Prospective Validation of Colored Non-Enhanced Head CT Images for Detecting Acute Stroke in the Setting of a Code Gray
Khalid Manzoul1, Charlotte Tylor1, Jeffery Hooker1, Elliot Varney1, Seth Lirette2, David Gordy1, Ramin Hamidi1, David Joyner1, Todd Nichols1, Ellen Parker1, Sam Baird1, Tamvir Rizvi1, Charlene Claudio1, Juliana Sitta1, Niki Patel1, Rana Gordji1, Amy Krecker1, Gerri Wilson1, Stella Powell1, Edward Florez1, Andrew Smith3, Candace Howard1
1Department of Radiology, University of Mississippi Medical Center, Jackson, MS
2Department of Data Science, University of Mississippi Medical Center, Jackson, MS
3Department of Radiology, University of Alabama at Birmingham, Birmingham, AL
Non-enhanced CT imaging has a very important diagnostic role in the setting of a Code-Gray, and rapid assessment and diagnosis is crucial for optimal patient outcome. For this prospective observational study, 100 consecutive Code Gray adult patients from 2/1/2018 to 3/17/2018 with a gold-standard confirmatory MRI were included in an effort to decrease time and maintain diagnostic accuracy of acute ischemic stroke detection. Grayscale NECT were collected and colorized NECT images were generated using a previously established method. All images were de-identified, and two randomized reading sets (100 grayscale NECT and 100 grayscale+colored NECT images) were generated. Four experienced readers independently assessed each reading set. Reading sessions were separated by >2weeks to minimize recall bias. The mean accuracy, sensitivity, specificity and time of assessment were compared between grayscale and grayscale+color images in a multivariate model. Among the 4 readers, the mean accuracy/sensitivity/specificity for correctly diagnosing acute ischemic stroke were 72%/46%/87% using only grayscale images and 69%/36%/86% using grayscale images+color NECT images (p=0.08/p=0.006/p=0.858). Mean time of interpretation of 59 seconds using grayscale only images decreased by 19 seconds for interpretation using grayscale+color images (p<0.001). Significantly decreasing the time of assessment without degrading diagnostic accuracy could widely applicable among general radiologists with no additional patient expense or radiation potentially improve stroke outcomes.