Higher education institutions require course leaders to continuously develop and refine academic modules to align with evolving educational standards and university guidelines. This paper presents a domain-specific AI-powered assistant developed for Teesside University that leverages a fine-tuned LLaMA-3.2-3B-Instruct model integrated with Retrieval-Augmented Generation and an agentic decision-making layer. The system generates policy-compliant course content using Low-Rank Adaptation and 4-bit quantisation for efficient training on limited hardware. A hybrid retrieval system combining FAISS for dense semantic search and BM25 for lexical matching grounds responses in institutional documents. Evaluation demonstrates a perplexity of 7.38, Hits@3 of$\mathbf{1. 0 0}$, cosine similarity of$\mathbf{0. 7 5 6 5}$, and hallucination rate of 0.00. Human evaluators rated outputs with an average score of 8.25 out of 10 across relevance, fluency, completeness, and organisation metrics, confirming the system's practical utility in academic planning while ensuring alignment with educational policy.

Modern football increasingly relies on data-driven decision-making to inform tactics, selection, and in-game adjustments. This work develops a club-specific framework that couples supervised machine learning (ML) with Business Intelligence (BI) dashboards to predict match outcomes and surface opponent tendencies that coaches can act on. Using$\mathbf{1, 5 4 4}$match-level rows and 1,355 engineered features covering technical, tactical, and event metrics for Change for a Championship Football Team, we trained and compared Logistic Regression, Random Forest, Gradient Boosting, XGBoost, Support Vector Machine (SVM), k-Nearest Neighbors (KNN), Naïve Bayes, and an Artificial Neural Network (ANN). Models were evaluated with accuracy, precision/recall/F1, confusion matrices, ROC-AUC (one-vs-rest), and 5-fold cross-validation; SHAP values supported post-hoc explainability. A Power BI dashboard operationalizes predictions via modules for overview KPIs, match statistics, opponent strategy, possession/passing dynamics, and forecasting.

This study applies machine learning techniques to predict football match outcomes in the EFL (English Football League) Championship, emphasising the identification of key performance metrics that contribute to team success. Using over 4,000 team-level match records, three models, Logistic Regression (LR), Decision Tree, and Neural Network, were evaluated. LR achieved the highest accuracy (72.34), outperforming the more complex models in predicting Wins, with an F1 of 0.82 and Losses (0.79), though all models struggled with Draws. Feature importance analysis revealed that offensive metrics, such as goals from set pieces, possession phases, and attacking transitions, were the most influential predictors of success. Defensively, factors like pressure on shots and vulnerability to bypassed defenders negatively impacted outcomes. While LR offered interpretable global feature insights, it lacks the ability to explain individual team predictions, a limitation that future work could address using tree-based or attention-driven models capable of producing local, instance-level explanations. These findings reinforce the value of data-driven approaches in football analytics, offering both predictive accuracy and tactical insight.

Physiotherapy education often suffers from inconsistencies and delays in delivering meaningful feedback during practical assessments. This paper presents an AI-driven, keyword-based feedback generation system designed to address these limitations by producing structured, personalised, and timely feedback aligned with clinical evaluation standards. The approach builds on previous keyword-tagged natural language feedback systems and introduces a more advanced methodology through the use of Meta's Llama 4 Scout, a large-scale language model fine-tuned using Low-Rank Adaptation (LoRA). A carefully preprocessed dataset of physiotherapy assessments was utilised, and the system was trained using a structured prompt framework that integrates assessment keywords and improvement cues to synthesise relevant textual responses. This methodology enables the generation of consistent feedback output across diverse clinical contexts while reducing the manual workload of educators. Qualitative evaluation by physiotherapy educators confirmed that the model maintains appropriate clinical terminology, structural consistency, and pedagogical tone, outperforming general-purpose models in domain-specific relevance. The model design emphasises scalability, efficiency, and domain alignment in physiotherapy training environments. This study contributes to the advancement of AI-assisted education tools and highlights the role of large language models in clinical training domains. While evaluation is ongoing, this work lays a robust foundation for future research in automated feedback for skill-based learning disciplines such as physiotherapy.

Lower back pain (LBP) is a global medical condition that affects more than $\mathbf{8 0 \%}$ of people at least once in their lifetime. Various abnormal conditions could cause LBP. One of these conditions is Lumbar Spinal Stenosis (LSS), which is usually a serious condition, hence the need for its prompt diagnosis. The visual assessment process of magnetic resonance imaging (MRI) is expensive in terms of time and effort and prone to delays. The integration of an AI-enabled system to help clinicians in diagnosing patients with LBP is expected to alleviate the burden on radiologists and foster a less time-consuming and cost-efficient diagnostic process. In this paper, a convolutional neural network (CNN) model has been built to diagnose T2-weighted (T2W) axial MRI scans. The model detects LSS in these scans and has achieved remarkable accuracy and recall of $\mathbf{9 1 \%}$ and $\mathbf{9 6 \%}$, respectively. Explainable AI (XAI) using LIME’s ImageExplainer was also implemented to ensure the model offers explainable insights. This is to ensure the model is not only accurate but also reliable.

The rapid evolution of artificial intelligence (AI) has ushered in new opportunities for enhancing physiotherapy education. This study introduces an innovative AI-based tool that integrates computer vision and speech recognition to provide a comprehensive and objective assessment of physiotherapy students’ practical skills. Designed to surpass the limitations of traditional, subjective methods, this system analyses both physical movements and verbal interactions during exams, categorizing student performance into ‘Good’, ‘Average’, or ‘Brief’ to offer a multifaceted evaluation of their clinical abilities. The research aimed to develop and validate a system that accurately captures and analyses nuanced physical and verbal behaviours in physiotherapy practical exams. The system was trained and tested on a curated dataset of recorded sessions, where it demonstrated high precision, achieving an overall accuracy of 89% in identifying specific actions performed by students. These results underscore the system’s capability to enhance the objectivity and depth of practical assessments significantly. This study marks a significant contribution to healthcare education by showing the effectiveness of AI technologies in assessing practical and communication skills. The AI-assisted tool not only improves the educational process but also better prepares students for clinical practice. The findings suggest a model for future AI integration in professional healthcare training and assessment, highlighting the transformative potential of AI in advancing physiotherapy education.

Eye tracking and head pose estimation (HPE) have previously lacked reliability, interpretability, and comprehensibility. For instance, many works rely on traditional computer vision methods, which may not perform well in dynamic and realistic environments. Recently, a widespread trend has emerged, leveraging deep learning for HPE specifically framed as a regression task; however, considering the real-time applications, the problem could be better formulated as classification (e.g., left, centre, right head pose and gaze) using a hybrid approach. For the first time, we present a complete facial profiling approach to extract micro and macro facial movement, gaze, and eye state features, which can be used for various applications related to comprehension analysis. The multi-model approach provides discrete human-understandable head pose estimations utilising deep transfer learning, a newly introduced method of head roll calculation, gaze estimation via iris detection, and eye state estimation (i.e. , open or closed). Unlike existing works, this approach can automatically analyse the input image or video frame to produce human-understandable binary codes (e.g., eye open or close, looking left or right, etc.) for each facial component ( aka face channels). The proposed approach is validated on multiple standard datasets, indicating outperformance compared to existing methods in several aspects, including reliability, generalisation, completeness, and interpretability. This work will significantly impact several diverse domains, including psychological and cognitive tasks with a broad scope of applications, such as in police interrogations and investigations, animal behaviour, and smart applications, including driver behaviour analysis, student attention measurement, and automated camera flashes.

Where traditional motorways contain hard shoulders to provide refuge for broken-down vehicles, smart motorways instead use live outer lane to ease congestion. The live lanes can be closed due to accidents or breakdowns which is communicated to other road users through overhead gantry signs. This can only occur if the traffic management control center is made aware of the stationary vehicle(s), through either notification via phone call or Motorway Incident Detection and Automatic Signaling (MIDAS) induction loop technology. Alternatively, radar-based stopped vehicle detection is used to identify non-moving objects on the highway. However, this technology is unable to recognize objects or distinguish between congestion and break down etc., which leads to generate false alarms. For the first time, we propose a fully autonomous computer vision and deep learning-based solution to detect stationary vehicles on highways and local roads. We employ deep transfer learning to build a custom-trained vehicle detection model using a newly prepared dataset comprising over 105,000 annotated vehicle instances. DeepSort algorithm is employed for real-time vehicle tracking through associating instances between time series frames, followed by a rule-based algorithm to identify the current state of detected vehicles. Experimental outcomes show our approach as outperforming the state-of-the-art methods in terms of efficient and reliable detection of stationary vehicles (with 98.3% accuracy) as well as distinguish them from congestions when evaluated over video streams captured in realistic dynamic and diverse conditions.

This paper addresses the central problem of automatic segmentation of lumbar spine Magnetic Resonance Imaging (MRI) images to delineate boundaries between the anterior arch and posterior arch of the lumbar spine. This is necessary to efficiently detect the occurrence of lumbar spinal stenosis as a leading cause of Chronic Lower Back Pain. A patch-based classification neural network consisting of convolutional and fully connected layers is used to classify and label pixels in MRI images. The classifier is trained using overlapping patches of size 25×25 pixels taken from a set of cropped axial-view T2-weighted MRI images of the bottom three intervertebral discs. A set of experiment is conducted to measure the performance of the classification network in segmenting the images when either all or each of the discs separately is used. Using pixel accuracy, mean accuracy, mean Intersection over Union (IoU), and frequency weighted IoU as the performance metrics we have shown that our approach produces better segmentation results than eleven other pixel classifiers. Furthermore, our experiment result also indicates that our approach produces more accurate delineation of all important boundaries and making it best suited for the subsequent stage of lumbar spinal stenosis detection.

In this paper, a novel application of machine learning algorithms including Neural Network architecture is presented for the prediction of flood severity. Floods are considered natural disasters that cause wide-scale devastation to areas affected. The phenomenon of flooding is commonly caused by runoff from rivers and precipitation, specifically during periods of extremely high rainfall. Due to the concerns surrounding global warming and extreme ecological effects, flooding is considered a serious problem that has a negative impact on infrastructure and humankind. This paper attempts to address the issue of flood mitigation through the presentation of a new flood dataset, comprising 2000 annotated flood events, where the severity of the outcome is categorised according to 3 target classes, demonstrating the respective severities of floods. The paper also presents various types of machine learning algorithms for predicting flood severity and classifying outcomes into three classes, normal, abnormal, and high-risk floods. Extensive research indicates that artificial intelligence algorithms could produce enhancement when utilised for the pre-processing of flood data. These approaches helped in acquiring better accuracy in the classification techniques. Neural network architectures generally produce good outcomes in many applications, however, our experiments results illustrated that random forest classifier yields the optimal results in comparison with the benchmarked models.

Disc herniation is considered as the main cause for lower back pain (LBP), a health issue that affects a very high proportion of the UK population and is costing the UK government over £1.3 million per day in health care cost. Currently, the process to diagnose the cause of LBP involves a visual examination of a large number of Magnetic Resonance Images (MRI) but this process is both expensive in terms time and effort. Automatic detection of the lumbar disc herniation will reduce the time to diagnose and detect the cause of LBP by the orthopedist. There has been very limited progress towards automatic detection of disc herniation and all of the proposed techniques still require substantial manual intervention in many of the stages. Our analysis of the problem suggests that using the axial view of the MRI could potentially improve the outcome as opposed to the sagittal view used by these techniques. In this paper, we propose using the Centroid Distance Function as a shape feature of a segmented disc MRI taken from the axial view. Visual observation of the feature indicates that the feature could be used as a suitable indicator of the presence of herniation in the lumbar disc