Unified high-dimensional modelling in stroke patients v1.0
Research type
Research Study
Full title
Unified high-dimensional modelling of chronic and acute injury in stroke.
IRAS ID
190353
Contact name
Yee Mah
Contact email
Duration of Study in the UK
1 years, 6 months, 21 days
Research summary
Research Summary
In the UK, over 150,000 people suffer a stroke each year, making it the country’s third biggest killer. A stroke is when the blood supply to part of your brain is cut off. Without blood, brain cells can be damaged or die, leading to a variety of symptoms depending on where the damage has happened.
When a patient presents to a hospital with a stroke, they will receive a brain scan—a three dimensional picture of the brain—to locate the region of damage. Around 25-30% of stroke patients will have had more than one stroke and it is not uncommon to find additional older areas of damage in the brain other than the one that has brought them to hospital.
This project aims to use hundreds to thousands of brain scans to create complex models that look for links between the pattern of brain injury and the symptoms they produce. To attain these large datasets, we will use brain scans that are already acquired in clinical stroke practice and the medical assessments routinely performed whilst in hospital. In this way we will make better use of the information we already have and avoid any additional interventions for patients. Through the development of automated image analysis techniques, the regions of brain damage can be extracted. Unlike previous models that have only examined the volume of damage caused by a recent stroke, we will use information arising from both recent and past injuries.
It is hoped these mathematical models can better predict the symptoms a patient will experience based on their pattern of brain injury, thus assisting patients, families and health services in planning for the future.Summary of Results
Acute stroke is often superimposed on chronic damage from previous cerebrovascular events. This background will inevitably modulate the impact of acute injury on clinical outcomes to an extent that will depend on the precise anatomical pattern of damage. Previous attempts to model the relationship between stroke injury and functional outcomes have used simple models that ignore a lot of the anatomical details, and individuals’ past medical history.
In this project, we have designed and validated a new method to automatically identify the pattern of chronic stroke injury from non-contrast CT scans (lesion segmentation). By applying this new lesion segmentation routine to a large clinical dataset of CT brain scans, we were able to train a prediction model, to predict with moderate levels of accuracy, the pre-admission and discharge level of functional independence of the patients defined by the modified Rankin Scale score. Importantly, we show that with sufficiently large sample sizes, increasing the complexity of the model by incorporating more variables, improves the predictive performance of the models.REC name
London - Camden & Kings Cross Research Ethics Committee
REC reference
15/LO/1955
Date of REC Opinion
8 Dec 2015
REC opinion
Favourable Opinion