University of Twente Student Theses

Login

Computer-aided diagnosis for CT based clinical triage of ischemic and hemorrhagic stroke patients : a deep learning and quantitative image analysis approach

Bergmans, R.H.J. (2019) Computer-aided diagnosis for CT based clinical triage of ischemic and hemorrhagic stroke patients : a deep learning and quantitative image analysis approach.

[img] PDF
3MB
Abstract:Stroke is sudden development in cerebral disturbance due to insufficient blood supply by either blockage (ischemic) or bleeding (hemorrhagic) of an artery in the brain. The aim of this thesis is to explore the ability of deep learning and quantitative image analysis to aid in triaging these patients. For ischemic stroke, therapy decision depends on the extent of the infarct core, which is best characterized on diffusion-weighted imaging (DWI), while CT is the commonly available imaging modality in emergency radiology. A deep learning system has been developed that is able to predict the extent of DWI infarct core from 3-phase CT data. Predicted infarct core lesion volumes led to low incorrect therapy decision rates, implying potential clinical usability. For hemorrhagic stroke, likelihood of hematoma expansion is an important factor in therapy selection. The recently proposed I2-score prediction model for hematoma expansion, based on quantitative dual-energy CT (DECT) angiography iodine features, has been modified to a model combining DECT iodine features with quantitative unenhanced CT texture features. Feature selection and ranking showed that DECT iodine features perform better than NCCT texture features, and that a combined iodine-texture model was not able to outperform the I2-score.
Item Type:Essay (Master)
Clients:
Massachusetts General Hospital, Boston, USA
Faculty:TNW: Science and Technology
Subject:44 medicine, 50 technical science in general, 54 computer science
Programme:Technical Medicine MSc (60033)
Link to this item:http://purl.utwente.nl/essays/77302
Export this item as:BibTeX
EndNote
HTML Citation
Reference Manager

 

Repository Staff Only: item control page