Alberta Stroke Program Early CT Score Region Segmentation

Proceedings of ‏The 4th International Conference on Modern Research in Science, Engineering and Technology

Year: 2021

DOI:

[Fulltext PDF]

Alberta Stroke Program Early CT Score Region Segmentation

Rafael de Freitas Brito, Paulo Antonio Guimarães Bettero , Antônio Cláudio Paschoarelli Veiga1 and João Batista Destro Filho

 

ABSTRACT: 

Stroke is one of the leading causes of death and disability in the world, affecting low and middle-income countries in particular. The Alberta Stroke Program Early CT Score (ASPECTS) quantifies the extent of early ischemic changes in computed tomographies (CTs) of stroke patients and is widely used as a patient selection tool in stroke care. The score, however, isn’t very reliable (has low interrater agreement), such that computational methods for assisted or even automated diagnostics could improve its reliability and, consequently, stroke care. Some automated ASPECTS scoring tools have already been developed to this day, both in an academic and a commercial context, and although almost all of them perform some sort of region segmentation, this step is often very briefly discussed in ASPECTS automation papers. Given this gap in literature, this paper proposes and evaluates an ASPECTS region segmentation algorithm, using a public CT template, CT database (CQ500) and a public library for image processing (SimpleITK). The proposed method is composed by four steps: pre-processing/slice selection, linear registration, non-linear registration and display/evaluation. Overall, it obtained a mean Dice coefficient of 0.6587 with a standard deviation of 0.0595 and a mean Hausdorff distance of 14.3903 with a standard deviation of 4.4366, for the 10 CTs evaluated.

Keywords: ASPECTS; Atlas Registration; Computer-aided Diagnosis; Computed Tomography; Digital Image Processing.