RSS
Логотип
Баннер в шапке 1
Баннер в шапке 2
2018/02/28 17:15:45

ATLAS (Anatomical Tracings of Lesions After Stroke)

.

2018: Emergence in a public access

At the end of February, 2018 it became available to downloading of ATLAS (Anatomical Tracings of Lesions After Stroke) — the largest open base of MRT-pictures of a brain of patients who had a stroke. Already 33 research groups loaded into base data for development and testing of the algorithms which are automatically processing MRT-images worldwide.[1]

Historically it developed so that sharing of data was contrary to the normal competition between research groups. However for training and testing of artificial intelligence extensive databases are required.

Fragment of the ATLAS database
File:Aquote1.png
Sponsors and scientific communities at last recognized the value of team work and data exchange for promotion of scientific achievements — noted one of authors of the research Suk-Ley Lew (Sook-Lei Liew).
File:Aquote2.png

Lew and her command are going to use algorithms for accomplishment of meta-analysis of several thousand MRT-pictures to study how separate anatomic damages influence the recovery period. This information will help scientists to define biological markers on which it will be possible to foretell what rehabilitation technicians and the personalized plans of treatment will be effective at certain patients.

Earlier scientists had to outline manually borders of damages that, certainly, cannot be done on thousands of MRT-images. Developers of algorithms will be able to automate this process.

Lew and her command already test all existing algorithms of hands-off processing of similar images to reveal the most exact systems. Meanwhile the best results were shown by an algorithm "Identification of Damages with Data Analysis of the Neighbouring Areas" (LINDA).

File:Aquote1.png
Nevertheless, even the best algorithms still fail in some cases therefore we insist that at all stages of the automated processing visual and manual control from the researcher is necessary, - Lew says. - But even such approach substantially saves time of researchers.[2]
File:Aquote2.png

Notes