Research project

Real-Data Reconstruction to Remove Rician Bias from Diffusional Kurtosis Imaging of the Prostate

Programme
STP
Specialty
Imaging Non-Ionising Radiation
Author
Rosie Goodburn
Training location
Cambridge University Hospitals NHS Foundation Trust

Purpose: To compare prostate diffusional kurtosis imaging (DKI) metrics estimated from phase-corrected real data with those estimated from magnitude data with and without noise compensation (NC). Methods: Diffusion-weighted MR images were acquired at 3T in 16 prostate cancer patients, measuring 6 b-values (0–1500 s/mm2), each acquired with 6 signal averages and along 3 diffusion directions. To allow NC, additional noise-only images were collected by removing radiofrequency pulses. In addition to the conventional magnitude-averaged images, the reconstruction software was adapted to remove large-scale phase variations with a low-pass filter during Homodyne reconstruction, and allowing phase-corrected real data to be averaged. Pixelwise maps of apparent diffusion (D) and apparent kurtosis (K) were calculated for phase-corrected real data, and for magnitude data both with and without noise compensation. Mean values for these three cases were compared in tumour (TUM), normal transition zone (NTZ) and normal peripheral zone (NPZ). Results: Relative to metrics estimated from magnitude data without NC, median NC K were significantly (P<0.001) lower by 4/6/6% in tumor/NPZ/NTZ, while real-data K were significantly (P<0.001) lower by 16/19/13%, where the differences between NC K and real-data K were also significant (P<0.001). Conclusion: Compared to magnitude data with NC, real data produces significantly lower kurtosis values in prostate for both normal tissue and tumor.

Last updated on 10th September 2020