (CJS).(Saudi Digital Library, 2025) Jabalawi, Bayan; Brooks-Hay, OonaArtificial intelligence (AI) increasingly influences criminal justice systems (CJSs), offering efficiency, speed, and innovative decision-making opportunities. However, it also raises critical ethical, social, and technical challenges, particularly regarding bias, transparency, accountability, and the preservation of human judgment. Despite growing literature on AI and justice, limited research has examined the perspectives of future practitioners. This study addresses that gap by exploring the understandings, perceptions, and readiness of postgraduate criminology and criminal justice students in Scottish universities toward AI integration in CJSs. The research draws on semi-structured student interviews by adopting a qualitative grounded theory approach. Data were analysed using systematic coding procedures, including open, axial, and selective coding, to identify categories and construct a theoretical storyline. The findings reveal a dual perception of AI: it is seen simultaneously as a tool for institutional efficiency and a threat to human justice. Key concerns centred on algorithmic bias, privacy violations, and the erosion of empathy, while perceived benefits included enhanced efficiency, reduced institutional burdens, and novel analytical insights. Students’ knowledge was primarily shaped by fragmented sources, such as media and limited academic exposure, amplifying fascination and mistrust. Overall, students expressed conditional readiness to engage with AI, accepting it only as a supportive assistant while emphasising the need for training, safeguards, and human oversight. The study concludes that successful integration of AI in criminal justice requires technological advancement, transparent, ethical practices, and educational reforms that reinforce trust and safeguard the human dimension of justice.17 0Item Restricted Knowledge Graphs and AI in Criminal Investigations: Advancing Sensemaking and Decision Support(Saudi Digital Library, 2024) Alotaibi, Ahad Shabib; Baber, ChrisIntelligence analysis increasingly contends with large, heterogeneous, and ambiguous data—especially in criminal investigations—placing pressure on human sensemaking. This thesis examines how integrating human cognitive strategies with computational tools—Large Language Models (LLMs) and Knowledge Graphs (KGs)—may support investigative reasoning. The work is organised around three questions: (1) how KGs/AI affect human decision-making in intelligence analysis; (2) how people frame information for intelligence tasks and how technology shapes that framing; and (3) how the analysis context (e.g., crime type) influences reasoning. The Data–Frame Model (DFM) provides the organising account of framing/reframing; observed outputs are mapped to deductive, inductive, and abductive types of reasoning without claims about internal cognition. An ELICIT-derived study contrasts self-generated with provided frames in card sorting and a short reasoning task with colour-coded reliability cues. Within this task, provided frames increased cross-participant agreement at a single time point, whereas self-generated frames were more stable across sessions; participants generally preferred green (high-confidence) items, used yellow cautiously, and avoided red. A KG/Cypher modelling step approximated participants’ justifications while making assumptions (e.g., reliability thresholds) explicit. These findings, from a small novice sample, indicate benefits to preserving personal framing while keeping reliability weighting inspectable and revisable. A comparative study of human versus LLM-assisted query generation shows complementary strengths: the model rapidly proposes structured, comprehensive question sets over large text corpora, while humans more readily connect financial, behavioural, and interpersonal facets and tailor queries to local context. This motivates a mixed-initiative approach in which AI supports breadth and consistency and analysts contribute contextual nuance and judgement. To bridge cognition and computation, a provenance-aware KG (~300 nodes/~600 relationships) was designed for the North by Southwest scenario, encoding both declarative facts and procedural elements (follow-the-money / follow-the-crime). Three filter families—analysis procedures, analysts’ heuristics, and graph algorithms (e.g., centrality, community detection)—were expressed via Bloom phrases and Cypher templates. A controlled user study (novice N = 30) then examined how filter form (question-based vs algorithmic) and context (money vs crimes) were associated with what participants foregrounded, how anchors persisted, and when decisions converged. Three patterns emerged: (i) context reliably shifted focus (e.g., financial actors under money; operational roles under crimes); (ii) filter affordances shaped anchoring and paths to convergence (salience encodings in algorithmic views were frequently cited); and (iii) alignment between filter and context often coincided with higher convergence, though not uniformly. Opportunities for data→frame updating were limited by interface and workflow constraints—a design-relevant limitation. Contributions are threefold: (i) a conceptual synthesis aligning the DFM with observable reasoning types and an operational definition of candidate reframing (new evidence/rationale plus persistence); (ii) methodological assets (a transparent KG/filters pipeline with reproducible Bloom/Cypher materials, stability and convergence measures, and reliability-cue instrumentation); and (iii) empirical evidence on human–AI complementarity and on when KGs help (and constrain) investigative work. Limitations include simulated scenarios, novice samples, small cells, fixed orders, and tool-specific affordances. Future work targets professional-analyst studies, richer interaction for reframing, larger/live datasets, anti-anchoring and “why-explanations” for algorithms, and continued attention to legal, ethical, and privacy considerations. Together, these contributions arise from three empirical studies: an ELICIT-based framing study, a comparative analysis of human versus AI query generation, and a knowledge-graph experiment on filters and investigative contexts.29 0Item Restricted AI-driven Resource Anomaly Detection Framework for Robotic CI/CD Pipeline in the Edge-Cloud Continuum(Saudi Digital Library., 2025) Alhawas, Abdulaziz; Singh Gill, SukhpalContinuous integration and continuous delivery (CI/CD) can improve the reliability of robotic software. Practical pipelines still struggle with consistent simulation, runtime anomaly detection, and actionable feedback. This project integrates unsupervised anomaly detection into a ROS 2 CI/CD pipeline and evaluates it in Webots over six scenarios ranging from basic tasks to complex configurations. The system collects CPU and memory usage at 2 Hz with derived rolling and slope features, applies an Isolation Forest model, and displays results in a Streamlit dashboard. We deploy the pipeline to an edge virtual machine for continuous anomaly detection and introduce an early warning layer that predicts anomalies before they occur and sends recommendations to mitigate. The pipeline detects realistic resource anomalies in various scenarios. Results show that the proposed system performs competitively in terms of precision, recall, and F1-scores as compared to the baselines. The work also addresses practical and ethical issues by highlighting the importance of human oversight when interpreting AI decisions. Future work includes dynamic tuning of the framework parameters, online learning and drift handling, and validation on physical hardware.8 0Item Restricted Artificial Intelligence (AI) Assimilation in the Public Sector:An Attention-Based Exploration of Decision Making, Leadership, and Communication in Saudi Arabia(Saudi Digital Library, 2025) Alshahrani, Albandari Fahad; Griva, Anastasia; Dennehy, DenisThe rapid development of Artificial Intelligence (AI) has opened new possibilities for public sector organisations to improve service delivery, strengthen decision-making processes, and enhance operational efficiency. However, successfully assimilating AI in government contexts presents distinct challenges that differ markedly from those faced by private sector organisations. Public institutions operate within complex frameworks shaped by multiple stakeholder expectations, stringent regulatory requirements, accountability obligations, and often risk-averse organisational cultures — all of which significantly influence technology assimilation outcomes. Despite growing interest in AI within government, there is still limited understanding of how organisational attention dynamics shape AI assimilation processes. This PhD thesis addresses this critical gap by applying the Attention-Based View (ABV) theory to explore how leadership attention allocation, communication practices, and institutional contexts influence AI integration in public sector organisations. This doctoral thesis, structured as an article-based PhD, comprises three interrelated studies that collectively advance understanding of AI assimilation through the lens of organisational attention. The research pursues five Research Objectives (ROs): identifying organisational and governance challenges in public sector AI assimilation through a systematic literature review (RO1); investigating leadership attention allocation mechanisms in AI initiatives (RO2); examining communication channels as attention management mechanisms in public sector AI integration (RO3); analysing how national policies and institutional contexts influence AI assimilation outcomes (RO4); and providing practical insights for AI-driven governance (RO5). Methodologically, the research combines a systematic literature review with qualitative case studies conducted in the Saudi Arabian public sector, focusing on organisations implementing AI under the Vision 2030 transformation agenda. The first study presents a systematic literature review of 61 peer-reviewed articles published between 2012 and 2023, mapping the current state of AI research in public administration. This review identifies seven major challenges including infrastructure limitations, data governance issues, workforce readiness gaps, regulatory complexities, cultural resistance, cybersecurity concerns, and resource constraints, and five primary benefits, such as enhanced decision-making, greater efficiency, cost optimisation, increased transparency, and improved citizen engagement. This study lays a foundational understanding of AI assimilation challenges and underscores the need for attention-based perspectives. The second study applies ABV theory to examine attention-related challenges in AI assimilation within Saudi public sector organisations. Using in-depth qualitative analysis of a single case study, the research identifies five core attention-based challenges, divided into internal (situated) and external (structural) categories. Internally, challenges include fragmented leadership attention, competing priorities, and resource conflicts; externally, they involve regulatory demands, stakeholder expectations, and institutional pressures. These findings highlight the importance of understanding how attention allocation shapes AI outcomes and underscore the central role of leadership focus in managing assimilation challenges. The third study extends this analysis by exploring how leadership practices and communication channels facilitate AI integration across multiple Saudi public sector organisations. The research shows that leaders coordinate organisational attention through structural frameworks (formal systems), situated practices (contextual engagement), and communication-mediated mechanisms (information flow management). The study introduces the concept of leaders as "attention architects" who design and manage attention structures to support digital transformation. Findings reveal how formal and informal communication channels function not only as conduits but as active mechanisms shaping attention, fostering alignment, and sustaining commitment to AI initiatives. Theoretically, this thesis advances ABV by applying the theory to public sector AI assimilation and developing communication channels as attention regulators. It offers the first thorough application of ABV in public sector AI assimilation, highlighting distinct dynamics compared to private sector contexts. The study also underscores the role of national transformation agendas in shaping attention allocation and assimilation trajectories, providing insights relevant to Global South and developing country contexts. Furthermore, this thesis establishes a novel theoretical framework that integrates organisational attention with institutional theory, demonstrating how cultural and political factors systematically influence attention allocation patterns in complex technological transformations (Ocasio et al., 2018; Taras et al., 2020). The thesis also contributes to communication theory by conceptualising formal and informal communication networks as co-equal drivers of attention distribution, challenging traditional hierarchical models of organisational attention and proposing a more dynamic, multi-channel approach to understanding attention flows in public sector contexts (Putnam & Mumby, 2014; Cornelissen et al., 2020). Practically, the findings provide actionable guidance for public sector leaders and policymakers. They suggest strategies for designing attention structures, managing competing demands, and leveraging communication channels to enable successful AI integration. The focus on Saudi Arabia's Vision 2030 offers valuable lessons for other governments pursuing digital transformation under complex institutional and cultural constraints. This thesis acknowledges limitations, including its focus on a single national context, the literature review's temporal scope (up to 2023), and the qualitative nature of empirical studies. These limitations present opportunities for future work, such as cross-country comparative studies, longitudinal analyses of attention dynamics, and quantitative validation of the developed frameworks. In sum, this thesis makes significant contributions to both theory and practice by demonstrating the critical role of organisational attention in public sector AI assimilation. It reveals that successful integration demands strategic attention management, effective communication systems, and leadership practices that align organisational focus with implementation goals. The findings offer a strong foundation for future studies on attention dynamics in technology assimilation and provide practical insights to support leaders and policymakers striving for AI-enabled governance transformation. By integrating theoretical depth with practical relevance, this PhD research advances academic understanding and offers concrete guidance for navigating public sector digital transformation.7 0Item Restricted Automatic Essay Scoring in Arabic: Development, Evaluation, and Advanced Techniques(University of Bristol, 2025) Ghazawi, Rayed; Simpson, EdwinAutomated Essay Scoring (AES) has advanced considerably due to recent progress in natural language processing (NLP). This thesis examines key challenges in AES, with a particular focus on the Arabic language, and proposes practical approaches informed by both computational techniques and educational theory. First, the research investigates how the formulation of essay questions affects the accuracy of automated scoring systems. A set of question-design criteria, derived from educational principles, is introduced and empirically tested. Experiments show that adherence to these criteria can significantly improve AES performance, with improvements of up to 40% observed using BERT-based models for English essays. Given the limited resources for Arabic AES, this thesis introduces the AR-AES dataset, consisting of 2046 essays from undergraduate students across multiple courses, annotated independently by two university instructors. This resource alleviates the scarcity of Arabic-language datasets for AES, supporting model development and evaluation. Experimental analyses using pretrained Arabic NLP models demonstrate that transformer-based approaches achieve the highest levels of agreement with human scores. In many cases, their predictions show greater consistency with the gold scores than the agreement observed between the human annotators themselves. This high level of agreement with human scores indicates that, under appropriate conditions, the proposed AES system may be suitable for assisting human markers in real-world educational settings. Additionally, the thesis explores the potential of large language models (LLMs), including ChatGPT, Llama, Aya, Jais, and ACEGPT for Arabic AES. Experiments with different training approaches, zero-shot, few-shot, and fine-tuning, demonstrate the importance of prompt engineering. A mixed-language prompting strategy, combining Arabic essays with English scoring guidelines, was found to notably enhance model performance. Nonetheless, fine-tuned AraBERT consistently yielded the strongest results, indicating that LLMs may not yet be the most effective option for Arabic AES tasks when training data is limited. Finally, an active learning framework is introduced, integrating AraBERT with uncertainty- and diversity-based sampling strategies. This human-in-the-loop approach prioritises essays that most benefit from expert review, reducing the need for extensive manual annotation while preserving high-scoring accuracy. Rather than replacing human markers, the system complements their efforts, offering a more efficient and consistent approach to large-scale essay evaluation. Overall, this thesis advances AES by introducing explicit criteria for effective essay question design, while also addressing specific challenges in Arabic AES. It contributes a comprehensively annotated dataset, presents a systematic evaluation of state-of-the-art NLP models, and effectively integrates active learning to balance automated scoring accuracy and human involvement.13 0Item Restricted Development of BMX siRNA Lipid Nanoparticles using Novel Ionizable, and Cleavable Lipids Discovered through AI and Experimentation for Cancer Therapy(Saudi Digital Library, 2025) Alasmari, Abdulaziz Nasser; Abu Fayyad, AhmedProstate cancer is the most commonly diagnosed male malignancy across the world. It is also considered the fifth leading cause of cancer death in men. in 2020 there was 1,414,249 newly diagnosed cases and 375,000 deaths worldwide from prostate cancer (Leslie SW.2023). The Tec family nonreceptor tyrosine kinase BMX is activated downstream of PI3K and has been implicated in regulation of multiple pathways and in the development of cancers including prostate cancer (Chen S. 2018). The available science supports the role of BMX in advanced prostate cancer, head & neck cancer, aggressive brain cancer, and many other immunological diseases (Cenni, Gutmann et al. 2012). RNA interference plays an important role in regulating the gene expression level within the cells (Agrawal N. 2003). However, delivering small interfering RNA (siRNA is challenging due to many obstacles, such as extremely short half-life in vivo, rapid elimination via glomerular filtration, and inability to cross cell membranes (due to the hydrophilic nature and negative charge of siRNA molecules). In order to use siRNA as a treatment for prostate cancer an effective delivery system is needed. Here, we demonstrate that BMX expression in prostate cancer is suppressed directly by siRNA using the delivery system. The delivery system used is composed of the negatively-charged siRNA encapsulated into a multi-component structure that contains (DOPE), 1,2-Dioleoyl-3 trimethylammonium propane (DOTAP), Cholesterol and Phosphatidylcholine. To further enhance the activity of the BMX siRNA lipid nanoparticle compositions two novel lipids; a cleavable PEGylated lipid, and an ionizable cationic lipid were synthesized and characterized by our team (Abu-Fayyad and Nazzal 2017) and then added to the formulation. The goal of incorporating the novel lipid is to overcome shielding effect PEGylation imparts to the formulation by the presence of PEG2000 in the composition since it represents an obstacle for the formula’s cellular uptake, and the subsequent engulfment by the endosome to release its contents (Kulkarni, Witzigmann et al. 2019).7 0Item Restricted Deep Learning based Cancer Classification and Segmentation in Medical Images(Saudi Digital Library, 2025) Alharbi, Afaf; Zhang, QianniCancer has significantly threatened human life and health for many years. In the clinic, medical images analysis is the golden stand for evaluating the prediction of patient prog- nosis and treatment outcome. Generally, manually labelling tumour regions in hundreds of medical images is time- consuming and expensive for pathologists, radiologists and CT scans experts. Recently, the advancements in hardware and computer vision have allowed deep-learning-based methods to become main stream to segment tumours automatically, significantly reducing the workload of healthcare professionals. However, there still remain many challenging tasks towards medical images such as auto- mated cancer categorisation, tumour area segmentation, and relying on large-scale labeled images. Therefore, this research studies theses challenges tasks in medical images proposing novel deep-learning paradigms that can support healthcare professionals in cancer diagnosis and treatment plans. Chapter 3 proposes automated tissue classification framework called Multiple Instance Learning (MIL) in whole slide histology images. To overcome the limitations of weak super- vision in tissue classification, we incorporate the attention mechanism into the MIL frame- work. This integration allows us to effectively address the challenges associated with the inadequate labeling of training data and improve the accuracy and reliability of the tissue classification process. Chapter 4 proposes a novel approach for histopathology image classification with MIL model that combines an adaptive attention mechanism into an end-to-end deep CNN as well as transfer learning pre-trained models (Trans-AMIL). Well-known Transfer Learning architectures of VGGNet [14], DenseNet [15] and ResNet[16] are leverage in our framework implementation. Experiment and deep analysis have been conducted on public histopathol- ogy breast cancer dataset. The results show that our Trans-AMIL proposed approach with VGG pre- trained model demonstrates excellent improvement over the state-of-the-art. Chapter 5 proposes a self-supervised learning for Magnetic resonance imaging (MRI) tu- mour segmentation. A self-supervised cancer segmentation framework is proposed to re- duce label dependency. An innovative Barlow-Twins technique scheme combined with swin transformer is developed to perform this self supervised method in MRI brain medical im- ages. Additionally, data augmentation are applied to improve the discriminability of tumour features. Experimental results show that the proposed method achieves better tumour seg- mentation performance than other popular self- supervised methods. Chapter 6 proposes an innovative Barlow Twins self supervised technique combined with Regularised variational auto-encoder for MRI tumour images as well as CT scans images segmentation task. A self-supervised cancer segmentation framework is proposed to reduce label dependency. An innovative Barlow-Twins technique scheme is developed to represent tumour features based on unlabeled images. Additionally, data augmentation are applied to improve the discriminability of tumour features. Experimental results show that the pro- posed method achieves better tumour segmentation performance than other existing state of the art methods. The thesis presents four approaches for classifying and segmenting cancer images from his- tology images, MRI images and CT scans images: unsupervised, and weakly supervised methods. This research effectively classifies histopathology images tumour regions based on histopathological annotations and well-designed modules. The research additionally comprehensively segments MRI and CT images. Our studies comprehensively demonstrate label-effective automatic on various types of medical image classification and segmentation. Experimental results prove that our works achieve state-of-the-art performances on both classification and segmentation tasks on real world datasets18 0