Calculating the BPF

Having outlined the brain in every image slice, you can now compute the brain parenchymal fraction, an index of brain atrophy. The computation involves selection of an intensity threshold that best separates the brain parenchyma from the CSF surrounding the brain and in the ventricles. This is done by examining the pixel intensity histogram, and separating the parenchyma peak from the CSF peak.

Note: if you have previously created the brain outline, and saved the ROIs to disk, you can simply reload them from disk at this stage, without re-finding them.

When performing fully-automated analysis, the BPF calculation has two parameters that may need to be adjusted, depending on the type of image being analysed.

Note: fully-automated analysis is strongly recommended to aid consistency of BPF measurements.

  1. Fitting Separate Grey- and White-Matter Peaks

    Some, more heavily T1-weighted, images have two peaks for the parenchyma: one for the grey matter, and one for the white matter. If this is the case for your images, then select brain_fit_grey_white

    To check whether this setting is appropriate for your images, you need to view the image intensity histogram of the brain. First mask the image to isolate just the brain and CSF. Then view the intensity histogram for the whole image by opening the image statistics dialog. Below is a 3-D FLASH image that requires separate peaks fitted to the grey and white matter.

    brain_SPGR

    brain_SPGR_histo

    Heavily T1-weighted 3-D FLASH image.

    Intensity histogram for the image to the left (all slices), with separate peaks for CSF, grey matter and white matter.

  2. Setting the Threshold Between CSF and Parenchyma

    When calculating the BPF, an intensity threshold is set which divides CSF from brain: all pixel with intensity below the threshold are considered to be CSF, and all those with intensity above are considered to be brain. The threshold is located between the histogram peaks for the CSF and brain (or grey-matter, if separately fitting GM and WM). The exact threshold location is determined by the CSF/brain threshold fraction:

    bpf_threshold_fraction

    Use the slider to set a value between 0 and 1. A value of 0 will put the threshold at the position of the CSF peak in the intensity histogram. A value of 1 will put the threshold at the position of the brain (or GM) peak in the intensity histogram. A setting of 0.5 (the default value) will put threshold half way between the CSF and brain peaks.

    If you find that too much brain tissue is being classed as CSF (the BPF is too low), then reduce the threshold fraction. Conversely, if you find that too much CSF is being classed as brain tissue (the BPF is too high), then increase the threshold fraction.

Note: whatever setting you use your images, for both cross-sectional and serial studies it is important to maintain the same setting for all patients scanned on any given scanner. Because of differences in contrast from scanner to scanner, it may be necessary to adjust the settings on a scanner-by-scanner basis, but for serial studies it is imperative that the same setting is always used for any given patient/scanner. You also have three further options:

  • If you do not want the CSF / parenchyma intensity threshold to be selected automatically, and you have another method for determining the threshold, then select:

brain_manual_threshold

and enter the threshold in the field provided. Any pixels with intensities above the threshold will be counted as parenchyma, and those below will be counted as CSF.
Note: when the threshold is set manually, no further image uniformity correction is performed as part of the BPF computation.

  • If it is clear from the result that BPF calculation is problematic, then you can help Jim to find the BPF by manually setting the initial guesses for the CSF, and parenchyma peak positions. Click the "Set initial guesses for peak locations manually" check box.

brain_set_initial_guesses

Using the whole-brain intensity histogram as a guide, enter your estimates of the peak positions for the CSF and parenchyma or (if you are fitting grey and white matter peaks separately) grey and white matter peaks.

  • For the purposes of verification, you can create images showing the distribution of parenchyma and CSF. To create these images, select brain_create_images. Images are created with the original image name plus suffixes "CSF" and "Brain" (or, if you are fitting grey and white matter peaks separately) "GM" and "WM".

Finally, click on the calc_bpf_button button. This fits a distribution of intensities to the histogram, assigns the peaks to CSF and parenchyma (or CSF, grey matter and white matter, if selected) and then chooses the intensity threshold between the CSF and parenchyma (or grey matter) peak positions according to the BPF CSF/brain threshold fraction (see above). The threshold is then applied to separate CSF from parenchyma, and the BPF calculated as:

BPF=Parenchyma Volume⁄(Parenchyma Volume + CSF Volume)

When complete, a window pops up showing the result, and giving you the opportunity to write these results as a permanent record to disk in a report.

bpf_result

To make a permanent record of these results, you can:

  • Write to a text file report. A File chooser will appear, for you to choose a log file name. The default file extension for log files is ".log". If the chosen file already exists, an entry will be appended to the log file.
  • Write to a PDF file report. A File chooser will appear, for you to choose a PDF file name. If the chosen file already exists, an entry will be appended to the PDF file. A PDF report will also include an illustration the brain/CSF segmentation.
  • Send a PDF report to a DICOM server. The DICOM server to which it will be sent must already have been set up, and is the current server selected in the DICOM Query/Retrieve tool. See these instructions for how to set up the connection to remote DICOM servers.
  • Save PDF and Send to DICOM. Both save to disk and send to a DICOM server.
  • Select the "Dismiss" option if you do not want to write a report.

A graph will also pop up showing the quality of the fit that is used to estimate the CSF/brain threshold.

brain_bpf_graph

The black line represents the image intensity histogram (after uniformity correction); the blue line represents the fit to the two (or three) peaks in the histogram; and the vertical dashed line is the threshold that divides CSF from brain.