"lyngby" a Matlab toolbox for fMRI analysis: Results of a user questionnaire
NEUROIMAGE
Functional mapping by MRI and PET provides unique access to spatio-temporal patterns of activity in the working human brain, The intricate combination of biological and engineering mechanisms involved in the acquission of functional brain images gives room for rich and expanding modeling research activity. To further this research we have developed a software toolbox based on the popular Matlab platform. The toolbox “lyngby” was first introduced in 1999[ I]. In 2000 we extended the set of models for spatio-temporal analysis of image sets now including:
. Baseline estimation for detrending.
. Single pixel activity maps: t, F, and KS tests.
. Single pixel hemodynamic response models: Lange-Zeger, FIR filter, regularized FIR filters and artiticial neural networks.
. Clustering: time series, correlation, or feature based.
. Multivariate models: CVA, CRA, artificial neural networks. To further enhance the usability of the toolbox we emailed a questionnaire to 270 registered users. We obtained substantial feedback from N=28 users. While the main part the results is relevant only to the “lyngby” development team, we here briefly summarize some aspects of more generic neuroinformatics interest.
The 28 respondents were about equally split across medical and technical backgrounds. About 90% of the usage is concerned with fMR1. Of the respondents 60% are concemdd with modeling fMR1 time series, while 75% report interest in the more conventional neuroimaging objective of functional localization. More than 70% used their own in-house analysis code, well aligned with the fact that “lyngby” is conceived as a platform for integrating and comparing multiple models.
We also asked the users about which of “1yngby”'s models they used and the result is summarized in the figure. More than half of the respondents use crosscorrelation, while about 40% report analysing time seties with flexible linear convolution methods (FIR filters). About a third use time series clustering (K-means) and a similar fraction use the SVD for multivariate analysis. 20% report that they use non-linear time series analysis by artificial neural networks.
Finally, we asked about one of the major sources of user problems, namely, input data formats. “lyngby” is currently supporting several formats including “Analyze”. Analyse is the preferred format among the users, however, more than 20% mentioned DICOM in their feedback. Our current aims include making the Analyse input filter more robust and we plan to add support for basic DICOM functionality.
Availability. “lyngby” is freely available to the functional brain imaging community for non-commercial use, and we invite you to email lyngby@hendrix.imm.dtu.dk to obtain further information and a personal pasacode to the lyngby Web site. The toolbox documentation is available in postscript form (including installation guide): http://hendrix.imm.dtu.dk/software/lyngby.
13 (6)
2001
S145
Hansen L.K., Nielsen F.A., Liptrot M., Strother S.C., Lange N., Gade A., Rottenberg D.A. & Paulson O.B.
Seventh annual meeting of the Organization for Human Brain Mapping :
OHBM
Meeting Abstract
Brighton, UK (June 10–14, 2001)
Part: Part 2 Suppl. S, Supplement: Part 2 Suppl. S Publisher: ACADEMIC PRESS INC, 525 B ST, STE 1900, SAN DIEGO, CA 92101-4495 USA ISSN: 1053-8119
Baycrest is an academic health sciences centre fully affiliated with the University of Toronto
Privacy Statement | Disclaimer | © 2012 Baycrest. All Rights Reserved.