MRI images are normally analyzed using one of two methods: 1) the perimeter of the tissue of interest is traced manually (6, 7), and the area within the perimeter is calculated by multiplying the number of pixels in the region of interest by their known area, or 2) image segmentation algorithms are used to identify all pixels within a selected range of intensities believed to be representative of a specific tissue. However, the latter approach is considered more problematic when applied to MRI images for three reasons: 1) distributions of pixel intensity (greyscale) values for different tissues overlap more for MRI than for CT images, 2) noise due to respiratory motion blurs the borders between tissues in the abdomen to a greater extent in MRI than in CT, and 3) inhomogeneity in the magnetic field can produce shading at the peripheries of MRI images (6).
With multiple MRI images, tissue volumes can be calculated by integrating cross-sectional area data from consecutive images. Because of the cost of image acquisition and analysis, images are typically collected with gaps between images (usually ranging from 20 to 40 mm), and volumes are then calculated using various modeling equations (3, 4, 8, 9). Tissue densities for adipose tissue, skeletal muscle, and organs are fairly constant from person to person, and volume measures for these tissues can be converted to mass units by multiplying the volume by assumed tissue density values (10, 11).