Reproducibility of fat area measurements in young, non-obese subjects by computerized analysis of magnetic resonance images

Abstract

OBJECTIVES: To assess reproducibility, expressed as both inter-observer variability and intra-observer variability, of fat area measurements on images obtained by magnetic resonance (MR); to compare variability between fat area measurements, calculated from a single image per body region and from the average fat area of three images, and to determine reproducibility of image acquisition at the abdominal level.

SUBJECTS: Thirty young, non-obese subjects (reproducibility of image analysis) and nine young, non-obese subjects (reproducibility of image acquisition).

METHODS: Three MR images at the level of the abdomen (in 30 subjects) and at the level of the hip and thigh (in 14 of them). Quantification of subcutaneous fat depots (abdomen, hip and thigh) and visceral fat depots using an image-analyzing computer program. Assessment of variability of image analysis for fat area measurements between two observers and within observers. Assessment of reproducibility of image acquisition at the abdominal level (in nine subjects).

RESULTS: Subcutaneous fat areas in all body regions were quantified with coefficients of variation (CV) ranging from only 2.1%–4.9%. By contrast, visceral fat area measurements showed markedly higher CVs (range: 9.4%–17.6%). Moreover, relative variability was much larger in small visceral fat areas (CVs up to 25.6%). The majority of CVs, calculated for intra-observer variability and calculated from the average fat area measurements of three images, was lower than calculated for inter-observer variability and for one single image, respectively. In particular, for the visceral fat depot, this reduction in variability had practical consequences for the number of subjects required for a study. Variation of repeated image acquisition was in the same range as variation of repeated measurements on the same image.

CONCLUSION: One image per body site is sufficient to obtain a reliable estimate of subcutaneous fat depots. For estimations of the visceral fat depot, the average area measurements of three images reduces variability and increases statistical power. The availability of one single experienced observer during a study adds to accuracy.