Patients were studied using a 2-tesla whole-body magnet (Oxford Magnet Technology, Oxford, United Kingdom), which was interfaced to a Bruker Avance spectrometer (Bruker Medical GmbH, Ettlingen, Germany). Patients were examined at rest and lay prone on a custom-designed bed incorporating a gantry that allowed a range of surface coils to be positioned beneath the precordium without moving the subject. Patients were positioned with their heart at the isocenter of the magnet, and this was confirmed using standard multislice spin-echo proton images acquired using a double-rectangular surface coil placed around the chest (relaxation time [TR] = heart rate, echo time [TE] = 25 ms, slice = 10 mm, 5 slices, 15-mm spacing). These images were also used to verify that the diaphragm would not be within the field of view during the 31P acquisition. The imaging coil was exchanged for a circular proton surface coil (15-cm diameter) placed beneath the chest, and shimming was performed to optimize the magnetic field homogeneity over the heart. Finally, the coil was exchanged for a 31P surface coil (8-cm diameter). To ensure full localization of the 31P-MRS data to the heart, spectra were acquired using a slice-selective, one-dimensional chemical shift imaging (1D-CSI) sequence, including spatial presaturation of lateral skeletal muscle, as follows. An oblique-angled saturation slab was positioned from the proton images across any skeletal muscle lateral to the surface coil, and all 31P signals were saturated using two pairs of sinc pulses at a 1:2 flip/angle ratio (13). An 8-cm thick transverse slice was then excited, followed by one-dimensional phase encoding into the chest to subdivide signals into 64 coronal layers, each 1-cm thick (TR = heart rate, 16 averages). All proton and 31P data at acquisition were cardiac gated using a pulse oximeter probe placed on the subject’s finger.