SMS-HSL: Simultaneous multislice aliasing separation exploiting hankel subspace learning

Suhyung Park, Jaeseok Park

Research output: Contribution to journalArticlepeer-review

7 Citations (Scopus)


Purpose: To develop a novel, simultaneous multislice reconstruction method that exploits Hankel subspace learning (SMS-HSL) for aliasing separation in the slice direction. Methods: An SMS signal model with the Hankel-structured matrix was proposed. To efficiently suppress interslice leakage artifacts from a signal subspace perspective, a null space was learned from the reference data combined over all slices other than a slice of interest using singular value decomposition. Given the fact that the Hankel-structured matrix is rank-deficient while the magnitude between the reference and its estimate is similar in k-space, the SMS-HSL was reformulated as a constrained optimization problem with both low-rank and magnitude priors. SMS signals were projected onto a slice-specific subspace while undesired signals were eliminated using the null space operator. The simulations and experiments were performed with increasing multiband factors up to 6 using the SMS-HSL and the split slice-GRAPPA. Results: Compared with the split slice-GRAPPA, the SMS-HSL shows superior performance in suppressing aliasing artifacts and noises at high multiband factors even with: insufficient reference signals, a small number of coils, and a short distance between aliasing voxels. Conclusion: We successfully demonstrated the effectiveness of the SMS-HSL over the split slice-GRAPPA for aliasing separation in the slice direction. Magn Reson Med 78:1392–1404, 2017.

Original languageEnglish
Pages (from-to)1392-1404
Number of pages13
JournalMagnetic Resonance in Medicine
Issue number4
Publication statusPublished - 2017 Oct


  • Hankel matrix
  • low rank
  • magnetic resonance imaging
  • parallel imaging
  • simultaneous multislice

ASJC Scopus subject areas

  • Radiology Nuclear Medicine and imaging


Dive into the research topics of 'SMS-HSL: Simultaneous multislice aliasing separation exploiting hankel subspace learning'. Together they form a unique fingerprint.

Cite this