[Mristudio-users] Why deleting b0 from one of the runs increases FA a lot.

Fri Sep 27 20:06:17 EDT 2013

each scan does gain calibration independently.
Remember, the intensity scale is an arbitrary unit.

For example, 16-bit digitizer has 2^16 scales to record the signal
intensity. That is +/- 32K. The scanner record a signal as tiny as 1 and as
large as 32,000. This is called dynamic range.

Throughout the scan, you have to make sure that the most intense signal has
to be smaller than 32,000 while you want to use the dynamic range as wide
as possible. So, during the pre-scan, the scanner estimate the largest
signal and attenuate the gain (signal amplifier) so that the entire signal
fits within 32,000. Of course, the scanner makes very conservative
estimation because the worst thing happens is the signal clips the 32,000
scale range (like the signal is 40,000, which can't be recorded).

Even if the scanner is perfectly reproducible, generating the same amount
of signal, the final recorded signals in the digitizer depends on the gain
and this gain setting is influenced by the short prescan, which is affected
by noise.

So, long story short, the absolute intensity scale between two separate
scans can hardly an indicator of scanner performance unless you turn off
this auto-gain calibration.

On Fri, Sep 27, 2013 at 7:19 PM, Dorian P. <alb.net at gmail.com> wrote:

> Thank you Dr. Mori. Your explanation was very helpful.
>
> About DWI intensities, I am still perplexed to why the same subject gets
> so different DWI intensities, only 5 minutes apart. There must be something
> to do with the machine, all else being equal.
>
> Best regards.
>
> 2013/9/27 Susumu Mori <smoriw at gmail.com>
>
>> I think it's happens. This is an expected range. Unless you turn off
>> auto-gain setting in the second run, the scanner does automatic gain
>> setting (something like automated determination of the volume of your
>> stereo) and the signal intensity (equivalent of the loudness of the sound)
>> varies.
>> If you combine these two, the first scan with higher intensities receive
>> more weighting (when both scans have corresponding b0s). If one of the
>> scans lacks b0, then you have a problem.
>> I thought Hangyi once tried to implement cross-scan intensity
>> normalization, but not sure if it is a default setup of the current version.
>> In your case, however, because you combine the two scans as one set, this
>> type of cross-scan intensity normalization won't work. So you definitely
>> have a problem by combining the two scans.
>>
>>
>> On Fri, Sep 27, 2013 at 4:06 PM, Dorian P. <alb.net at gmail.com> wrote:
>>
>>> Average DWI for the arcuate fibers, extracted from
>>>
>>> Session 1 alone
>>> mean DWI = 42.8095
>>> mean b0 = 85.4266
>>> FA= 0.5014
>>>
>>>
>>> Session 2 alone
>>> mean DWI = 54.6673
>>> mean b0 = 109.7234
>>> FA = 0.4830
>>>
>>>
>>> Is it abnormal to have these differences between DTI runs? They were
>>> acquired one after the other, no other MRI sequence in between.
>>>
>>>
>>>
>>> P.s. Calculations are from a subject that has both sessions complete, no
>>> b0 missing.
>>>
>>>
>>>
>>> 2013/9/27 Susumu Mori <smoriw at gmail.com>
>>>
>>>> Then I believe the two sets have intensity differences by 10-20%.
>>>> Create the sum of all DWIs from the two scans and compare the intensity.
>>>>
>>>>
>>>> On Fri, Sep 27, 2013 at 3:27 PM, Dorian P. <alb.net at gmail.com> wrote:
>>>>
>>>>> MD seem to go down for the partial dataset:
>>>>>
>>>>> Trace:
>>>>> 2 DTI runs (normal) = 2.528e-003
>>>>> 2 DTI runs (exclude 1 b0) = 2.061e-003
>>>>> 1 DTI run (normal) = 2.493e-003
>>>>>
>>>>> Same ROIs were used to extract the fasciculus in each case.
>>>>>
>>>>>
>>>>> 2013/9/27 Susumu Mori <smoriw at gmail.com>
>>>>>
>>>>>> how about MD?
>>>>>> Maybe intensities of two scans were different.
>>>>>>
>>>>>>
>>>>>> On Fri, Sep 27, 2013 at 2:56 PM, Dorian P. <alb.net at gmail.com> wrote:
>>>>>>
>>>>>>> Hi Dr. Mori
>>>>>>>
>>>>>>> That was my strategy. I  combined the two runs, b0+32DWI+32DWI. This
>>>>>>> was done of course after registration, so that I could use the new gradient
>>>>>>> table. Images were saved in raw format, and loaded back with the new
>>>>>>> gradient table.
>>>>>>>
>>>>>>> I tried both using all images with gradient 100, 100, 100 for the
>>>>>>> bad b0 or completely deleting the bad b0 (and the corresponding gradient
>>>>>>> line). Either way FA goes up a lot for the same measured tract.
>>>>>>>
>>>>>>> Here are some more tests on left arcuate from a good subject with 2
>>>>>>> DTI runs. Values are FA:
>>>>>>> 2 DTI runs (normal) = 0.489
>>>>>>> 2 DTI runs (exclude 1 b0) = 0.55
>>>>>>> 1 DTI run (normal) = 0.499
>>>>>>>
>>>>>>> It seems that excluding one b0 is much worse than excluding the
>>>>>>> whole DTI run. In other words, adding DWI images from a sequence that
>>>>>>> doesn't have its own b0 may invalidate the findings. If seems that a set of
>>>>>>> DWI images without a b0 acquired during the same sequence is not useful to
>>>>>>> to include in DTI calculation.
>>>>>>>
>>>>>>> Does this make sense?
>>>>>>>
>>>>>>> Dorian
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> 2013/9/27 Susumu Mori <smoriw at gmail.com>
>>>>>>>
>>>>>>>> Do you mean, you used different gradient tables for different runs?
>>>>>>>> I wonder if the gradient table and images are not matched for a
>>>>>>>> part of data.
>>>>>>>> Suppose you have 1 b0 + 12 DWIs x 2 runs and you lost one b0 in the
>>>>>>>> second run.
>>>>>>>> The only way to analyze this data, I can think of, is, combine the
>>>>>>>> two runs as 1 b0 + 12 DWIs + 12 DWIs and crate one 25-element gradient
>>>>>>>> table.
>>>>>>>>
>>>>>>>> In general, if you have less data, you have lower SNR and FA goes
>>>>>>>> up but I doubt if the loss of one b0 could cause such a large FA increase.
>>>>>>>>
>>>>>>>>
>>>>>>>> On Fri, Sep 27, 2013 at 10:38 AM, Dorian P. <alb.net at gmail.com>wrote:
>>>>>>>>
>>>>>>>>> Dear developers and Dr. Mori,
>>>>>>>>>
>>>>>>>>> I am comparing FA for some tracts at two time points. At each time
>>>>>>>>> point I acquire 2-3 DTI runs to increase SNR. However, the scanner has
>>>>>>>>> malfunctioned for one of the runs in one subject. As a result the B0 image
>>>>>>>>> is not saved, but all other DWIs are there. So, I set the collapsed B0 to
>>>>>>>>> gradient 100, 100, 100 and align all DWIs to the only B0 available. For
>>>>>>>>> some results, the resulting FA is increased a lot.
>>>>>>>>>
>>>>>>>>> I tried the same procedure with another subject, when I set one of
>>>>>>>>> the B0 to gradient 100, 100, 100, the resulting FA of the tract goes up
>>>>>>>>> from 0.51 to 0.62.
>>>>>>>>>
>>>>>>>>> Is there any explanation why this is happening, and whether I
>>>>>>>>> should delete completely the collapsed B0 instead of setting it to 100,
>>>>>>>>> 100, 100?
>>>>>>>>>
>>>>>>>>> Thank you.
>>>>>>>>> Dorian
>>>>>>>>> TJU
>>>>>>>>>
>>>>>>>>> _______________________________________________
>>>>>>>>> mristudio-users mailing list
>>>>>>>>> mristudio-users at mristudio.org
>>>>>>>>> http://lists.mristudio.org/mailman/listinfo/
>>>>>>>>> Unsubscribe, send a blank email to:
>>>>>>>>> mristudio-users-unsubscribe at mristudio.org
>>>>>>>>>
>>>>>>>>>
>>>>>>>>
>>>>>>>> _______________________________________________
>>>>>>>> mristudio-users mailing list
>>>>>>>> mristudio-users at mristudio.org
>>>>>>>> http://lists.mristudio.org/mailman/listinfo/
>>>>>>>> Unsubscribe, send a blank email to:
>>>>>>>> mristudio-users-unsubscribe at mristudio.org
>>>>>>>>
>>>>>>>>
>>>>>>>
>>>>>>> _______________________________________________
>>>>>>> mristudio-users mailing list
>>>>>>> mristudio-users at mristudio.org
>>>>>>> http://lists.mristudio.org/mailman/listinfo/
>>>>>>> Unsubscribe, send a blank email to:
>>>>>>> mristudio-users-unsubscribe at mristudio.org
>>>>>>>
>>>>>>>
>>>>>>
>>>>>> _______________________________________________
>>>>>> mristudio-users mailing list
>>>>>> mristudio-users at mristudio.org
>>>>>> http://lists.mristudio.org/mailman/listinfo/
>>>>>> Unsubscribe, send a blank email to:
>>>>>> mristudio-users-unsubscribe at mristudio.org
>>>>>>
>>>>>>
>>>>>
>>>>> _______________________________________________
>>>>> mristudio-users mailing list
>>>>> mristudio-users at mristudio.org
>>>>> http://lists.mristudio.org/mailman/listinfo/
>>>>> Unsubscribe, send a blank email to:
>>>>> mristudio-users-unsubscribe at mristudio.org
>>>>>
>>>>>
>>>>
>>>> _______________________________________________
>>>> mristudio-users mailing list
>>>> mristudio-users at mristudio.org
>>>> http://lists.mristudio.org/mailman/listinfo/
>>>> Unsubscribe, send a blank email to:
>>>> mristudio-users-unsubscribe at mristudio.org
>>>>
>>>>
>>>
>>> _______________________________________________
>>> mristudio-users mailing list
>>> mristudio-users at mristudio.org
>>> http://lists.mristudio.org/mailman/listinfo/
>>> Unsubscribe, send a blank email to:
>>> mristudio-users-unsubscribe at mristudio.org
>>>
>>>
>>
>> _______________________________________________
>> mristudio-users mailing list
>> mristudio-users at mristudio.org
>> http://lists.mristudio.org/mailman/listinfo/
>> Unsubscribe, send a blank email to:
>> mristudio-users-unsubscribe at mristudio.org
>>
>>
>
> _______________________________________________
> mristudio-users mailing list
> mristudio-users at mristudio.org
> http://lists.mristudio.org/mailman/listinfo/
> Unsubscribe, send a blank email to:
> mristudio-users-unsubscribe at mristudio.org
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.mristudio.org/pipermail/mristudio-users/attachments/20130927/7703040d/attachment-0001.html 