Conference proceeding
Numerical simulations of spectral phase and spectral distance behavior as scalar descriptors of multispectral data
Proceedings of SPIE, Vol.5370(1), pp.1914-1923
Medical Imaging 2004: Image Processing
05/12/2004
DOI: 10.1117/12.535769
Abstract
Multispectral imaging is the acquisition of images of the same view with
multiple contrast mechanisms. Each voxel is a vector, each element of the
vector being the value at that location for a particular spectral channel. We
refer to this vector space as the feature space. Scalar measures computed
from the feature space have been shown to improve visualization of objects. A
wide variety of choices for the scalar measures are available. We used
simulations to examine two simple scalar measures: spectral distance (SD) and
spectral phase (SP). We assumed that we had two tissue compartments:
background and lesion. The tissue characteristics in each compartment were
described by Gaussian distributions with different means and variance of one.
We used the mean background value as the reference point in the feature
space. We compared SD and SP in terms of dependence on absolute locations of
the spectral channels, correlation between spectral channels, and the number
of spectral channels used (dimension). Each model was characterized by the
L2 and L-infinity norms of the input signal-difference-to-noise ratio
(SDNR) and the resulting SDNR. SP produced peak SDNR values which exceeded SD
values. However, SP had a lower average SDNR value across the examined
parameters. SP demonstrated a complex dependence on the correlation
coefficient. For low SDNR in the input channels, formation of SD and SP
images decreased the SDNR. The minimum additional SDNR depended on both the
dimensionality of the model and the SDNR of the primary channel.
Details
- Title: Subtitle
- Numerical simulations of spectral phase and spectral distance behavior as scalar descriptors of multispectral data
- Creators
- Brian E Chapman - Woman's HospitalClaudia Mello-Thoms - Woman's Hospital
- Resource Type
- Conference proceeding
- Publication Details
- Proceedings of SPIE, Vol.5370(1), pp.1914-1923
- Conference
- Medical Imaging 2004: Image Processing
- DOI
- 10.1117/12.535769
- ISSN
- 0277-786X
- Language
- English
- Date published
- 05/12/2004
- Academic Unit
- Roy J. Carver Department of Biomedical Engineering; Radiology
- Record Identifier
- 9984318695702771
Metrics
6 Record Views