Dr William Marsh

Teaching

Embedded Systems (Postgraduate/Undergraduate)

This module provides a practice-oriented introduction to embedded real-time systems. The main topics are (1) Modelling and simulation in UML and state-of-the-art tools; (2) Basic concepts of micro-controllers; (3) Real-time systems with interrupts and schedulers; (4) Real-time operating systems: processes and communication; (5) Energy aware design and construction; (6) Debugging and testing as part of software development processes.

Statistics for Artificial Intelligence and Data Science (Postgraduate)

This module has two components. The first introduces students to the use of probability and statistics in the context of data analysis. The module starts with basics of descriptive statistics and probability distributions. Then we go on with applied statistics techniques, such as visualisation, fitting probability distributions, time-series analysis, and hypothesis testing, which are all fundamental to the exploration, insight extraction, and modelling activities that are fundamental in handling data, of any size. The second covers some basic matrix algebra, including matrix multiplication and diagonalisation.

Research

Research Interests:

Medical Decision Support Models: Data, Knowledge and Evidence

Can data be used for decision-making? In many applications there are not enough data, key values are not directly observed or the problem requires reasoning about change. In these cases, it is better to combine data and knowledge for building a decision model.

Many medical decision problems fit this pattern. However, given the long history of clinical trials, clinicians are reluctant to assume an understanding of causes even when trials are completely impractical. Recent work on decision making in trauma surgery has shown the potential of causal models implemented using Bayesian networks. However, there are still many challenges before the use of these models can become routine.

Safety, Reliability And Risk: Modelling Accidents & Incident Causes

Analysing what can go wrong is fundamental for assessing risk in safety systems. Existing approaches have a number of deficiencies: (1) human behaviour and technical failures are poorly integrated (2) model created for system approval are often not used when a system is in operation (3) information on incidents and procedural compliance is not used to update risks.

The aim of the research is to extend existing accident-based modelling techniques to resolve these problems. Recent work has proposed a new model structure using a Bayesian network for causal modelling from accident / incident data, with the aim of predicting the likely safety / reliability improvement that would be achieved by changes in the operation of a system at a specific location.

Publications

• Hill A, Morrissey D, Marsh W (2025). “It’s the future, come on!”: a think aloud study exploring clinicians’ use of knowledge-based AI decision support. nameOfConference

  
  

  DOI: 10.1016/j.ijmedinf.2025.106089

  
  

  QMRO: qmroHref
• Marsden MER, Perkins ZB, Pisirir E et al. (2025). Early clinical evaluation of a machine-learning system for risk prediction of trauma-induced coagulopathy in the prehospital setting. nameOfConference

  
  

  DOI: 10.1136/emermed-2024-214396

  
  

  QMRO: qmroHref
• Kyrimi E, McLachlan S, Wohlgemut JM et al. (2025). Explainable AI: definition and attributes of a good explanation for health AI. nameOfConference

  
  

  DOI: 10.1007/s43681-025-00668-x

  
  

  QMRO: https://qmro.qmul.ac.uk/xmlui/handle/123456789/105550
• Kyrimi E, Mossadegh S, Wohlgemut JM et al. (2025). Counterfactual reasoning using causal Bayesian networks as a healthcare governance tool. nameOfConference

  
  

  DOI: 10.1016/j.ijmedinf.2024.105681

  
  

  QMRO: https://qmro.qmul.ac.uk/xmlui/handle/123456789/101474
• Şakar CT, Keith‐Jopp C, Yet B et al. (2024). A Multi‐Criteria Decision Support Tool for Shared Decision Making in Clinical Consultation. nameOfConference

  
  

  DOI: 10.1002/mcda.70001

  
  

  QMRO: qmroHref
• Alptekin C, Wohlgemut JM, Perkins ZB et al. (2025). Presenting predictions and performance of probabilistic models for clinical decision support in trauma care. nameOfConference

  
  

  DOI: 10.1016/j.ijmedinf.2024.105702

  
  

  QMRO: qmroHref
• Wohlgemut JM, Pisirir E, Stoner RS et al. (publicationYear). A scoping review, novel taxonomy and catalogue of implementation frameworks for clinical decision support systems. nameOfConference

  
  

  DOI: 10.1186/s12911-024-02739-1

  
  

  QMRO: https://qmro.qmul.ac.uk/xmlui/handle/123456789/106311
• Kyrimi E, McLachlan S, Wohlgemut JM et al. (2024). Explainable AI: Definition and attributes of a good explanation for health AI. nameOfConference

  
  

  DOI: 10.48550/arxiv.2409.15338

  
  

  QMRO: qmroHref
• Hill A, Morrissey D, Marsh W (2024). What characteristics of clinical decision support system implementations lead to adoption for regular use? A scoping review. nameOfConference

  
  

  DOI: 10.1136/bmjhci-2024-101046

  
  

  QMRO: https://qmro.qmul.ac.uk/xmlui/handle/123456789/100793
• Lowe C, Sephton R, Marsh W et al. (publicationYear). Evaluation of a Musculoskeletal Digital Assessment Routing Tool (DART): Crossover Noninferiority Randomized Pilot Trial. nameOfConference

  
  

  DOI: 10.2196/56715

  
  

  QMRO: https://qmro.qmul.ac.uk/xmlui/handle/123456789/100495