NCIGT Project Week Fast Imaging Library

© NIH National Center for Image-Guided Therapy

NCIGT Project Week

Fast Imaging Library

V E R I TAS

W. Scott Hoge and Bruno Madore


Overview

• Fast Imaging: reconstruct MR images from limited data• MR applications limited by time

– Reduce data aquired == reduce acquisition time– Can be used to increase spatial resolution, temporal resolution,

or both.

• Aims to ‘fill the gaps’ left by scanner manufacturers• Library currently supports:

– Parallel MR Imaging (SENSE, GRAPPA, …)– Temporal Processing (UNFOLD)– plus additional utility functions (Partial Fourier, DICOM, …)

• Linux, Windows, and Mac distributions available• Link from stand-alone C/C++ applications or call from

Matlab scripts


Library Functions: Parallel Imaging

• pMRI: Parallel MR Imaging employs spatial encoding to complement Fourier encoding and reduce image data requirements.

• SENSE and GRAPPA are widely known variants

• Library supports:– Variable-density sampling– Sophisticated regularization– 2D and 3D acquisitions


Library Functions: UNFOLD

• Temporal encoding can “tag” acquisition artifacts• One can then suppress the artifacts by applying

temporal filters to an image series.

• Library provides both time domain and frequency domain Fermi filters for UNFOLD


Library Functions: Utility Routines

• Partial Fourier reconstruction (homodyne)• Basic building blocks

– FFT, SVD, Conjugate Gradient, LSQR• Magnetic field corrections (GradWarp)• Basic DICOM reader/editor/writer


Applications

• A wide variety of applications have been supported to date:– Accelerated Cardiac Perfusion CINE– Real-time cardiac imaging– Accelerated temperature monitoring– Real-time self-calibrated EPI– Accelerated PROPELLER EPI for

perfusion imaging


App: Cardiac Perfusion

• Utilizing library building blocks, one can mix-and-match algorithms for best image quality

UNFOLD+ SPACE RIP (lo)+ SPACE RIP (mid)

UNFOLD+ GRAPPA (lo)+ GRAPPA (mid)

UNFOLD+ GRAPPA (lo)+ SPACE RIP (mid)

3.5x acceleration, 8 channel cardiac coil, 1.5T, SSFP sequence

Credit: B. Madore


App: Real-time Cardiac Tracking

• 3.2x acceleration, 128x128, 8 channel coil, SSFP

• Completely self-referenced

• 6 images/sec with 1 frame of latency

Credit: E. Samset, R. Chu


App: Temperature Monitoring

• 8x acceleration using 2DRF, SENSE+UNFOLD

8 times 8 times accelerationacceleration

Heating Result (20w for 60sec)

1/81/8FOVFOV

Before SENSE+UNFOLD

After enhanced reconstruction

Reconstructed phase image

Gel Phantom

Transducer

Water tank

Phantom after 2DRF Excitation

Reduced FOVReduced FOV

Pha

se d

irect

ion

Focus of Heating

2DRF Excitation Profile4.3 msec RF pulse

Credit: C-S Mei, J. Yuan, N. McDannold, L. Panych, B. Madore


App: Real-time self-cal EPI

• Phase-labeling w/ UNFOLD for Nyquist ghost correction

• 3.2x acceleration, 128x128, SS-GE-EPI, 1.5T

• Completely self-referenced with low latency

1x: 3.2x:

Credit: H.Tan, B. Kraft (WFU)

note ReducedDistortion


App: PROPELLER EPI for Perfusion

2007: before collaboration

No parallel imaging

Blade dimension = 32 x 128Number of blades = 64

2009: with Fast Imaging Library

Acceleration factor = 2x, GRAPPA

Blade dimension = 96 x 128Number of blades = 64with distortion correction.Credit: H.Tan, B. Kraft (WFU)


What’s in the Package?

• Linkable binary library file• Documentation (via Doxygen)

– Quick Start Guide– Function Reference Manual

• Demos– Matlab MEX functions w/ test data– C/C++ demos: FFT, data I/O, DICOM tag

editing


Quick Setup Example

• Download from

http://www.ncigt.org/pages/Downloads

“Imaging Toolkit”• Unzip into working directory• Start Matlab• Build MEX files• Run Matlab demo scripts


Acknowledgements

• Funding provided by NCIGTNIH U41 RR019703-01A2 (PI:Jolesz)

• For more information:[email protected]

[email protected]

Documents

NCIGT Project Week Fast Imaging Library