Abstract

Recently W. T. Dixon (Radiology 153, 189 (1984))introduced a simple method of proton chemical-shift imaging which requires only two images, a conventional (in-phase) image and an image in which fat and water protons are 180 degrees out of phase during signal acquisition, to separate the signals from fat and water protons. We have tested the application of this method to the quantitative determination of fat content and fat and water longitudinal relaxation times, and analyzed the effects of random and systematic errors. Ten phantoms were constructed with a range of fat contents (0-50% by weight) and water T1's (300-800 ms). Fat and water T1's were measured with a 0.6-T clinical imaging system in two ways: using the system as a spectrometer with all gradients off, and from least-squares fits to in-phase and out-of-phase image data made with six values of TR. The image-derived values of water T1 agreed well with spectrometer-derived values (r = 0.97) and the image derived fat fraction correlated strongly with the fat fraction by weight (r = 0.995). The effects of random and systematic errors were analyzed for a minimum data set of four images: in-phase and out-of-phase images at two values of TR. The pair of TR values which minimize the variance in water T1 were calculated, and for these pulse sequences the effects of two potential systematic errors were calculated: inhomogeneities in the main field, which will reduce the intensity in out-of-phase images compared to in-phase images even for pure water samples, and an incorrect shift of the 180 degrees pulse in the out-of-phase pulse sequence, corresponding to an inaccurate assumed chemical shift. With careful attention to such systematic effects the Dixon method is capable of producing reliable quantitative measurements.