File(s) not publicly available
Object strength - an accurate measure for small objects that is insensitive to partial volume effects
journal contribution
posted on 2023-06-07, 13:49 authored by P. S. Tofts, N. C. Silver, G. J. Barker, A. GassThere are currently four problems in characterising small nonuniform lesions or other objects in Magnetic Resonance images where partial volume effects are significant. Object size is over- or under-estimated; boundaries are often not reproducible; mean object value cannot be measured; and fuzzy borders cannot be accommodated. A new measure, Object Strength, is proposed. This is the sum of all abnormal intensities, above a uniform background value. For a uniform object, this is simply the product of the increase in intensity and the size of the object. Biologically, this could be at least as relevant as existing measures of size or mean intensity. We hypothesise that Object Strength will perform better than traditional area measurements in characterising small objects. In a pilot study, the reproducibility of object strength measurements was investigated using MR images of small multiple sclerosis (MS) lesions. In addition, accuracy was investigated using artificial lesions of known volume (0.3-6.2 ml) and realistic appearance. Reproducibility approached that of area measurements (in 33/90 lesion reports the difference between repeats was less than for area measurements). Total lesion volume was accurate to 0.2%. In conclusion, Object Strength has potential for improved characterisation of small lesions and objects in imaging and possibly spectroscopy.
History
Publication status
- Published
Journal
Magnetic Resonance Materials in Physics, Biology and MedicineISSN
0968-5243Publisher
Springer VerlagExternal DOI
Issue
3Volume
18Page range
162-169Department affiliated with
- BSMS Publications
Full text available
- No
Peer reviewed?
- Yes
Legacy Posted Date
2007-02-28Usage metrics
Categories
No categories selectedKeywords
Licence
Exports
RefWorks
BibTeX
Ref. manager
Endnote
DataCite
NLM
DC