TY - JOUR
T1 - Automated Calculation of Alberta Stroke Program Early CT Score
T2 - Validation in Patients With Large Hemispheric Infarct
AU - Albers, Gregory W.
AU - Wald, Michael J.
AU - Mlynash, Michael
AU - Endres, Juergen
AU - Bammer, Roland
AU - Straka, Matus
AU - Maier, Andreas
AU - Hinson, Holly E.
AU - Sheth, Kevin N.
AU - Taylor Kimberly, W.
AU - Molyneaux, Bradley J.
PY - 2019/11/1
Y1 - 2019/11/1
N2 - Background and Purpose- We compared the Alberta Stroke Program Early CT Score (ASPECTS), calculated using a machine learning-based automatic software tool, RAPID ASPECTS, as well as the median score from 4 experienced readers, with the diffusion-weighted imaging (DWI) ASPECTS obtained following the baseline computed tomography (CT) in patients with large hemispheric infarcts. Methods- CT and magnetic resonance imaging scans from the GAMES-RP study, which enrolled patients with large hemispheric infarctions (82-300 mL) documented on DWI-magnetic resonance imaging, were evaluated by blinded experienced readers to determine both CT and DWI ASPECTS. The CT scans were also evaluated by an automated software program (RAPID ASPECTS). Using the DWI ASPECTS as a reference standard, the median CT ASPECTS of the clinicians and the automated score were compared using the interclass correlation coefficient. Results- The median CT ASPECTS for the clinicians was 5 (interquartile range, 4-7), for RAPID ASPECTS 3 (interquartile range, 1-6), and for DWI ASPECTS 3 (2-4). Median error for RAPID ASPECTS was 1 (interquartile range, -1 to 3) versus 3 (interquartile range, 1-4) for clinicians (P<0.001). The automated score had a higher level of agreement with the median of the DWI ASPECTS, both for the full scale and when dichotomized at <6 versus 6 or more (difference in intraclass correlation coefficient, P=0.001). Conclusions- RAPID ASPECTS was more accurate than experienced clinicians in identifying early evidence of brain ischemia as documented by DWI.
AB - Background and Purpose- We compared the Alberta Stroke Program Early CT Score (ASPECTS), calculated using a machine learning-based automatic software tool, RAPID ASPECTS, as well as the median score from 4 experienced readers, with the diffusion-weighted imaging (DWI) ASPECTS obtained following the baseline computed tomography (CT) in patients with large hemispheric infarcts. Methods- CT and magnetic resonance imaging scans from the GAMES-RP study, which enrolled patients with large hemispheric infarctions (82-300 mL) documented on DWI-magnetic resonance imaging, were evaluated by blinded experienced readers to determine both CT and DWI ASPECTS. The CT scans were also evaluated by an automated software program (RAPID ASPECTS). Using the DWI ASPECTS as a reference standard, the median CT ASPECTS of the clinicians and the automated score were compared using the interclass correlation coefficient. Results- The median CT ASPECTS for the clinicians was 5 (interquartile range, 4-7), for RAPID ASPECTS 3 (interquartile range, 1-6), and for DWI ASPECTS 3 (2-4). Median error for RAPID ASPECTS was 1 (interquartile range, -1 to 3) versus 3 (interquartile range, 1-4) for clinicians (P<0.001). The automated score had a higher level of agreement with the median of the DWI ASPECTS, both for the full scale and when dichotomized at <6 versus 6 or more (difference in intraclass correlation coefficient, P=0.001). Conclusions- RAPID ASPECTS was more accurate than experienced clinicians in identifying early evidence of brain ischemia as documented by DWI.
KW - brain
KW - cerebral infarction
KW - machine learning
KW - magnetic resonance imaging
KW - tomography
UR - http://www.scopus.com/inward/record.url?scp=85074243241&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85074243241&partnerID=8YFLogxK
U2 - 10.1161/STROKEAHA.119.026430
DO - 10.1161/STROKEAHA.119.026430
M3 - Article
C2 - 31500555
AN - SCOPUS:85074243241
SN - 0039-2499
VL - 50
SP - 3277
EP - 3279
JO - Stroke
JF - Stroke
IS - 11
ER -