An Optimization of System for Automatic Recognition of Ischemic Stroke Areas in Computed Tomography Images

Informatica ◽  
2007 ◽  
Vol 18 (4) ◽  
pp. 603-614
Author(s):  
Darius Grigaitis ◽  
Vaida Bartkutė ◽  
Leonidas Sakalauskas
Keyword(s):  
Computed Tomography ◽  
Ischemic Stroke ◽  
Automatic Recognition ◽  
Computed Tomography Images

scholarly journals e-ASPECTS software is non-inferior to neuroradiologists in applying the ASPECT score to computed tomography scans of acute ischemic stroke patients

2016 ◽  
Vol 12 (6) ◽  
pp. 615-622 ◽  
Author(s):  
Simon Nagel ◽  
Devesh Sinha ◽  
Diana Day ◽  
Wolfgang Reith ◽  
René Chapot ◽  
...  
Keyword(s):  
Computed Tomography ◽  
Ischemic Stroke ◽  
Correlation Coefficients ◽  
Ground Truth ◽  
Stroke Patients ◽  
Computed Tomography Images ◽  
Contrast Enhanced ◽  
Enhanced Computed Tomography ◽  
Sensitivity Specificity ◽  
Contrast Enhanced Computed Tomography

Background The Alberta Stroke Program Early Computed Tomography Score (ASPECTS) is an established 10-point quantitative topographic computed tomography scan score to assess early ischemic changes. We performed a non-inferiority trial between the e-ASPECTS software and neuroradiologists in scoring ASPECTS on non-contrast enhanced computed tomography images of acute ischemic stroke patients. Methods In this multicenter study, e-ASPECTS and three independent neuroradiologists retrospectively and blindly assessed baseline non-contrast enhanced computed tomography images of 132 patients with acute anterior circulation ischemic stroke. Follow-up scans served as ground truth to determine the definite area of infarction. Sensitivity, specificity, and accuracy for region- and score-based analysis, receiver-operating characteristic curves, Bland-Altman plots and Matthews correlation coefficients relative to the ground truth were calculated and comparisons were made between neuroradiologists and different pre-specified e-ASPECTS operating points. The non-inferiority margin was set to 10% for both sensitivity and specificity on region-based analysis. Results In total 2640 (132 patients × 20 regions per patient) ASPECTS regions were scored. Mean time from onset to baseline computed tomography was 146 ± 124 min and median NIH Stroke Scale (NIHSS) was 11 (6–17, interquartile range). Median ASPECTS for ground truth on follow-up imaging was 8 (6.5–9, interquartile range). In the region-based analysis, two e-ASPECTS operating points (sensitivity, specificity, and accuracy of 44%, 93%, 87% and 44%, 91%, 85%) were statistically non-inferior to all three neuroradiologists (all p-values <0.003). Both Matthews correlation coefficients for e-ASPECTS were higher (0.36 and 0.34) than those of all neuroradiologists (0.32, 0.31, and 0.3). Conclusions e-ASPECTS was non-inferior to three neuroradiologists in scoring ASPECTS on non-contrast enhanced computed tomography images of acute stroke patients.

scholarly journals Automatic Recognition of Anatomical Regions in Computed Tomography Images

2018 ◽  
Vol 62 (4) ◽  
pp. 117-125 ◽  
Author(s):  
Márton József Tóth ◽  
László Ruskó ◽  
Balázs Csébfalvi
Keyword(s):  
Neural Network ◽  
Computed Tomography ◽  
Human Body ◽  
Spatial Coherence ◽  
Automatic Recognition ◽  
Processing Method ◽  
Post Processing ◽  
Computed Tomography Images ◽  
The Neural Network ◽  
Initial Classification

This paper presents a method that can recognize anatomy regions in Computed Tomography (CT) examinations. In this work the human body is divided into eleven regions from the foot to the head. The proposed method consists of two main parts. In the first step, a Convolutional Neural Network (CNN) is used to classify the axial slices of the CT exam. The accuracy of the initial classification is 93.4 %. As the neural network processes the axial slices independently from each other, no spatial coherence is guaranteed. To ensure the contentious labeling the initial classification step is followed by a post-processing method that incorporates the expected order and size of the anatomical regions to improve the labeling. In this way, the accuracy is increased to 94.0 %, the confusion of non-neighboring regions dropped from 1.5 % to 0.0 %. This means that a continuous and outlier free labeling is obtained. The method was trained on a set of 320 CT exams and evaluated on another set of 160 cases.

scholarly journals Texture analysis of computed tomography images of acute ischemic stroke patients

2009 ◽  
Vol 42 (11) ◽  
pp. 1076-1079 ◽  
Author(s):  
M.S. Oliveira ◽  
P.T. Fernandes ◽  
W.M. Avelar ◽  
S.L.M. Santos ◽  
G. Castellano ◽  
...  
Keyword(s):  
Computed Tomography ◽  
Ischemic Stroke ◽  
Acute Ischemic Stroke ◽  
Texture Analysis ◽  
Stroke Patients ◽  
Computed Tomography Images

scholarly journals Free-form tumor synthesis in computed tomography images via richer generative adversarial network

2021 ◽  
Vol 218 ◽  
pp. 106753
Author(s):  
Qiangguo Jin ◽  
Hui Cui ◽  
Changming Sun ◽  
Zhaopeng Meng ◽  
Ran Su
Keyword(s):  
Computed Tomography ◽  
Free Form ◽  
Generative Adversarial Network ◽  
Adversarial Network ◽  
Computed Tomography Images
Sign in / Sign up

Export Citation Format

Share Document