Enabling in vivo comparisons of different four-dimensional magnetic resonance imaging sequences for radiotherapy guidance using visual biofeedback

Managing respiratory motion is essential for effective radiotherapy in the abdominothoracic regions. Respiratory-correlated four-dimensional magnetic resonance imaging (4D-MRI) can provide accurate motion estimation to help define treatment volumes for adaptive radiotherapy. However, validating and comparing 4D-MRI sequences in vivo is challenging due to the presence of breathing variability. This study combines visual biofeedback (VBF) with 4D-MRI sequences to facilitate in vivo comparisons. Fourteen healthy volunteers and one patient were scanned on a 1.5 T Unity MR-linear accelerator (Elekta AB, Stockholm, Sweden) at two institutions. A radial stack-of-stars (SoS), a simultaneous multi-slice (SMS), and a Cartesian acquisition with spiral ordering (CASPR) 4D-MRI sequence were acquired. These acquisitions were performed without and with VBF based on an interleaved one-dimensional respiratory navigator (1D-RNAV) acquisition. Breathing variability across sequences was quantified using 1D-RNAV-derived breathing waveforms. Reconstructed 4D-MRI data were used to extract the motion amplitude, which was compared intra-volunteer across sequences and to the amplitudes of the breathing waveforms. Breathing variability across sequences decreased by 37% (amplitude, p = 0.039) and 64% (period, p < 0.003), and the median intra-volunteer 4D-MRI-derived motion amplitude agreement improved from 3.5 mm to 1.8 mm (p = 0.064) across sequences due to VBF guidance. Four-dimensional MRI-derived amplitudes were smaller than breathing waveform amplitudes, with median differences of -31% (SoS), -17% (SMS), and -9% (CASPR). The average breathing waveform amplitude was 8% larger than instructed. This methodology enables in vivo comparisons of 4D-MRI sequences for adaptive radiotherapy, with guidance improving anatomical consistency and ensuring more reliable comparisons.