The management of acutely unwell patients is challenging for many clinicians, as clinical decisions must be made quickly and must account for a large number of interdependent factors. A comprehensive clinical decision support system (CDSS) would be highly desirable in such a setting. The number of interdependent and unpredictable factors involved in such scenarios results in complex adaptive systems. Conventional, algorithm-driven systems have limited utility under such circumstances. Machine learning-based systems represent a potential solution to this unmet need, but the difficulty in obtaining large datasets in the domain of acute care is a limiting factor. (More generally, there are many areas of medicine where machine learning-based systems could significantly improve patient care but where large datasets to train such systems are not readily available.) We propose that there may be a way to circumvent this issue, drawing upon the theory behind systems such as DeepMind’s AlphaGo.


We propose to develop the existing prototype of the Virtu-ALS resuscitation simulator to incorporate a more comprehensive range of clinical scenarios and the ability to log and upload user actions within the software. This will be done in close liaison with the clinical staff and resuscitation training programme at Craigavon Area Hospital, where we will seek expert consultation on the clinical design of the training system and undertake early product testing and development with experienced healthcare professionals.

Clinical accuracy of the system will be key to the success of the project, and in an area where high-level evidence is sparse this will rely heavily on individual experience and expertise. When development is complete, the app will be updated on the publicly app stores. Over the last six months the app has had approximately 15 000 downloads. Based upon this, we estimate that we will be able to capture data from over 100 000 clinical decisions, along with information about scenario outcomes following these decisions. This data will come from users with widely varying and un-validated levels of experience and competence, but will be used as pre-training data for the machine-learning-based system and therefore does not need to be of especially high quality.

Following the pre-training with data accumulated from human users, we will further train the system through exposure to the resuscitation simulator. Its goal will be to optimise patient haemodynamics at any given point and minimise the time taken to reach predefined criteria for patient “stabilisation”. Assuming the simulator adequately reflects real-world clinical practice, the resultant “intelligent” system should have developed skills that are transferable to real-world situations. We will aim to demonstrate this by exposing the system to the scenarios used for the assessment of widely recognised resuscitation courses such as Advanced Life Support and Advanced Cardiac Life Support and objectively evaluating its performance. NB – the criticism could be levelled that an algorithm-based system could pass an ALS course assessment, given the highly-controlled nature of such a scenario. We will assess the system in other ways to demonstrate that its capabilities go beyond this.

Essential criteria

  • To hold, or expect to achieve by 15 August, an Upper Second Class Honours (2:1) Degree or equivalent from a UK institution (or overseas award deemed to be equivalent via UK NARIC) in a related or cognate field.


This project is funded by: the EU INTERREG VA Programme

This project is supported by the European Union's INTERREG VA Programme, managed by the Special EU Programmes Body (SEUPB).

The scholarships will cover tuition fees and a maintenance award of not less than £14,553 per annum for three years (subject to satisfactory academic performance). Applications are invited from UK, European Union and overseas students.

Other information

The Doctoral College at Ulster University