Using Brain Imaging And Machine Learning To Improve Prognostic Models In Intracerebral Haemorrhage
- The Natalie Kate Moss Trust
- Jun 3
- 4 min read
In June 2025, The Natalie Kate Moss Trust began their support of Olivia Murray, a brilliant researcher at The University of Manchester, with her three year Postdoctoral Research Fellowship where she will use machine learning to improve prognostic models in intracerebral haemorrhage. This research could help to transform the treatment received by patients, improving survival rates and lessening disability post Intracerebral Haemorrhage.
Why is accurate prognostication important?
Prognostication is a prediction of the recovery trajectory of a patient. By assessing features such as a patient’s brain scan, age, ICH volume, and level of consciousness on admission, clinicians can estimate a patient’s potential for recovery. This initial prediction is important, as it informs the treatment a patient goes on to receive.
However, there is evidence that current methods of prognostication for ICH are overly pessimistic and fail to identify some patients who have a chance of good motor recovery.
Olivia’s research shows that healthcare professionals are more likely to withdraw life-sustaining treatment for ICH patients than for those with other types of strokes potentially creating a ‘self-sustaining prophecy’. Providing accurate and personal prognoses to healthcare professionals and families could thus improve patient outcomes.
White matter injury assessment
The integrity of important white matter tracts in the brain is crucial for recovery after stroke, and stroke survivors who have sustained damage to their white matter tracts often have poorer motor recovery. Damage to these white matter tracts is best assessed using diffusion MRI brain scans, a type of imaging where they can measure the diffusion of water along the white matter tract. However, advanced MRI imaging is not usually available to ICH patients.
Olivia’s PhD project involved using machine learning models trained on diffusion MRI to assess white matter damage from routine diagnostic CT scans alone, making rapid white matter injury assessment available to all ICH patients.
They discovered that our markers of white matter injury, assessed from CT imaging, significantly predicted motor outcome in both the short term and the long term after ICH. Moreover, they found these markers were more predictive than hematoma volume, which is a factor that traditional ICH prognostic models rely on. This showed that utilising information available from diagnostic imaging could meaningfully inform clinical practices.
What next?
Moving forward, with the support of NKMT, Olivia will be developing this prognostic modelling further, by incorporating additional information available from CT imaging. She plans to develop AI tools to quickly and accurately assess background brain health, looking at factors such as brain atrophy and pre-existing white matter disease. Improving on the models developed during her PhD, she will analyse the haematoma and the surrounding brain tissue on CT to predict the likelihood of haematoma expansion, a dangerous complication after ICH. The prognostic power of these models will be tested and validated using data from previous ICH clinical trials. If their models show significant improvement over the standard prognostic tools used in the clinic, they could be trialled in a patient population alongside the ABC ICH study.
If successful, future work would involve working with the NKMT to:
Test the prognosis tool in an existing database of all ICH patients in Scotland over a 7-year period. This will show how well the tool works in a wide range of real patients in NHS hospitals.
Decide with clinicians and those affected by ICH when the tool should suggest actions, like referrals for surgery or intensive care.
Develop software to run the tool in the NHS in collaboration with healthcare professionals, people with ICH experience, and companies with similar software. This will ensure the tool fits existing systems and meets the needs of everyone involved.
Introduce the tool and recommended actions in NHS hospitals through a clinical trial to test if it improves care, reduces deaths and disability for ICH patients, and is cost-effective.
Why NKMT’s support matters
Olivia Murray: “The support given by the Natalie Kate Moss Trust will allow us to transition our initial ideas from my PhD into validated tools with the potential to improve ICH patient care. Their funding will allow us to test and calibrate our models in both clinical trials and real-world ICH patient datasets, ensuring they are robust and applicable in real-world clinical settings. This investment in translating ideas from the bench into the clinic is crucial for improving ICH treatment and will ultimately lead to enhanced quality of life for ICH survivors and their families.”
A bit about Olivia
Olivia has a background in physics and completed her Master's degree in physics in Manchester in 2021. It was during her degree that she developed an interest in applied machine learning. During the final year of her degree, she was introduced to medical imaging and conducted a research project using AI to analyse CT images of head and neck cancer patients. Over the course of this project she developed a goal for her career; to use her knowledge of physics and machine learning to improve the lives of patients.
In 2021, she began her 3.5-year PhD project, aimed at improving prognostic modelling by investigating the role of white matter injury in recovery after intracerebral haemorrhage (ICH), under the supervision of Professor Adrian Parry-Jones.
