Purpose
Stroke is a major disease of the 21st century. Worldwide, it affects about 15 million people each year [1,2]. It is a leading cause of long-term disability and the second leading cause of death [3]. In the past 20 years, there has been a significant increase in the absolute number of stroke cases, with an increase of 70.0% in incidence, 102.0% in prevalence, and 43.0% in the number of deaths [4,5].Stroke is classified in two types: hemorrhagic stroke which is caused by a leak of...
Methods and materials
All patients who underwent NCCT and/or CTA scans for stroke suspicion between March 2019 and July 2020 at two hospitals (La Timone and Hopital-Nord, Marseille, France) were retrospectively and consecutively collected. In order to establish the ground truth (GT), two board-certified neuroradiologists analyzed the scans and defined by consensus the presence of ICHs (and their subtypes: Intraparenchymal, Intraventricular and/or Extra-axial) and/or LVOs (and their locations in the Internal Carotid and/or Middle Cerebral Arteries segments: ICA, MCA-M1 and/or MCA-M2). In addition, if the case presented a...
Results
A total of 405 patients were included in the study. Mean age was 64.9 yo ± SD:18.9 and 52.6% were female. Among the 405 patients, there were 373 NCCT and 331 CTA. According to the GT, there were 27.9% positive NCCT for ICH, 25.1% positive CTA for LVO and the mean ASPECT Score was 9.3.The overall per-case accuracy of the algorithm was 93.3% [95% CI: 90.3% - 95.6%] for ICH, and 86.4% [95% CI: 82.2% - 89.9%] for LVO. The algorithm performances for each type...
Conclusion
The AI-based application demonstrated high performance for the detection of cases with ICH and LVO. This high performance was consistent among ICH subtypes and LVO segments. Similarly, the software presented a high per-region accuracy for the computation of the ASPECT score and all the ASPECTS regions had individual accuracies higher than 81%. These results indicate not only that the algorithm performs adequately but that its high accuracy is generalizable across heterogeneous subgroups of stroke types. Regarding IS, the algorithm was capable of correctly classifying almost...
Personal information and conflict of interest
A-A. El-Ahmadi:
Nothing to disclose
G. Brun:
Nothing to disclose
A. Ayobi:
Employee: Avicenna.AI, La Ciotat, France
S. Quenet:
Employee: Avicenna.AI, La Ciotat, France
Y. Chaibi:
Employee: Avicenna.AI, La Ciotat, France
A. Reyre:
Nothing to disclose
A. Jacquier:
Nothing to disclose
N. Girard:
Nothing to disclose
References
[1] T.F. Hasan, H. Hasan, R.E. Kelley, Overview of Acute Ischemic Stroke Evaluation and Management, Biomedicines 9 (2021) 1486. https://doi.org/10.3390/biomedicines9101486.[2] J. Puig, J. Shankar, D. Liebeskind, M. Terceño, K. Nael, A.M. Demchuk, B. Menon, D. Dowlatshahi, C. Leivaâ€ÂSalinas, M. Wintermark, G. Thomalla, Y. Silva, J. Serena, S. Pedraza, M. Essig, From “Time is Brain” to “Imaging is Brain”: A Paradigm Shift in the Management of Acute Ischemic Stroke, J. Neuroimaging 30 (2020) 562–571. https://doi.org/10.1111/jon.12693.[3] D. Zhang, X. Zou, C. Sy, H. Qin, Y. Wang, X....