Luca Giancardo1, Sunil A Sheth2, Alexandra L Czap2, Mersedeh Bahr-Hosseini3, Noopur Singh4, Jose-Miguel Yamal4, May Nour3, Stephanie Parker2, Youngran Kim2, Lucas Restrepo3, Rania Abdelkhaleq2, Sergio Salazar-Marioni2, Kenny Phan2, Ritvij Bowry2, Suja S Rajan5, James C Grotta6, Jeffrey L Saver3. 1. Center for Precision Health, UTHealth School of Biomedical Informatics, UTHealth McGovern Medical School, Houston, TX (L.G.). 2. Department of Neurology, UTHealth McGovern Medical School, Houston TX (A.L.C., S.P., Y.K., R.A., S.S.-M., K.P., R.B., S.A.S.). 3. Department of Neurology and Comprehensive Stroke Center, UCLA, Los Angeles, CA (M.B.-H., M.N., L.R., J.L.S.). 4. Department of Biostatistics and Data Science, School of Public Health, University of Texas Health Sciences Center at Houston (N.S., J.-M.Y.). 5. Department of Management, Policy and Community Health, School of Public Health, University of Texas Health Sciences Center at Houston (S.S.R.). 6. Clinical Innovation and Research Institute, Memorial Hermann Hospial Texas Medical Center, Houston (J.C.G.).
Abstract
BACKGROUND: Prehospital automated large vessel occlusion (LVO) detection in Mobile Stroke Units (MSUs) could accelerate identification and treatment of patients with LVO acute ischemic stroke. Here, we evaluate the performance of a machine learning (ML) model on CT angiograms (CTAs) obtained from 2 MSUs to detect LVO. METHODS: Patients evaluated on MSUs in Houston and Los Angeles with out-of-hospital CTAs were identified. Anterior circulation LVO was defined as an occlusion of the intracranial internal carotid artery, middle cerebral artery (M1 or M2), or anterior cerebral artery vessels and determined by an expert human reader. A ML model to detect LVO was trained and tested on independent data sets consisting of in-hospital CTAs and then tested on MSU CTA images. Model performance was determined using area under the receiver-operator curve statistics. RESULTS: Among 68 patients with out-of-hospital MSU CTAs, 40% had an LVO. The most common occlusion location was the middle cerebral artery M1 segment (59%), followed by the internal carotid artery (30%), and middle cerebral artery M2 (11%). Median time from last known well to CTA imaging was 88.0 (interquartile range, 59.5-196.0) minutes. After training on 870 in-hospital CTAs, the ML model performed well in identifying LVO in a separate in-hospital data set of 441 images with area under receiver-operator curve of 0.84 (95% CI, 0.80-0.87). ML algorithm analysis time was under 1 minute. The performance of the ML model on the MSU CTA images was comparable with area under receiver-operator curve 0.80 (95% CI, 0.71-0.89). There was no significant difference in performance between the Houston and Los Angeles MSU CTA cohorts. CONCLUSIONS: In this study of patients evaluated on MSUs in 2 cities, a ML algorithm was able to accurately and rapidly detect LVO using prehospital CTA acquisitions.
BACKGROUND: Prehospital automated large vessel occlusion (LVO) detection in Mobile Stroke Units (MSUs) could accelerate identification and treatment of patients with LVO acute ischemic stroke. Here, we evaluate the performance of a machine learning (ML) model on CT angiograms (CTAs) obtained from 2 MSUs to detect LVO. METHODS: Patients evaluated on MSUs in Houston and Los Angeles with out-of-hospital CTAs were identified. Anterior circulation LVO was defined as an occlusion of the intracranial internal carotid artery, middle cerebral artery (M1 or M2), or anterior cerebral artery vessels and determined by an expert human reader. A ML model to detect LVO was trained and tested on independent data sets consisting of in-hospital CTAs and then tested on MSU CTA images. Model performance was determined using area under the receiver-operator curve statistics. RESULTS: Among 68 patients with out-of-hospital MSU CTAs, 40% had an LVO. The most common occlusion location was the middle cerebral artery M1 segment (59%), followed by the internal carotid artery (30%), and middle cerebral artery M2 (11%). Median time from last known well to CTA imaging was 88.0 (interquartile range, 59.5-196.0) minutes. After training on 870 in-hospital CTAs, the ML model performed well in identifying LVO in a separate in-hospital data set of 441 images with area under receiver-operator curve of 0.84 (95% CI, 0.80-0.87). ML algorithm analysis time was under 1 minute. The performance of the ML model on the MSU CTA images was comparable with area under receiver-operator curve 0.80 (95% CI, 0.71-0.89). There was no significant difference in performance between the Houston and Los Angeles MSU CTA cohorts. CONCLUSIONS: In this study of patients evaluated on MSUs in 2 cities, a ML algorithm was able to accurately and rapidly detect LVO using prehospital CTA acquisitions.
Authors: Bruce C V Campbell; Søren Christensen; Brian M Tress; Leonid Churilov; Patricia M Desmond; Mark W Parsons; P Alan Barber; Christopher R Levi; Christopher Bladin; Geoffrey A Donnan; Stephen M Davis Journal: J Cereb Blood Flow Metab Date: 2013-05-08 Impact factor: 6.200
Authors: Henry Zhao; Skye Coote; Damien Easton; Francesca Langenberg; Michael Stephenson; Karen Smith; Stephen Bernard; Dominique A Cadilhac; Joosup Kim; Christopher F Bladin; Leonid Churilov; Douglas E Crompton; Helen M Dewey; Lauren M Sanders; Tissa Wijeratne; Geoffrey Cloud; Duncan M Brooks; Hamed Asadi; Vincent Thijs; Ronil V Chandra; Henry Ma; Patricia M Desmond; Richard J Dowling; Peter J Mitchell; Nawaf Yassi; Bernard Yan; Bruce C V Campbell; Mark W Parsons; Geoffrey A Donnan; Stephen M Davis Journal: Stroke Date: 2020-02-12 Impact factor: 7.914
Authors: Beau Norgeot; Giorgio Quer; Brett K Beaulieu-Jones; Ali Torkamani; Raquel Dias; Milena Gianfrancesco; Rima Arnaout; Isaac S Kohane; Suchi Saria; Eric Topol; Ziad Obermeyer; Bin Yu; Atul J Butte Journal: Nat Med Date: 2020-09 Impact factor: 53.440
Authors: Alexandra L Czap; Noopur Singh; Ritvij Bowry; Amanda Jagolino-Cole; Stephanie A Parker; Kenny Phan; Mengxi Wang; Sunil A Sheth; Suja S Rajan; Jose-Miguel Yamal; James C Grotta Journal: Stroke Date: 2020-04-16 Impact factor: 7.914
Authors: Emmad M Qazi; Sung Il Sohn; Sachin Mishra; Mohammed A Almekhlafi; Muneer Eesa; Christopher D d'Esterre; Abdul A Qazi; Josep Puig; Mayank Goyal; Andrew M Demchuk; Bijoy K Menon Journal: Can J Neurol Sci Date: 2015-09-14 Impact factor: 2.104
Authors: Michael T Froehler; Jeffrey L Saver; Osama O Zaidat; Reza Jahan; Mohammad Ali Aziz-Sultan; Richard P Klucznik; Diogo C Haussen; Frank R Hellinger; Dileep R Yavagal; Tom L Yao; David S Liebeskind; Ashutosh P Jadhav; Rishi Gupta; Ameer E Hassan; Coleman O Martin; Hormozd Bozorgchami; Ritesh Kaushal; Raul G Nogueira; Ravi H Gandhi; Eric C Peterson; Shervin R Dashti; Curtis A Given; Brijesh P Mehta; Vivek Deshmukh; Sidney Starkman; Italo Linfante; Scott H McPherson; Peter Kvamme; Thomas J Grobelny; Muhammad S Hussain; Ike Thacker; Nirav Vora; Peng Roc Chen; Stephen J Monteith; Robert D Ecker; Clemens M Schirmer; Eric Sauvageau; Alex Abou-Chebl; Colin P Derdeyn; Lucian Maidan; Aamir Badruddin; Adnan H Siddiqui; Travis M Dumont; Abdulnasser Alhajeri; M Asif Taqi; Khaled Asi; Jeffrey Carpenter; Alan Boulos; Gaurav Jindal; Ajit S Puri; Rohan Chitale; Eric M Deshaies; David H Robinson; David F Kallmes; Blaise W Baxter; Mouhammad A Jumaa; Peter Sunenshine; Aniel Majjhoo; Joey D English; Shuichi Suzuki; Richard D Fessler; Josser E Delgado Almandoz; Jerry C Martin; Nils H Mueller-Kronast Journal: Circulation Date: 2017-09-24 Impact factor: 29.690
Authors: Samuel G Thorpe; Corey M Thibeault; Nicolas Canac; Kian Jalaleddini; Amber Dorn; Seth J Wilk; Thomas Devlin; Fabien Scalzo; Robert B Hamilton Journal: PLoS One Date: 2020-02-06 Impact factor: 3.240
Authors: Paul Reidler; Lena Stueckelschweiger; Daniel Puhr-Westerheide; Katharina Feil; Lars Kellert; Konstantinos Dimitriadis; Steffen Tiedt; Moriz Herzberg; Jan Rémi; Thomas Liebig; Matthias P Fabritius; Wolfgang G Kunz Journal: Clin Neuroradiol Date: 2020-09-16 Impact factor: 3.649