A4 Refereed article in a conference publication
Automatic Quantification of CT Images for Traumatic Brain Injury
Authors: Koikkalainen J, Lotjonen J, Ledig C, Rueckert D, Tenovuo O, Menon D
Editors: IEEE
Conference name: International Symposium on Biomedical Imaging
Publication year: 2014
Journal: International Symposium on Biomedical Imaging
Book title : 2014 IEEE 11th International Symposium on Biomedical Imaging (ISBI)
Journal name in source: 2014 IEEE 11TH INTERNATIONAL SYMPOSIUM ON BIOMEDICAL IMAGING (ISBI)
Journal acronym: I S BIOMED IMAGING
Series title: IEEE International Symposium on Biomedical Imaging
First page : 125
Last page: 128
Number of pages: 4
eISBN: 978-1-4673-1961-4
ISSN: 1945-7928
DOI: https://doi.org/10.1109/ISBI.2014.6867825
Abstract
Traumatic brain injury (TBI) is a major health problem and the most common cause of permanent disability in people under the age of 40 years. In this paper, we present a fully automatic framework for the analysis of acute computed tomography (CT) images in TBI. Different pathologies common in TBI are quantified and all the information is combined for clinical outcome prediction in individual patients. We propose a multi-template approach for the registration of CT data, which improves the robustness and accuracy of spatial normalization. This is especially important for noisy CT data and TBI images with large areas of pathology. The tissue segmentation methods we use have been optimized to deal with these challenges. The methods we describe have been evaluated on acute CTs from 104 TBI patients. We demonstrate on this dataset that the prediction of dichotomized favorable or unfavorable outcome can be made with an accuracy of 79%.
Traumatic brain injury (TBI) is a major health problem and the most common cause of permanent disability in people under the age of 40 years. In this paper, we present a fully automatic framework for the analysis of acute computed tomography (CT) images in TBI. Different pathologies common in TBI are quantified and all the information is combined for clinical outcome prediction in individual patients. We propose a multi-template approach for the registration of CT data, which improves the robustness and accuracy of spatial normalization. This is especially important for noisy CT data and TBI images with large areas of pathology. The tissue segmentation methods we use have been optimized to deal with these challenges. The methods we describe have been evaluated on acute CTs from 104 TBI patients. We demonstrate on this dataset that the prediction of dichotomized favorable or unfavorable outcome can be made with an accuracy of 79%.
UTU Research Portal