Creation of a Dataset of MSCT-Images and Clinical Data for Acute Cerebrovascular Events

Background The use of neuroimaging methods is an integral part of the process of assisting patients with acute cerebrovascular events (ACVE), and computed tomography (CT) is the «gold standard» for examining this category of patients. The capabilities of the analysis of CT images may be significantly expanded with modern methods of machine learning including the application of the principles of radiomics. However, since the use of these methods requires large arrays of DICOM (Digital Imaging and Communications in Medicine)-images, their implementation into clinical practice is limited by the lack of representative sample sets. Inaddition, at present, collections (datasets) of CT images of stroke patients, that are suitable for machine learning, are practically not available in the public domain.

Aim of study Regarding the aforesaid, the aim of this work was to create a DICOM images dataset of native CT and CT-angiography of patients with different types of stroke. Material and meth ods The collection was based on the medical cases of patients hospitalized in the Regional Vascular Center of the N.V. Sklifosovsky Research Institute for Emergency Medicine. We used a previously developed specialized platform to enter clinical data on the stroke cases, to attach CT DICOMimages to each case, to contour 3D areas of interest, and to tag (label) them. A dictionary was developed for tagging, where elements describe the type of lesion, location, and vascular territory.

Results A dataset of clinical cases and images was formed in the course of the work. It included anonymous information about 220 patients, 130 of them with ischemic stroke, 40 with hemorrhagic stroke, and 50 patients without cerebrovascular disorders. Clinical data included information about type of stroke, presence of concomitant diseases and complications, length of hospital stay, methods of treatment, and outcome. The results of 370 studies of native CT and 102 studies of CT-angiography were entered for all patients. The areas of interest corresponding to direct and indirect signs of stroke were contoured and tagged by radiologists on each series of images.

Conclusion The resulting collection of images will enable the use of various methods of data analysis and machine learning in solving the most important practical problems including diagnosis of the stroke type, assessment of lesion volume, and prediction of the degree of neurological deficit.

1. Kornienko VN, Pronina IN. (eds.) Diagnosticheskaya neyroradiologiya. Moscow: Izdatel’stvo:T. M. Andreeva Publ.; 2006. (In Russ.)

2. Suslina ZA, Piradov MA. (eds.) Insul’t: diagnostika, lechenie, profilaktika. Moscow: MEDpress–inform Publ.; 2008. (In Russ.)

3. Piradov MA, Krylov VV, Belkin AA, Petrikov SS. Insul’ty. In: Gel’fanda BR, Zabolotskikh IB (eds.) Intensivnaya terapiya. 2nd ed., rev. and exp. Moscow: GEOTAR-Media Publ.; 2017. Ch. 2. pp. 288–309. (In Russ.)

4. Suslina ZA, Varakin YuYa. Clinico-epidemiological studies a perspective direction for investigation of cerebrovascular disorders. Annals of Clinical and Experimental Neurology. 2009;3(3):4–11. (In Russ.)

5. Vereshchagin NV, Piradov MA, Suslina ZA. Printsipy diagnostiki i lecheniya bol’nykh v ostrom periode insul’ta. Atmosfera. Nervnye bolezni. 2002;(1):8–14. (In Russ.)

6. Chilamkurthy S, Ghosh R, Tanamala S, Biviji M, Campeau NG, Venugopal VK, et al. Development and Validation of Deep Learning Algorithms for Detection of Critical Findings in Head CT Scans. arXiv:1803.05854 [cs. CV]. 2018. Available at: https://arxiv.org/abs/1803.05854 [Accessed 21 May 2020].

7. Hajimani E, Ruano MG, Ruano AE. An intelligent support system for automatic detection of cerebral vascular accidents from brain CT images. Comput Methods Programs Biomed. 2017;146:109-123. PMID: 28688480 https://doi.org/10.1016/j.cmpb.2017.05.005

8. ISLES: Ischemic Stroke Lesion Segmentation Challenge. Available at: http://www.isles-challenge.org/ [Accessed 21 May 2020].

9. Dolotova D, Donitova V, Arhipov I, Sharifullin F, Zagriazkina T, Kobrinskii B, et al. A Platform for Collection and Analysis of Image Data on Stroke. Stud Health Technol Inform. 2019;262:312–315. PMID: 31349330 https://doi.org/10.3233/SHTI190081

10. Mikhail P, Le MGD, Mair G. Computational Image Analysis of Nonenhanced Computed Tomography for Acute Ischaemic Stroke: A Systematic Review. J Stroke Cerebrovasc Dis. 2020;29(5):104715. doi:10.1016/j.jstrokecerebrovasdis.2020.104715

11. Hssayeni M. Computed Tomography Images for Intracranial Hemorrhage Detection and Segmentation (version 1.3.1). PhysioNet. 2020. https://doi.org/10.13026/4nae-zg36.

12. Enigma stroke recovery. Available at: http://enigma.ini.usc.edu/ongoing/enigma-stroke-recovery/ [Accessed 21 May 2020].

13. Liew S, Anglin J, Banks N, et al. A large, open source dataset of stroke anatomical brain images and manual lesion segmentations. Sci Data. 2018;5:180011. https://doi.org/10.1038/sdata.2018.11