Multiple sclerosis lesion enhancement and white matter region estimation using hyperintensities in FLAIR images

Autor(es)

Freire, Paulo G. L. ^[1] ; Ferrari, Ricardo J. ^[1]

Afiliação do(s) autor(es)

^[1] Univ Fed Sao Carlos, Dept Comp, Rod Washington Luis, Km 235, BR-13565905 Sao Carlos, SP - Brazil

• Tipo de documento: Artigo Científico
• Fonte: Biomedical Signal Processing and Control; v. 49, p. 338-348, MAR 2019.
• Citações Web of Science: 0

Resumo

FLAIR images are usually acquired in the context of multiple sclerosis (MS) because lesions appear hyperintense in this particular image weight. However, their intensity profile overlaps with white matter (WM) and gray matter (GM) tissues, posing difficulties for accurate segmentation. We propose a lesion enhancement technique to dim down WM and GM regions and highlight hyperintensities by creating a hyperintensity probability map. A byproduct of our proposal is the estimation of a mask encompassing WM and MS lesions. We observed that the lesion intensity profile in FLAIR was 25% and 19% brighter than WM and GM, respectively; comparatively, the same profile in our enhanced images was 444% and 264% brighter. On the mask estimation, we achieved a sensitivity of 99% and specificity of 98%. The results indicate significant improvement in the intensity distinction between lesions and tissues and can aid both experts and automated techniques in segmentation tasks. (C) 2018 Elsevier Ltd. All rights reserved. (AU)

• Processo FAPESP: 16/15661-0 - Esquema computacional automático para segmentação e identificação de processo inflamatório agudo em lesões de esclerose múltipla sem uso de agente de contraste
• Beneficiário: Paulo Guilherme de Lima Freire
• Modalidade de apoio: Bolsas no Brasil - Doutorado