Magnetic Resonance Image Reconstruction from Incomplete K-Space Data


In magnetic resonance imaging (MRI), a technique called partial fourier (PF) can be used to reduce the acquisition time by half. Only half of the data is needed to reconstruct the full image using the conjugate symmetric property of Fourier transform.

  • Magnetic Resonance Image Reconstruction from Incomplete K-Space Data 0
  • Magnetic Resonance Image Reconstruction from Incomplete K-Space Data 1
Commercialisation opportunities
Problem addressed

MRI scanning is time-consuming and radiologists are limited. Each day, only a handful of MRI scans can be completed with a machine. With this technology, the scanning time can be effectively halved while still capturing enough data for image reconstruction with low error rate, doubling number of scans per day.

  • The proposed CNN model shows noticably higher signal-to-noise ratio than conventionaly POCS method.
  • The error rate remains very low even at 0.51 fractions of sampling, while traditional methods render unusable images when the fraction is below 0.6.
Key impact
  • Double number of scans from an MRI machine per day
  • By halving the time taken for each scan, the cost per each MRI scan can be significantly discounted
  • MRI Scanning


  • PRC Application No. 202010810587.8
  • US Application No. 17/797,119
The University of Hong Kong (HKU)

Founded in 1911, The University of Hong Kong (HKU) is the first and oldest institution of higher education in Hong Kong. For over a century, the University has dedicated itself to creating knowledge, providing education, and serving society. Today, HKU has an established worldwide reputation for being a research-led comprehensive University with ten Faculties and a strong commitment to intellectual freedom, liberty and diversity. HKU has a proud record of academic recognition in researches through honours and awards received from both local and international bodies.