[Insight-users] Fasting marching algorithm provided by ITK

Dan Mueller dan.muel at gmail.com
Wed Jun 16 13:27:55 EDT 2010


Hi Yang,

Fast marching expansion is guided by a speed function. The expanding
front propagates based on the *intensity* of the given input image
(called speed image). Front leakage is a common problem of the fast
marching method. The advantage is that it is fast (:P), and is good
for obtaining an initial contour for further refinement. The front
propagation is NOT controlled by advection or curvature terms (like a
level set). The front can only expand, it can not contract once it has
leaked into unwanted areas.

Some options:
    (1) Use a mapping function (sigmoid, window-level) to enhance the
speed image so that areas you want the front to expand are 1.0 (white)
and areas you want the front to stop are 0.0 (black)
    (2) Consider subtracting the gradient magnitude (or some edge
measure) of the speed image from itself (this may help prevent
leakage)
    (3) Consider using a level set or alternative method.

I would recommend reading
    Level Set Methods and Fast Marching Methods
    J. A. Sethian, Cambridge University Press, 1999
    http://math.berkeley.edu/~sethian/2006/Publications/Book/2006/book_1999.html

HTH

Cheers, Dan

2010/6/16 Xiaopeng Yang <yxp233 at postech.ac.kr>:
> Hi Dan,
>
> I set the StoppingValue as 100. I attached my segmentation result which made
> me confused. I want to segment liver, but even the result has covered the
> part outside of liver, still there is a gap between the liver contour and
> the segmentation result. Could you explain this?
>
> Thank you Dan.
>
> Yang
>
> -----邮件原件-----
> 发件人: Dan Mueller [mailto:dan.muel at gmail.com]
> 发送时间: Wednesday, June 16, 2010 9:53 PM
> 收件人: Xiaopeng Yang
> 抄送: insight-users at itk.org
> 主题: Re: [Insight-users] Fasting marching algorithm provided by ITK
>
> Hi Yang,
>
> A picture clearly explaining your problem would be helpful...
>
> From your description, it seems the fast marching filter is
> terminating earlier than you expect. One possible reason is due to the
> StoppingValue parameter. Are you setting the StoppingValue? By default
> this value is set very high to ensure the marching does not terminate
> earl, however if you override this value...
>
> Please let us know.
>
> Regards, Dan
>
> On 16 June 2010 14:30, Xiaopeng Yang <yxp233 at postech.ac.kr> wrote:
>> Dear ITK users,
>>
>> About the fasting marching algorithm provided by ITK, does anybody can
>> explain why it always cannot reach the contour?
>>
>> Thanks.
>>
>> Yang


More information about the Insight-users mailing list