A0929
Title: The importance of morphology data in predicting the risks of aneurysm rupture
Authors: Ehsan Kharatikoopaei - Durham University (United Kingdom) [presenting]
Nasima Akhter - Durham University (United Kingdom)
Amanda Ellison - Wolfson Research Institute for Health and Wellbeing Department of Psychology Durham University (United Kingdom)
Nicki Richards - Teesside Aneurysm Group (United Kingdom)
Boguslaw Obara - Newcastle University (United Kingdom)
Adetayo Kasim - Durham University (United Kingdom)
Steve Steve Bonner - Department of Neurocritical care The James Cook University Hospital Middlesbrough (United Kingdom)
Edel McCauley - Teesside Aneurysm Group (United Kingdom)
Nitin Mukerji - South Tees Hospitals NHS Trust-The James Cook University Hospital-Middlesbrough (United Kingdom)
Abstract: Spontaneous Subarachnoid Haemmorrhage (SAH), mainly caused by rupture of Intracranial aneurysms (IA), is a common cerebrovascular disorder affecting worldwide mortality and morbidity. Approximately, 3.5 million people in the UK are likely to have unruptured IA. While the management of unruptured IA is controversial and not all aneurysms rupture, there is a lot more to understand about the possibility of SAH and risk factors associated with it. Aneurysm morphology may influence aneurysm rupture, and advanced methods investigating the relative importance of morphological data in predicting aneurism rupture have been underutilized. Random forest (RF) classification is applied to a database of over 400 patients attending James Cook Hospital in North East England, UK, to predict rupture of aneurysms. RF showed that when demographic data were used in addition to the morphological data, the prediction accuracy goes up to 71\% with age at haemorrhage, aneurysm dome diameter (d5), log ratio of d5 and aneurysm dome height, years of smoking, log ratio of aneurysm neck diameter and d5 being the top five most important variables. This shows the non-ignorable importance of the factors which are not commonly noted and can be useful to support clinical decisions in this context.
