Noise reduction in MR angiography with nonlinear anisotropic filtering

Purpose To evaluate three‐dimensional nonlinear anisotropic filtering in suppressing image noise in high spatial resolution magnetic resonance angiograms (MRA) acquired with hybrid undersampled projection reconstruction and phase contrast vastly undersampled isotropic projection reconstruction (PC‐VIPR). Materials and Methods Three‐dimensional nonlinear anisotropic filtering was quantitatively analyzed and evaluated through the measurement of contrast to noise ratio (CNR) in PC‐VIPR images and contrast enhanced peripheral MRA images. To filter MRA images with ultra‐high spatial resolution and poor CNR, a spatial frequency dependent nonlinear anisotropic filtering algorithm was proposed that uses two‐step processing to filter the whole spatial frequency data. Results Three‐dimensional nonlinear anisotropic filtering was shown to be effective in suppressing noise and improving CNR in MRA with isotropic spatial resolution. Higher CNR was achieved using spatial frequency dependent nonlinear anisotropic filtering. A typical CNR gain of between 50–100% was shown in our studies. Conclusion Three‐dimensional nonlinear anisotropic filtering significantly improved CNR in MRA images with isotropic spatial resolution. Spatial frequency dependent nonlinear anisotropic filtering further improved CNR for MRA images with ultra‐high spatial resolution and low CNR. J. Magn. Reson. Imaging 2004;19:632–639. © 2004 Wiley‐Liss, Inc.