timz
New Member
Posts: 14 
|
Post by timz on Jan 24, 2014 20:08:54 GMT -5
I think the reason to use a magnet instead of a fixed connection is the user can adjust the position of the chuck. But, when the rotational stage rotates, there is a rotation axis in the space anyway. Even if the sample is positioned off the rotation axis, in principle, the data can still be reconstructed (if the position of the rotation axis is known). However, once setup, the rotation axis should not change unless the stage is moved. So, why does the user need to adjust the position of the chuck (assume the sample is small and won't get out of FOV no matter where the rotation axis is)? So, in the current configuration and software, is it assumed that the rotation axis falls right on the middle of the FOV? How critical is the assumption to the reconstruction?
|
|
|
|
Post by JDazai on Jan 29, 2014 10:24:35 GMT -5
When using a high NA, the depth of focus is quite small. So, if the sample is off centre it may become out of focus during the rotation. If you can use a lower NA and lose resolution then it may work. Also, if you can consistently mount the sample directly in the middle of the chuck then there would be no need to move it using a magnetic chuck. However, we have not yet found a method to do so.
|
|
timz
New Member
Posts: 14
|
Post by timz on Jan 29, 2014 19:54:02 GMT -5
So, from the specifications of the T100 system, at 0.52X the DOF is only 6.2 mm. But I guess the system mostly works at high magnifications. Then at 5.2X the DOF is only 62 microns. So, if the sample has to be in focus during the rotation, the sample has to be thinner than DOF. I don't have much experience with the samples, but I guess most samples should be within 1 ~ 10mm. And in the paper, the sample seems to be thicker/wider than 10mm, then how did you achieve that?
|
|
|
|
Post by mwong on Feb 6, 2014 14:18:12 GMT -5
Hi Tim,
Depth of Field is a difficult thing to interpret because its not a standardized measure. What is acceptable during OPT imaging in terms of depth of field is not equivalent to the value listed by the microscope/objective manufacturer.
In addition, we do not image our samples at max magnification. With the system we have, we image E9.5 embryos at approximately 2.5x magnification out of the alloted 5.2x magnification. And for the E12.5 mouse embryo in the paper, the magnification is dramatically less than that, approximately 1-2x magnification. This affords us larger depth of field to cover our embryos. As well, we only use the center 1024 x 1024 crop of our 2048 x 2048 CCD to both afford us more depth of field (less magnification to fill the FOV with the embryo) as well as to counteract vignetting.
|
|
timz
New Member
Posts: 14
|
Post by timz on Feb 27, 2014 14:46:29 GMT -5
Thanks to Jun and Michael fot the answers. I have not checked the codes, neither the reconstruction software, but back to the original question: is there anywhere in the Matlab code or the reconstruction software that assume the rotational axis fall on the middle line? If it is off a bit (not so much that makes sample out of focus during rotation), will it affect the reconstruction? Does the reconstruction software figure out the position of the rotational axis each time it process the data?
|
|
|
|
Post by mwong on Feb 28, 2014 14:06:05 GMT -5
Neither the matlab/labview/reconstruction code can correct for a rotation axis that is not within 0.01-0.05 degrees of totally vertical. That is what the calibration with the phantom is for. We did not write in a compensation of this in any of the code, because through experience the rotational axis did not move/deviate over months of use. Therefore, we saw fit that just an initial alignment of the rotational axis along the y-axis is needed. However, the Nrecon code does correct for if the rotational axis is not exactly on the 512 centre line. It can compensate for as much as 50 pixels off the centre line. It does this by looking at a projection image at 0 degrees rotation and a mirrored projection at 180 degrees rotation. If these two projections are identical, that means that the axis of rotation is centered. The software tries to pick the best post-alignment value by looking for similarity between line profiles through these two projection images. But it is always best for the user to do a series of reconstruction of the centre slices and visually pick out the best post alignment value.
So in short, the axis of rotation does not need to be exactly on the 512 centre line, but should be made as close as possible. But the axis of rotations does need to be completely vertical.
|
|
