Imaging of the beating heart is one of the main challenges in fast MRI. In order to depict the heart in the different phases of the cardiac cycle and to avoid artefacts due to respiratory motion the data acquisition needs to be fast, yield high signal and provide excellent contrast between blood and myocardium. Balanced steady state free precession (bSSFP) sequences offer all these advantages. Unfortunately, they are sensitive to field inhomogeneities, which lead to banding artifacts, considerably reducing the image quality¹. In 2002, Foxall et al. demonstrated that the steady state of bSSFP tolerates a slow frequency change², thus allowing it to sweep through different frequencies in one acquisition. Such a dynamically frequency-sweeped radial acquisition proved suitable to create banding-free images even in the presence of high field inhomogeneities³. The purpose of this study was to apply a frequency modulated procedure to cardiac cine imaging and to use a multi-frequency reconstruction to obtain banding free images of the cardiac cycle.Radial bSSFP measurements were performed in a healthy volunteer on a 3T MRI system (Siemens MAGNETOM Prisma) using a 30-channel body array coil. The pulse sequence featured a shift in the offset frequency for each projection, covering a total offset frequency range of 360Hz. Other imaging parameters were: TE 1.4ms, TR 2.8ms, flip angle 36°, resolution 2.2x2.2x10mm3, total acquisition time 19s. Readouts of in vivo measurements were retrospectively assigned to 20 cardiac phases using the ECG information recorded during the measurement. All images were reconstructed by gridding a 2D Fourier transform and a square-root of the sum-of-squares combination of all coils. Additionally, all images were reconstructed for 100 different off-resonance frequencies in analogy to EPI corrections⁴ to correct for off-resonance effects in frequency-modulated acquisitions. The procedure first compensates for the linear drift in signal phase caused by the acquisition pattern and leads to a binary distribution of phase with one 180° phase shift per cycle of the off-resonance profile. The algorithm then eliminates this phase jump for different assumed frequencies. A square-root of the sum-of-squares combination of all coils generated an image stack, whose maximum intensity projection provided the final result.Figure 1 shows a short axis view in diastole and systole for standard bSSFP, frequency-modulated bSSFP with simple gridding reconstruction and the proposed frequency-modulated bSSFP with multi-frequency reconstruction. The images show banding artefacts in standard bSSFP images. Frequency-modulation removes bandings but gridding reconstruction is accompanied by signal losses. The proposed multi-frequency reconstruction removed bandings and features high signal intensity.Our study demonstrates the capabilities of a multi-frequency reconstruction. In addition to the possibility of reconstructing banding-free images of the cardiac cycle from one breath-hold CINE acquisition, the multi-frequency reconstruction provides an increased signal level compared to a simple gridding reconstruction. Main advantages of bSSFP, namely acquisition speed and image contrast, were successfully retained by the reconstruction. Future work will involve a combination of frequency-modulated bSSFP with acceleration techniques like parallel imaging or compressed sensing to further reduce streaking artifacts and to improve the spatial resolution.