Hc. Whalley et Jm. Wardlaw, Accuracy and reproducibility of simple cross-sectional linear and area measurements of brain structures and their comparison with volume measurements, NEURORADIOL, 43(4), 2001, pp. 263-271
Volumetric measurement of brain structure on brain images is regarded as a
gold standard, yet is very time consuming. We wondered whether simple linea
r and area measurements might be as accurate and reproducible. Two observer
s independently measured the cross-sectional area of the corpus callosum, l
entiform and caudate nuclei, thalamus, amygdalas, hippocampi, lateral and t
hird ventricles, and the width of the sylvian and frontal interhemispheric
fissures and brain stem on brain MRI of 55 patients using a program written
in-house; one observer also measured the volumes of the basal ganglia, amy
gdalo-hippocampal complex and ventricular system using Analyze, and perform
ed qualitative assessment of four regions (lateral and third ventricles, co
rtex, and medial temporal lobe) using the Lieberman score. All measures wer
e performed blinded to all other information. Test objects of known size we
re also imaged with MRI and measured by the two observers using the in-hous
e program. The true sizes of the test objects were measured using engineeri
ng calipers by two observers blind to the MRI results. Differences between
the two observers using the same measurement method, and one observer using
different methods, were calculated. The simple linear and cross-sectional
area measurements were rapid (20 min versus 5 h for volumetric); were highl
y accurate for test-object measurement versus true size; had excellent intr
aobserver reliability; and, for most brain structures, the simple measures
correlated highly significantly with volumetric measures. The simple measur
es were in general highly reproducible, the difference (as a percentage of
the area or width of a region) between the two raters being around 10%, ran
ge 0.1%-14.1%, (similar to inter-rater variability in previous studies of v
olume measurements). The simple linear and area measures are reproducible a
nd correlate well with the measured volumes, and there is a considerable ti
me saving with the former. In circumstances where a large volume of work pr
ecludes detailed volume measurement, simple methods are reliable and can be
used instead.