 |
 |
Hi Mark, Thank you very much for the explanations. >>2. Under Viewing Timeseries (4D Images) the paragraph above the Histogram Toolbox figure. " >>Could you please explain what is demeaning and how it works? >Demeaning = removing the mean (i.e. subtracting the mean) a) If the time series signal has a mean of 3.5, demeaning means subtracting each data point by 3.5. Am I right? b) How does it work when there are more than 1 time series? For example, the mean of series1 is 3.5, the mean of series2 is 100 while the mean of series3 is -20. How does demeaning work in this case? >>7. Performing "fslmaths image0 -sub image1 imdiff" and "fslmaths imdiff -div image0 -mul 100 >>imdiffpercent" and then viewing the histogram of imdiff and imdiffpercent show that there are >>voxels of negative intensities. What does negative intensity mean? >That the subtraction above formed a negative number. What is the physical interpretation of negative intensity? Thank you. Date: Mon, 7 Oct 2013 06:37:50 +0000 From: [log in to unmask] Subject: Re: [FSL] Questions from Intro Practical To: [log in to unmask] Hi, On 6 Oct 2013, at 06:38, brain human <[log in to unmask]> wrote: Hi Mark, Thank you very much for your explanations. 1. Under "Intro to FSLView" last line "They don't have to have the same number of timepoints." Supposing that one image file has 2 timepoints and the other has 100 timepoints. I load both to fslview to view them at the same time. Since they have different number of timepoints, how do they synchronize with each other? When possible, the same volume number is used in each. 2. Under Viewing Timeseries (4D Images) the paragraph above the Histogram Toolbox figure. " Could you please explain what is demeaning and how it works? Demeaning = removing the mean (i.e. subtracting the mean) 3. Similarly, "choose to turn on the "%" button [4], which shows the y-axis as a % of the mean level". Could you please explain what you mean by "mean level" and how to calculate the % of the mean level? By pressing the %, the limits of the y-axis in this example changes from [-600 600] to [-3 3]. Percentage signal change = 100*(signal - mean signal)/(mean signal) 4. In Section fslstats, "This is because there are lots of empty voxels in the image (because the data has been brain-extracted)." In the Ortho view, the brain is in the center of each of the 3 views. Does empty voxel means voxel in black color? Does that include region surrounding the brain (i.e. every voxel/pixel in each view window that is outside the brain) or just voxels in black inside the brain? Empty voxels contain zeros. They include all potential voxels, so mainly those outside of the brain. 5.The sentence above "Run fslstats highres -m -M" in Section fslstats. "..but there is a large tail of very low background voxels. Tell the histogram to use log scale..." What do you mean by "low background voxels"? The tail is to represent voxels of HIGH intensity values. Tails exist on both sides (right and left, or high and low). 6. Under Section fslmaths, 3rd paragraph: "the robust range is quiet small, with most change showing up as motion effects round the edge of the brain." The robust range is quiet small because "fslstats imdiffpercent -r -R" gives [-47.78 37.69] as the robust range. Am I right? How can you tell "most change showing up as motion effects rund the edge of the brain"? By looking at the image and seeing the changes at the edge. 7. Performing "fslmaths image0 -sub image1 imdiff" and "fslmaths imdiff -div image0 -mul 100 imdiffpercent" and then viewing the histogram of imdiff and imdiffpercent show that there are voxels of negative intensities. What does negative intensity mean? That the subtraction above formed a negative number. 8. Under "Using the gradient threshold option (-g)" in Section "Troubleshooting Brain Extraction (Optional)", a) What is mean by "cerebellum is underestimated"? That there should be more cerebellum in the mask than is found. b) "This option causes the overall fractional threshold (controlled globally by -f) to vary linearly with slice number - to get smaller to the bottom and bigger at the top, or vice versa." How do I change the direction? You cannot - it is fixed as the z-direction, but you can change the sign. c) In command line, we can set -f -g. How about in the bet GUI? It seems that under the GUI, we can only set the gradient threshold option (-g) under "Threshold gradient; positive values give larger brain outline at bottom, smaller at top." No, the fractional intensity threshold is available in the GUI (in the main part, not in the advanced options). 9. There is a typo in "fslsplit and fslmerge". It should be "rm vol0002* vol0006*" rather than "rm vol0002* vol0005*" No, the instructions are correct. Image numbering starts at zero, so vol0002 is the 3rd volume, and vol0005 is the 6th. 10. The line above Section fslroi. "Also try concatenating (merging) in z instead of t and view in fslview; why does nothing happen when you try to turn on the movie loop?" Because you have not merged in time. This is what you should see. a) I got 8 brain images lining up vertically on top of each other. Am I doing something wrong? No, this is correct. It is simply a demonstration, not something you would normally do in practice. b) Why does nothing happen when I try to turn on the movie loop? See above. All the best, Mark Thank you. Date: Sat, 5 Oct 2013 22:25:53 +0000 From: [log in to unmask] Subject: Re: [FSL] Questions from Intro Practical To: [log in to unmask] Hi, 1. In the last sentence of "Viewing Atlases", ".. with the full probability information shown as a gradation of colour." What is the convention used in the gradation of color? The colour maps are arbitrary. Red-Yellow is commonly used, which has yellow for larger values and red for smaller ones. 2. Under "Viewing the results" in BET. "Change the transparency and colour of the overlay so that you can get a better view of the successes of the brain extraction." How do I change the colour? Use the image info widget and change the lookup table - see the fslview documentation at: http://fsl.fmrib.ox.ac.uk/fsl/fslview/ 3. Under "fslstats" in FSLUTILS, "Slightly confusingly, the robust range doesn't seem to be doing the same thing for the left and right tails." What do you mean here? This is just meant to indicate that the values do not appear at similar looking points with respect to the ends of the histogram. 4. "Click on the histogram toolbox button [3] and turn off the ignore zeros and press OK." After I turned off "ignore zeros", I saw a spike of amplitude about 6.9e6 at zero. Is this because of "lots of empty voxels in the image (because the data has been brain-extracted)"? Yes. 5. "Now it is clear that the robust range limits make more sense." I don't understand. a. When I issued "fslstats highres -R -r", I got: 0.0 32767 0 16711.1699. In the histogram, there is a hill-like shape from 0 to 2.1e4. Then, the signal approaches zero with a small peak (a triangle) at around 3.3e4. So, an absolute range of [0 32767] makes sense. As for the robust range, it is [0 16711.1699]. From the histogram, 1.6e4 is not the tail. About 2.2e4 is. Could you please let me know why it gave 1.6e4 rather than a value around 2.2e4? The upper robust range is closer to the 98th percentile in this case. b.When I unchecked "ignore zero", I saw a large spike of amplitude about 6.85e6 at 0 intensity. Is this because the data has been extracted so that there are lots of empty voxels in the image? Yes. c. When I issued "fslstats highres -1 1 -R -r", I got: 11.0 32767.0 2696.99 18976.72. The histogram shows that there are non-zero number of voxels from Intensity 0 to about 2.3e4. Why a robust range of [2696.99 18976.72] makes more sense? It is more like the 2nd and 98th percentiles (ignoring the very ends of the range). d. When I issued "fslview egepi" and viewed the histogram, I found the curve approaching the x and y axes asymptotically. The number of voxels with zero intensity is 36345. In the case of highres, the number of voxels with zero intensity is 7e6. So, there are fewer voxels with exactly 0 intensity as the text mentioned. Why? It is to do with brain extraction. e. In egepi, the tail of the histogram can be considered from Intensity 3000 to 2e4. The tail of highres is from around 2e4 to 3.5e4 which is longer. By "large tail of very low background voxels", do you just look at the curve of egepi alone rather than comparing it with the curve in highres? Why highres has a longer tail and at a higher Intensity than the egepi curve? The background voxels have a small value and are reflected by the left tail. No comparison between these images was intended. All the best, Mark Date: Sat, 5 Oct 2013 22:25:53 +0000 From: [log in to unmask] Subject: Re: [FSL] Questions from Intro Practical To: [log in to unmask] Hi, 1. In the last sentence of "Viewing Atlases", ".. with the full probability information shown as a gradation of colour." What is the convention used in the gradation of color? The colour maps are arbitrary. Red-Yellow is commonly used, which has yellow for larger values and red for smaller ones. 2. Under "Viewing the results" in BET. "Change the transparency and colour of the overlay so that you can get a better view of the successes of the brain extraction." How do I change the colour? Use the image info widget and change the lookup table - see the fslview documentation at: http://fsl.fmrib.ox.ac.uk/fsl/fslview/ 3. Under "fslstats" in FSLUTILS, "Slightly confusingly, the robust range doesn't seem to be doing the same thing for the left and right tails." What do you mean here? This is just meant to indicate that the values do not appear at similar looking points with respect to the ends of the histogram. 4. "Click on the histogram toolbox button [3] and turn off the ignore zeros and press OK." After I turned off "ignore zeros", I saw a spike of amplitude about 6.9e6 at zero. Is this because of "lots of empty voxels in the image (because the data has been brain-extracted)"? Yes. 5. "Now it is clear that the robust range limits make more sense." I don't understand. a. When I issued "fslstats highres -R -r", I got: 0.0 32767 0 16711.1699. In the histogram, there is a hill-like shape from 0 to 2.1e4. Then, the signal approaches zero with a small peak (a triangle) at around 3.3e4. So, an absolute range of [0 32767] makes sense. As for the robust range, it is [0 16711.1699]. From the histogram, 1.6e4 is not the tail. About 2.2e4 is. Could you please let me know why it gave 1.6e4 rather than a value around 2.2e4? The upper robust range is closer to the 98th percentile in this case. b.When I unchecked "ignore zero", I saw a large spike of amplitude about 6.85e6 at 0 intensity. Is this because the data has been extracted so that there are lots of empty voxels in the image? Yes. c. When I issued "fslstats highres -1 1 -R -r", I got: 11.0 32767.0 2696.99 18976.72. The histogram shows that there are non-zero number of voxels from Intensity 0 to about 2.3e4. Why a robust range of [2696.99 18976.72] makes more sense? It is more like the 2nd and 98th percentiles (ignoring the very ends of the range). d. When I issued "fslview egepi" and viewed the histogram, I found the curve approaching the x and y axes asymptotically. The number of voxels with zero intensity is 36345. In the case of highres, the number of voxels with zero intensity is 7e6. So, there are fewer voxels with exactly 0 intensity as the text mentioned. Why? It is to do with brain extraction. e. In egepi, the tail of the histogram can be considered from Intensity 3000 to 2e4. The tail of highres is from around 2e4 to 3.5e4 which is longer. By "large tail of very low background voxels", do you just look at the curve of egepi alone rather than comparing it with the curve in highres? Why highres has a longer tail and at a higher Intensity than the egepi curve? The background voxels have a small value and are reflected by the left tail. No comparison between these images was intended. All the best, Mark