Format

Send to

Choose Destination
Neuroimage. 2016 Aug 15;137:1-8. doi: 10.1016/j.neuroimage.2016.05.055. Epub 2016 May 25.

Improved spatial accuracy of functional maps in the rat olfactory bulb using supervised machine learning approach.

Author information

1
Department of Ophthalmology, University of Pittsburgh School of Medicine, Pittsburgh, PA, USA; Department of Radiology, University of Pittsburgh School of Medicine, Pittsburgh, PA, USA. Electronic address: murphymc@pitt.edu.
2
Department of Radiology, University of Pittsburgh School of Medicine, Pittsburgh, PA, USA.
3
Department of Ophthalmology, University of Pittsburgh School of Medicine, Pittsburgh, PA, USA; Department of Bioengineering, University of Pittsburgh, Pittsburgh, PA, USA.
4
Center for Neuroscience Imaging Research, Institute for Basic Science, Sungkyunkwan University, Suwon, Republic of Korea; Department of Biomedical Engineering, Sungkyunkwan University, Suwon, Republic of Korea; Department of Biological Sciences, Sungkyunkwan University, Suwon, Republic of Korea.

Abstract

Functional MRI (fMRI) is a popular and important tool for noninvasive mapping of neural activity. As fMRI measures the hemodynamic response, the resulting activation maps do not perfectly reflect the underlying neural activity. The purpose of this work was to design a data-driven model to improve the spatial accuracy of fMRI maps in the rat olfactory bulb. This system is an ideal choice for this investigation since the bulb circuit is well characterized, allowing for an accurate definition of activity patterns in order to train the model. We generated models for both cerebral blood volume weighted (CBVw) and blood oxygen level dependent (BOLD) fMRI data. The results indicate that the spatial accuracy of the activation maps is either significantly improved or at worst not significantly different when using the learned models compared to a conventional general linear model approach, particularly for BOLD images and activity patterns involving deep layers of the bulb. Furthermore, the activation maps computed by CBVw and BOLD data show increased agreement when using the learned models, lending more confidence to their accuracy. The models presented here could have an immediate impact on studies of the olfactory bulb, but perhaps more importantly, demonstrate the potential for similar flexible, data-driven models to improve the quality of activation maps calculated using fMRI data.

PMID:
27236085
PMCID:
PMC4914461
DOI:
10.1016/j.neuroimage.2016.05.055
[Indexed for MEDLINE]
Free PMC Article

Supplemental Content

Full text links

Icon for Elsevier Science Icon for PubMed Central
Loading ...
Support Center