A technique is described for performing frequency-selective signal suppression with a high degree of tolerance to RF field inhomogeneity. The method is called B 1 -insensitive train to obliterate signal (BISTRO). BISTRO consists of multiple amplitude-and frequency-modulated (FM) pulses interleaved with spoiler gradients. BISTRO was developed for the purpose of accomplishing bandselective signal removal, as in water suppression and outer-volume suppression (OVS), in applications requiring the use of an inhomogeneous RF transmitter, such as a surface coil. In the present work, Bloch simulations were used to illustrate the principles and theoretical performance of BISTRO. Its performance for OVS was evaluated experimentally using MRI and spectroscopic imaging of phantoms and in vivo animal and human brain. By using FM pulses featuring offset-independent adiabaticity, BISTRO permitted high-quality, broadband suppression with one (or two) discrete borders demarcating the edge ( Key words: adiabatic pulse; pulse sequence; volume localization; outer-volume suppression; frequency modulation A technique to accomplish band-selective signal suppression is a common need in many types of NMR experiments. Examples of such application include chemicalshift selective suppression of water or fat resonances and outer-volume suppression (OVS) in localized spectroscopy and imaging. In the most common approach, signal suppression is achieved by exploiting one or more frequency-selective pulses that convert the initial magnetization M 0 into transverse magnetization M xy that is subsequently dephased by pulsed B 0 gradients (1-11). In OVS techniques, M xy is dephased after being excited in slabs at selected locations and orientations. RF pulses used to excite the unwanted magnetization are commonly the amplitude-modulated (AM) variety. Such OVS approaches generally provide excellent signal suppression, provided the RF field is relatively homogeneous so that each position in the sample experiences a flip angle of 90°. When the flip angle deviates from 90°, some longitudinal magnetization M z remains, giving incomplete signal suppression. The projection presaturation (PP) method of and related techniques (7,8) solve this problem to a certain degree by employing a series of identical, small flip-angle pulses. For example, by repeating the small flip 32 times, the PP technique can offer high-quality suppression (M z / M 0 Ͻ 0.01) on resonance despite an approximately fivefold variation of RF amplitude. Although this degree of compensation for RF inhomogeneity is more than sufficient for experiments performed with volume coils, it is inadequate for RF transmission with a surface coil, which typically subjects the sample to a Ͼ10-fold variation in RF amplitude. Furthermore, because the frequency response profile produced by AM pulses is highly dependent on flip angle, RF transmission with a surface coil can lead to out-of-slice signal excitation, and hence to an undesirable reduction of M 0 in the tissue region(s) to be detected.To overco...