PURR-TURBO: A Novel Pulse Sequence For Relaxographic Imaging

Introduction Recently, the concept of reluxu~ra~hic imaging (RI) that is based on the utilization of the relaxogram the distribution of the relaxation times has been introduced.’ This work revealed another fundamental dimension of NMR, which has potential in many various applications. It is believed that RI is particularly useful for the discrimination of isochronous signals with more than one relaxation time. In this report, we propose a new pulse sequence for RI that is a hybrid version of the “single-shot” IR snapshot-FLASH’ and the “oneshot” IR snapshot-FLASH sequences 1,3-5 These methods are also known as “multi-point” IR techniques. The new sequence proposed here collects multiple k-space rows in each inversion recovery time (TI) point during magnetization recovery rather than the entire data set (as in a “singleshot’ method) or just a single row (as in a “one-shot” method). Thus, it further reduces the scanning time without significantly giving up the advantage of the one-shot IR snapshot-FLASH methods. It also can afford the attainment of a higher spatial resolution in a reasonable measuring time. Furthermore, this method does not need a highperformance gradient amplifier. Methods The new sequence is a considerably speeded-up version of the original PURR (progressively unsaturated relaxation during perturbed recovery from inversion)’ pulse sequence (Figure, top), and we have termed this new sequence PURR-TURBO (Figure 1, bottom). The Figure illustrates the difference in these two sequences. In the original PURR, one k-space line (echo: index n) is acquired at each TI value (tk) following an inversion pulse and a spoiled gradient pulse. Unlike the original PURR, PURR-TURBO collects four (Figure) or six k-space lines (using an alternating center out scheme) at each TI point. This essentially decreases the acquisition time by almost 4 or 6 fold. However, increasing the number of the k-space lines per acquisition to more than 6 would begin to risk losing the capability of faithfully detecting short T, components, as the sequence becomes more like the original “singleshort” IR snapshot-FLASH sequence introduced by Haase.’ Two phantoms were used to test the sequence: a spherical phantom containing a 2 L solution of 2% agar, 75 mM NaCI, and 0.5 mM NaN,; and a spherical phantom containing a 2 L solution of 0.25 mM MnC& and 70 mM NaCl. Also seven normal human volunteers participated in this study after giving their informed consent, and were compensated for their time. All NMR measurements were performed with a 4T VarianESiemens whole body instrument equipped with a whole-body actively shield gradient using a quadrature circumscribing head volume coil. All studies were performed with the original PURR and the new segmented PURRTURBO sequences illustrated in the Figure. The typical data acquisition parameters were: 256 x 128 (or 192) data matrix, FOV = 24 x 18 cm’, TE was about 5 ms, read pulse flip angle was 5”. Thirty-two TI values varied from 38 ms up to at least 3 Ti with spacing increasing geometrically. All raw data were first DC corrected, Fourier transformed and the resulting complex data were phased in reference to the longest TI image by using Interactive Data Language (IDL; Research Systems, Inc., Boulder, CO) software. Then, recovery data were analyzed with a three-parameter non-linear fit assuming a single Ti value. Results and Discussion Regardless the inherent inhomogeneous Bi at 4T due to mostly the dielectric resonance, the Ti distributions throughout the images of the phantoms was very uniform and was very well fit with a Gaussian function. This suggests that the error in Ti measurement due to B1 inhomogenity is negligible at the very small read pulse flip angle used here. The Ti values measured from PURR and PURR-TURBO sequences were in very good agreement with a correlation coefficient of better than 0.98 using all imaging pixels. However, the correlation coefficient for human brain images between these two sequences is about 0.96. The smaller correlation found for in viva data could be because of subject movements and/or physiological noise differences between the two MR examinations. In general, the Ti value obtained from PURR-TURBO is slightly smaller than that from the original PURR. This may be because that the magnetization is slightly more saturated by the former than the latter sequence. No artifacts were observed in the images indicating no presence of coherent magnetization from previous scans.h The major improvement in the new sequence is that it collects every 4 (or 6) k-space lines, instead of just one as in the original PURR, after a nonselective adiabatic inversion pulse and a spoiled gradient. The acquisition scheme is a compromise between Haase’s’ and Labadie’s’ methods. The former provides fewer TI points and a shorter scanning time but longer sampling times; while the latter provides more TI points and a longer scanning time but very small sampling times. The new technique preserves essentially all of PURR’s advantages. In addition, it provides about 4 or 6 times more efficient in scanning time. This technique provides a tool for studying situations with multi-, exponential relaxation times with a very good temporal resolution. Another important feature is that one can obtain a relaxogram or a T, map with an in-plane resolution of less than (1 mm)* in a few minutes without requiring high-performance gradient amplifiers. [This work was supported by DOE (DE-AC02-98CH10884) and NIH (ROl GM32125).] References 1. Labadie C, Lee J-H, Vetek G, and Springer CS, J. Magn. Reson. B, 105,99-112 (1994). 2. Haase A, Mugn. Reson. Med. 13,77-89 (1990). 3. Kay I and Henkelman RM, Mugn. Reson. Med. 22,414-424 (1991). 4. Deichmann R and Haase A, J. Mugn. Reson. 96, 608-612 (1992). 5. Gowland PA and Leach MO, Magn. Reson. Med. 26,79-88 (1992). 6. Sammi MK, Felder CA, Fowler JS, Lee J-H, Levy AV, Li X, Logan J, Palyka I, Rooney WD, Volkow ND, Wang G-J, and Springer CS, PISMRM, submitted.