How to check image stability

Indications:

This is a basic quality assurance process for MR imaging. It was developed as part of our magnet purchase specifications. It would be best to see this run on a regular basis - weekly, for example.

Description

Using a very simple data collection protocol, we analyze the signal stability of a time series of images, to show the magnitudes of the signal fluctuations. The data collection parameters were designed such that this simple analysis should unmask instability:

  1. Transmitter power
  2. ADC linearity
  3. Gradient amplitude
  4. B0 (magnetic field fluctuation)
  5. Demodulation frequency / frequency synthesizer

Or other factors that could influence it.

Stability is measured on the average intensity of an ROI located near the center of the image. By working with the average, P2P becomes sensitive to the instrument instability, rather than to the scanner or phantome signal to noise ratio. The user can control the ROI size and location.

The software studies the images for systematic changes in signal intensity that may or may not be position dependent. It separately analyzes high frequency (25 second) and low frequency image intensity changes.

What to do:

Collect a series of images according to the following protocol:
General purpose phantom (e.g. w/resolution features) Single shot (EPI) acquisition
2 x 2 x 8 mm voxels (conjugate synthesis OK)* TE 45 msec
128 x 128 matrix (25 cm FOV) TR 2.5 seconds
Gradient Echo (no 180 pulse) 21 slice locations
Head receiver coil 240 serially collected images
45 degree flip angle
*or equivalent 32 mm3 voxel volume

Save the images and then analyze them using P2P (whose source can be found from our Central Software Download site)

To use P2P, type:

> P2P -i infile.ext -o outfile.ext [options]

where infile.ext is the raw image file in any of our standard formats.

P2P accepts the following options:

-N Noise threshold below which pixels will not be analyzed. Default is for P2P to determine this automatically.
-s Number of images to skip at the beginning of the file
-X, -Y Horizontal and vertical offset, in pixels, of the upper left corner of the ROI from the center of the image. Default is -4. Positive values move the ROI down and to the right.
-S Linear dimension, in pixels of the ROI (the ROI is always square)
Default is 8
-r Present data in native units. If not selected, the data will be expressed in percent.
-H Highlight factor. P2P displays an image with the ROI highlighted. -H controls the ratio of the highlighted region to the original image intensity. Default is 1.25

The program creates an output file of six images. Please note that the images shown on this web page are of a human head - not a typical set for stability assessment.

Image of the peak to peak intensity variations. Image of the Standard deviation of the image intensity.
The maximum signal intensity at each location. The minimum signal intensity at each location.
The average signal intensity at each location. An image with the ROI (subtly) highlighted.

Running with the following input:

> P2P -i SIEMENS_B0_CORRECTED_EPI_1.img -o StabCheck.img

P2P produces, as output, a display similar to this:


P2P RESULTS SUMMARY:
Col 1: slice number     Col 2: peak to peak change      Col 4: Mean over time   Col 5: Standard deviation over time
Col 6: squared correlation between this slice and the denoised average of all slices    Col 7: Col5/Col6
Col 8: Slope (b) of the regression over time +/- 95% confidence interval        Col 9: Probability of non-zero slope

Sl:  1   P-P:   3.026 %   Avg:  279.11   sd: 0.5467 %   r^2: 0.846      sd/r^2: 0.646   b: -0.00030 % +/- 0.001012      p < 0.55837
Sl:  2   P-P:   2.403 %   Avg:  299.00   sd: 0.4765 %   r^2: 0.419      sd/r^2: 1.137   b: -0.00245 % +/- 0.000825      p < 0.00000
Sl:  3   P-P:   2.258 %   Avg:  318.22   sd: 0.4718 %   r^2: 0.329      sd/r^2: 1.432   b: -0.00141 % +/- 0.000855      p < 0.00125
Sl:  4   P-P:   
                                                 ...>
                                                                                        b: 0.00059 % +/- 0.000778       p < 0.13126
Sl: 19   P-P:   1.881 %   Avg:  394.54   sd: 0.3904 %   r^2: 0.708      sd/r^2: 0.551   b: -0.00107 % +/- 0.000710      p < 0.00320
Sl: 20   P-P:   2.178 %   Avg:  370.19   sd: 0.4206 %   r^2: 0.683      sd/r^2: 0.616   b: -0.00109 % +/- 0.000766      p < 0.00516
Sl: 21   P-P:   2.114 %   Avg:  375.46   sd: 0.4263 %   r^2: 0.888      sd/r^2: 0.480   b: 0.00190 % +/- 0.000751       p < 0.00000


        TIME PTS in FILE: ..... 240
        NUMBER SKIPPED: ....... 0
        LOG FILE: ............. StabCheck.P2PLog.txt
The log file contains (for each slice):
        - the average signal in the RoI at each time point and its mean
        - the low pass filtered signal in the ROI and its mean
        - the deviation from the filtered time course, expressed as percent of the mean.

        THRESHOLD: ............ 83.12
        NON-NOISE PTS: ........ 15247

ROI LOCATION:
        left: 28        top: 28
        right: 36       bottom: 36

Image 0: P-P map        Image 1: Std. Deviation Image
Image 2: Maximum Image  Image 3: Minimum Image
Image 4: Average Image  Image 5: ROI-highlighted Image

The log file contains (for each slice):
        - the average signal in the RoI at each time point and their means
        - the low pass filtered signal in the ROI and their mean
        - the deviation from the filtered time course, expressed as percent of the mean.
The log file can be used to create graphs such as these:

Finally, P2P creates a processing record: outfile.P2P.proc.

Back to HowTo pages
This page is maintained by Mark Cohen [updated 2.4.04]