High speed data processing for imaging MS-based molecular histology using graphical processing units.
J Am Soc Mass Spectrom
; 23(4): 745-52, 2012 Apr.
Article
em En
| MEDLINE
| ID: mdl-22311727
ABSTRACT
Imaging MS enables the distributions of hundreds of biomolecular ions to be determined directly from tissue samples. The application of multivariate methods, to identify pixels possessing correlated MS profiles, is referred to as molecular histology as tissues can be annotated on the basis of the MS profiles. The application of imaging MS-based molecular histology to larger tissue series, for clinical applications, requires significantly increased computational capacity in order to efficiently analyze the very large, highly dimensional datasets. Such datasets are highly suited to processing using graphical processor units, a very cost-effective solution for high speed processing. Here we demonstrate up to 13× speed improvements for imaging MS-based molecular histology using off-the-shelf components, and demonstrate equivalence with CPU based calculations. It is then discussed how imaging MS investigations may be designed to fully exploit the high speed of graphical processor units.
Texto completo:
1
Coleções:
01-internacional
Base de dados:
MEDLINE
Assunto principal:
Espectrometria de Massas
/
Processamento de Imagem Assistida por Computador
/
Imagem Molecular
/
Histocitoquímica
Limite:
Animals
Idioma:
En
Revista:
J Am Soc Mass Spectrom
Ano de publicação:
2012
Tipo de documento:
Article
País de afiliação:
Holanda