Transformer-based precision in 3D MRI stroke lesion segmentation

DOI

Search this article

Description

<p>Strokes are the leading cause of mortality worldwide. According to the WHO, 15 million people encounter strokes each year. Of these, 5 million die, and another 5 million become permanently crippled, putting a strain on families and communities. Three-dimensional stroke lesion segmentation is a challenging task in magnetic resonance imaging (MRI) due to the complicated cerebral anatomy and complexity of stroke-affected tissue. To address these challenges, we present a new segmentation framework based on cutting-edge transformer architecture that is specifically designed for the challenging high-dimensional spatial context of MRI scans. We pioneered this approach by adapting transformer models that have previously been successful in NLP to capture the intricate spatial relationships found in volumetric brain imaging. The proposed method enables transformers' unique capabilities in processing large-scale, complex datasets, resulting in significant advances in automated stroke diagnosis and treatment planning.</p>

Journal

Details 詳細情報について

  • CRID
    1390583424884916224
  • DOI
    10.11239/jsmbe.annual62.226_1
  • ISSN
    18814379
    1347443X
  • Text Lang
    ja
  • Data Source
    • JaLC
  • Abstract License Flag
    Disallowed

Report a problem

Back to top