English   español  
Please use this identifier to cite or link to this item: http://hdl.handle.net/10261/104129
Share/Impact:
Statistics
logo share SHARE   Add this article to your Mendeley library MendeleyBASE
Visualizar otros formatos: MARC | Dublin Core | RDF | ORE | MODS | METS | DIDL
Exportar a otros formatos:

Title

In vivo proton magnetic resonance spectroscopy regional metabolic profiles: differences among hippocampal formation, prefrontal cortex and cerebellum in adult rats using a machine learning approach

AuthorsGuadaño-Ferraz, Ana ; Fernández-Lamo, Iván ; Pacheco-Torres, Jesús; Gómez-Andrés, David; Pulido-Valdeolivas, Irene; López-Larrubia, Pilar ; Montero-Pedrazuela, Ana
Issue Date2012
CitationNeuroscience 2012
Abstract1H-Magnetic resonance spectroscopy (1H-MRS) is a non-invasive technology that provides information about several metabolite levels among multiple brain regions. This technology is widely used in preclinical and neuroscience research and in diagnosis in neurological practice. Although CNS is highly heterogeneous, few studies have been focused on interregional differences in 1H-MRS metabolic profiles and fewer, if any, have used multivariate approaches. Our aim is the development of a methodology to compare 1H-MRS metabolic profiles in order to improve the 1H-MRS result interpretation. In vivo spectra were acquired with a 7T PharmaScan® System (voxel size 27 mm3 and PRESS sequence, TE=35 ms) and fitted with LCModel. To avoid variance in signals, metabolite levels were normalized to NAA (N-Acetyl Aspartate). Four ratios to NAA were obtained: NAA+NAAG (NAA+N-Acetyl Aspartyl Glutamate), Cr+PCr (total Creatine), GPC+PCh (total Choline) and Glx (Glutamate+Glutamine) in hippocampal formation, prefrontal cortex and cerebellum from adult Wistar rats, aged 80, 105 and 130 days (n=8 each condition). Linear discriminant analysis (LDA), random forests (RF) and supervised neural networks (SNN) were used to classify brain regions according to age and NAA ratios. Cohen¿s kappa and the area under the receiver operating characteristic curve (AUC) were respectively calculated to measure global performance of each approach and each region in particular. Confidence intervals were calculated by bootstrapping. To measure discriminating importance of each dependent variable, the correlations among them and linear discriminant functions in LDA and the mean decrease in accuracy in RF were used. LDA, RF and SNN were able to discriminate among regions (Table 1). LDA showed the best global fitting. Compared to the others, the cerebellum metabolic profile was the most different. Cr+PCr and Glx were important in separating cerebellum and hippocampal formation respectively. Age was not an important factor in the models. This study provides a valid methodology to explore 1H-MRS interregional metabolic profiles detecting significant regional differences that are important in 1H-MRS result interpretation. This fact should be considered in clinical context.
DescriptionResumen del póster presentado al Neuroscience celebrado en Nueva Orleans (USA) del 13 al 17 de octubre de 2012.
URIhttp://hdl.handle.net/10261/104129
Appears in Collections:(IIBM) Comunicaciones congresos
Files in This Item:
File Description SizeFormat 
accesoRestringido.pdf15,38 kBAdobe PDFThumbnail
View/Open
Show full item record
Review this work
 


WARNING: Items in Digital.CSIC are protected by copyright, with all rights reserved, unless otherwise indicated.