Fat Quantification Using Multiecho Sequences with Bipolar Gradients: Investigation of Accuracy and Noise Performance Pernilla Peterson* and Sven Ma ˚nsson Purpose: To investigate the accuracy and noise performance of fat quantification with multiple gradient-echo images acquired using bipolar read-out gradients and compare them with those of the well-established unipolar technique. Theory: The bipolar read-out technique induces phase and amplitude errors caused by gradient delays, eddy currents, and frequency-dependent coil sensitivity. In this study, these errors were corrected for jointly with the fat/water separation by modeling the impact of these effects on the signal. This approach did not require acquisition of reference data or mod- ification of the pulse sequence. Methods: Simulations and a phantom experiment were used to investigate the accuracy and noise performance of the technique and compare them with those of a well-established technique using unipolar read-out gradients. Also, the in vivo feasibility was demonstrated for abdominal applications. Results: The phantom experiment demonstrated similar accu- racy of the bipolar and unipolar fat quantification techniques. In addition, the noise performance was shown not to be affected by the added estimations of the phase and amplitude errors for most inter-echo times. Conclusion: The bipolar technique was found to provide accurate fat quantification with noise performance similar to the unipolar technique given an appropriate choice of inter- echo time. Magn Reson Med 71:219–229, 2014. V C 2013 Wiley Periodicals, Inc. Key words: fat quantification; magnetic resonance imaging; multiecho; bipolar gradient; eddy currents; chemical shift Magnetic resonance imaging-based fat/water separation and fat quantification noninvasively provides quantita- tive and spatial information on fat accumulation in the human body, meeting the critical need for diagnostic tools in obesity and diabetes (1). Chemical shift techni- ques, in particular, have gained much interest in this field and use multiple gradient-echo images with vari- ous echo times (TEs) and an iterative least-squares esti- mation of the fat and water signal components as well as correction for magnetic field inhomogeneities and T  2 relaxation (2,3). Starting from a dual-echo technique introduced by Dixon in 1984 (4), recent research on chemical shift techniques has resulted in a number of methods that benefit from a multiecho acquisition (3,5– 8). This raises a demand for a time- and noise-efficient acquisition strategy, especially at 3 T or 7 T and for breath-hold applications (9,10). The multiple gradient echoes for fat/water separation can all be acquired during a single repetition time (TR) with either unipolar or bipolar read-out gradients. Uni- polar sequences have been used extensively for both fat/ water separation and quantification (3,11–13), and pauses data acquisition during fly-back read-out gra- dients. Thus, all echoes are collected with the same gra- dient polarity. With a bipolar approach, echoes are acquired during both positive and negative polarities, causing every other echo to have opposed read-out gradi- ent polarity. The latter is associated with a number of advantages and disadvantages compared with the unipo- lar approach. The obvious advantage of a bipolar acquisition is the possibility to reduce the inter-echo time (DTE) and thus also the TR and the total scan time. A shorter scan time reduces motion artifacts and sensitivity to T  2 relaxation and makes the method more compatible with breath- hold imaging. Alternatively, the shorter DTE may be used to acquire a larger number of echoes in the same scan time. Or, DTE may be kept constant and the read- out bandwidth may be decreased. Importantly, a reduced DTE also increases the range of off-resonance frequencies that can be resolved unambiguously in the iterative least-squares estimation (9,10) and allows closer sam- pling of the fat/water phase evolution. These points are particularly important at 3 T or 7 T where the off-reso- nance frequencies are expected to be larger and where the fat phase evolves faster than at 1.5 T. However, the bipolar acquisition also causes a few problems. First, the chemical shift of the fat signal appears in opposite directions in even and odd echoes. This may be solved through the use of a k-space-based fat/water separation or with the use of a high read-out bandwidth (14,15). Second, field inhomogeneities cause spatial distortions in opposite directions in every other echo. Such distortions may be corrected using the esti- mated field map or, in cases of typical shimming condi- tions, be ignored (15). Third, gradient delays and eddy currents cause phase errors in all spatial directions that are mainly linear, but also include higher order terms (16,17). These errors exist, of course, also in unipolarly acquired echoes, but are in the bipolar case in opposite directions in even and odd echoes, causing serious Department of Medical Radiation Physics, Lund University, Ska ˚ ne Univer- sity Hospital, Malm€ o, Sweden. Grant sponsors: Magnus Bergvalls stiftelse, Direkt€ or Albert Pa ˚ hlssons stiftelse. *Correspondence to: Pernilla Peterson, M.Sc., Department of Medical Radi- ation Physics, Inga Marie Nilssons gata 49, Ska ˚ ne University Hospital, SE- 205 02 Malm€ o, Sweden. E-mail: pernilla.peterson@med.lu.se Received 5 October 2012; revised 20 December 2012; accepted 3 January 2013 DOI 10.1002/mrm.24657 Published online 14 February 2013 in Wiley Online Library (wileyonlinelibrary.com). Magnetic Resonance in Medicine 71:219–229 (2014) V C 2013 Wiley Periodicals, Inc. 219