Whole-body tomographic imaging by NMR

Abstract

Since producing our first whole-body nuclear magnetic resonance (NMR) line-scan images some two years ago (Mansfield et al., 1978), we have made a number of changes to our imaging machine which have made possible a considerable improvement in picture quality together with a substantial reduction in imaging time. Other approaches to NMR imaging are extensively discussed and reviewed elsewhere (Mansfield, 1976; Brunner and Ernst, 1979; Mansfield and Morris, 1981). The present technique utilizes a selective radio-frequency pulse and switched magnetic field selection gradient (Mansfield et al., 1976) used initially to define a thin slice through the specimen. The excited plane of nuclear spins arising from the distribution of protons in the tissues produces a transient signal in a second “read” gradient. This signal is Fourier transformed to give the projection of the total spin distribution in a direction orthogonal to the read gradient. Successive projections produced by mechanically rotating the read gradient are subsequently formed into a two-dimensional picture array using a filtered back projection algorithm (Brooks and di Chiro, 1976). This technique is the NMR analogue of the successful X-ray tomographic technique developed by Hounsfield (1973). In the present images, the slice thickness was set to a nominal 2.0 cm, but it can be varied as required. The imaging time was 9.5 minutes with a pulse repetition period of 0.32 seconds and an average of 16 shots per projection. The NMR operating frequency was 4.0 MHz corresponding to a static magnetic field of 950 G.