[Mristudio-users] ROI

[susumu mori]

Yes, of course, I would recommend our atlas-based automated ROI definition.
However, we measure the accuracy of such automated ROI placement against
manual ROI, meaning we usually consider manual ROI is more accurate.

This is actually true if your patient population has severe anatomical
abnormalities and you can't expect accurate results with automated methods.

Then next question is how we judge "severe anatomical abnormalities?"
Unfortunately, we don't know it until we run through the automated method.
That means, we need to check the results one-by-one to make sure the
automated method produced accurate results.

Then how we judge if it is accurate or not? It is based on subjective
judgment. Therefore, even if we do automated ROI placement, at the end of
the day, it is our subjective judgment to call if it is working or not.

So, you could argue why we don't do manual ROI from the beginning? If the
structures of your interest are only a small number, say 2 structures. I
would say it could be easier to perform manual ROI. Even with severe
pathology cases, usually you can identify the structure fairly accurately.

The downside of the manual ROI approach is, you would leave many brain
structures unmeasured. So, the manual ROI approach must be hypothesis
driven; e.g. "we suspect the hippocampus is the damaged structure and
therefore we measure hippocampus."

If you are not sure where is abnormality, you'd better measure the entire
brain using the automated approach.

Probably, the best way is to do both.
Hypothesis driven for a small set of structures with manual ROI.
Hypothesis generation using the automate ROI and if there is any
abnormality, confirm by manual ROI.

On Mon, Mar 11, 2013 at 1:36 PM, Dorian P. <alb.net at gmail.com> wrote:

> Dr. Mori, would you advise to overlay the white matter atlas and obtain
> regional FAs that way? If not, why?
>
> I am aware that tracts change in each person, but sounds the best way to
> categorize FA regions without going for tractography.
>
> Dorian
>
> 2013/3/11 susumu mori <susumu at mri.jhu.edu>
>
>> Unless you are using tractography-based pixel grouping, you have to visit
>> every slice to make a 3D ROI.
>>
>>
>> On Mon, Mar 11, 2013 at 7:23 AM, Shaimaa Abdelsattar <
>> shaimaa96 at hotmail.com> wrote:
>>
>>> Thank you very much,
>>>
>>> So if I draw 2 ROI s one  at the level A and another one at level B
>>> (higher by 3 cuts  for example.) using the option (+), to perform 3d ROI as
>>> you mentioned in tutorial of ROI editor , is the result represents the
>>> whole volume in between the two cuts or only these 2 cuts individually?, or
>>> should i draw multiple Rois in each cut in between?
>>>
>>> Thanks again
>>>
>>> On Mar 11, 2013, at 12:17 PM, "susumu mori" <susumu at mri.jhu.edu> wrote:
>>>
>>> Hi Shaimma,
>>>
>>> For anatomical definition, there is no "right" way to define structures
>>> because anatomy doesn't always have clear boundary. It is like a cloud;
>>> very often there is only a gradual transition from one structure to the
>>> other.
>>>
>>> I once heard a presentation about a European consortium to try to come
>>> up with a consistent criteria for hippocampus definition and found there
>>> was as much as 100% difference in the hippocampal sizes defined by
>>> neuroanatomists.
>>>
>>> When you try to define a structure, like "internal capsule", because WM
>>> often doesn't have clear boundary, you need to come up with a written
>>> protocol to define it based on anatomical landmarks. Then you need to find
>>> intra-rater and inter-rater reproducibility. Your measurements are trusted
>>> only if you present these reproducibility results.
>>>
>>> The ROI can be as simple as one ROI in one representative slice. This
>>> could lead to high reproducibility if you choose a slice in which your
>>> tract of interest has clear boundary, but reproducibility could be "0
>>> (meaning no overlap)" if another person happens to choose a adjacent slice.
>>> Also, you would leave many portion of the tract of your interest
>>> unmeasured. The reproducibility issue could be solved if your protocol
>>> says, "first normalize the brain into the MNI space using affine
>>> transformation and choose axial slice 90".
>>>
>>> If your protocol suggests, "all axial slices from the anterior
>>> commissure to the level of the upper boundary of the thalamus in the
>>> mid-sagittal slice", then you may have to define the internal capsule in 10
>>> axial slices.
>>>
>>> Again, I'm not saying one is better than the other. Location
>>> identification is the greatest challenge for image analysis and there is no
>>> single way to solve it. Voxel-based analysis can give you automated way
>>> with 100% reproducibility, but it doesn't solve the accuracy problem (can
>>> it really align structures accurately?).
>>>
>>>
>>>
>>>
>>> On Sun, Mar 10, 2013 at 1:35 PM, Shaimaa Abdelsattar <
>>> shaimaa96 at hotmail.com> wrote:
>>>
>>>> Hi,
>>>> In ROI editor program, if I want to measure FA ,for example in the
>>>> posterior limb of internal capsule, what is the correct method for this ,is
>>>> it  to draw multiple Roi s in all sequential axial planes till i finish it,
>>>> or draw ROI in only few cuts or draw one small  Roi within its substance,
>>>> or all can work?
>>>>
>>>> Thank you very much
>>>> _______________________________________________
>>>> 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/20130311/e9723f5c/attachment.html