Skip to content

Neuroimaging Informatics Technology Initiative

Sections
Personal tools
You are here: Home » Data Format Working Group » Presentations » NIfTI-1 Abstract

NIfTI-1 Abstract

Document Actions
submitted to the 10th Annual Meeting of the Organization for Human Brain Mapping
10th Annual Meeting of the Organization for Human Brain Mapping

Abstract Number: 214.110001
Submitted By: Robert Cox
Last Modified: 9 Jan 04




SPLTRAK Abstract Submission


A (Sort of) New Image Data Format Standard: NIfTI-1
Robert W Cox1 , John Ashburner,2 , Hester Breman,3 , Kate Fissell,4 , Christian Haselgrove,5 , Colin J Holmes,6 , Jack L Lancaster,7 , David E Rex,8 , Stephen M Smith,9 , Jeffrey B Woodward,10 , Stephen C Strother,11
1NIMH/NIH/DHHS/Bethesda, 2FIL/London, 3Brain Innovation/Maastricht, 4U Pittsburgh/Pittsburgh, 5MGH/Charlestown, 6SGI/Mountain View, 7RIC/UTHSCSA/San Antonio, 8LONI/UCLA/Los Angeles, 9FMRIB/Oxford, 10Dartmouth College/Hanover, 11U Minnesota/Minneapolis and DFWG Chair

The problem of interoperability of neuroimaging data analysis software is being addressed by the NIfTI (Neuroimaging Informatics Technology Initiative) task force. The Data Format Working Group (DFWG) within NIfTI was chartered to deal with the particular issue of data interoperability; that is, making it simpler to interchange image (etc.) data between analysis packages.

The DFWG decided to proceed in two phases: near- and long-terms. Since the ANALYZE™-7.5 file format (.hdr/.img file pairs) is simple, widely used, and contains some unused/little-used fields, for the first phase we decided to modify this format to add features that the DFWG agreed were highly desirable for FMRI analysis purposes. In outline, new features include:
(1) Two affine coordinate definitions relating voxel index to spatial location:
� → one orthogonal transform, to indicate orientation/location of data in scanner coordinates;
� → one general affine transform, to indicate mapping to a "normalized" space;
� → global coordinates are +x=Right, +y=Anterior, +z=Superior.
(2) Codes to indicate spatial normalization type (e.g., MNI-152).
(3) Codes to indicate units of spatial and temporal dimensions.
(4) Codes to indicate temporal and spatial acquisition ordering for MRI.
(5) "Complete" set of 8..128 bit data type codes.
(6) Standardized way to store vector-valued datasets (e.g., a matrix at each voxel).
(7) Affine data scaling (i.e., true-data-value≡α*data-in-file+β; α, β stored in header).
(8) Codes and parameters for data "meaning" (e.g., values are t-statistics with 23.7 DOF):
� → 21 types of statistics are encodable; parameters can be global or voxel-dependent.
(9) "Magic" string indicates if header is NIfTI-1 compliant:
� → i.e., contains "extra" data not defined in ANALYZE™-7.5.
(10) Single or dual file storage (.nii or .hdr/.img).

The NIfTI-1 format has been carefully designed so that unmodified FMRI analysis programs that read ANALYZE™-7.5 files are likely to work with NIfTI-1 files; for example, the NIfTI-1 header is still 348 bytes long. Another goal was to reduce the uncertainty about diverse ANALYZE™-7.5 extensions used in this community. The developers of Brain Voyager, FSL, SPM, and AFNI have agreed to adapt their codes to read and write NIfTI-1 files; other sites (including those represented by the DFWG committee) have also agreed to support this format.

The complete definition of the NIfTI-1 format (as a heavily-commented C .h file) is available at the Web site nifti.nimh.nih.gov. In addition, sample C source code for reading and writing NIfTI-1 files is available at this site; Matlab code is planned.

The primary goal of NIfTI-1 is to foster near-term data interchangeability between FMRI analysis and visualization tools. The format is simple, and programs already using the ANALYZE™-7.5 format should be easily adapted to use NIfTI-1. We are aware that many desiderata are missing from this format; work is underway to define and agree upon a more ambitious and extensible NIfTI-2 format.

Acknowledgments: NIfTI and the DFWG are sponsored by NINDS and NIMH. Mark Jenkinson of the FMRIB aided significantly in formulating some aspects of NIfTI-1. The Mayo Foundation granted permission to use components of the ANALYZE™-7.5 header dbh.h.

Created by Alex Clark
Last modified 2005-04-14 18:47
 

Powered by Plone

This site conforms to the following standards: