Tuesday, July 26, 2011

Inner ear tumor segmentation procedure

As it was described in previous sections the character of analyzed (limited volume, various
intensity conditions) MRI data made implementation of common used algorithms very
difficult. One of the ways solving such problem may be a-priori knowledge of the place of
potential occurance of acoustic schwannoma tissue. Such objects as seen at Fig.1a are
localized in the neighborhood of head centre and inner ear. It is important information that
can improve processing due to decreased number of potential objects that may be classified
as tumor. For that issue volume of interest need to be created according to volume. This
information supported by local and  global data histogram processing may be useful in
recognizing tissues of interest. Segmentation algorithm proposed followed by recognizing
procedure.
The contrast injected into patient before MRI scanning changes intensity response within
acoustic schwannoma tissue. However the intensity does not increase enough to create in a
histogram a clear separate peak describ- ing the mentioned tumor. This peak lies slightly
beyond second peak that describes brain This observation allows to divide the
segmentation process into 2 steps. In the first step the peak of brain, tumor and other
objects possessing the same intensity range, are treated as a single peak. The global
histogram (Fig.4a) can be treated as bimodal. Under such assumption the best solution
would be dynamically adapting threshold value to every histogram.
This demand is fulfilled by Otsu method [7]. The effect of thersholding The voxels of interest
that are used to build new histogram are a part of binary mask and are placed inside a
sphere with a centre and a radius defined empirically. Radius of a sphere delimities a
position in relation In the second segmentation stage the histogram shows bimodal structure
also. Such bimodal character of local histogram emerges when huge background peak is
removed. Huge peak seen in lower intensities represents brain and the smaller one
represents tumor and other structures covering the same intensity range. Character of local
histogram allows usage of Otsu method for second stage segmentation threshold. An
exemplary MRI image consists of tumor segmented structures is presented below (Fig.6a):
 Local histogram thresholding operation finalizes fully automated inner ear tumo
segmentation process. The next step is interactive and requires intervention.
 
As it was described in previous sections the character of analyzed (limited volume, various
intensity conditions) MRI data made implementation of common used algorithms very
difficult. One of the ways solving such problem may be a-priori knowledge of the place of
potential occurance of acoustic schwannoma tissue. Such objects as seen at Fig.1a are
localized in the neighborhood of head centre and inner ear. It is important information that
can improve processing due to decreased number of potential objects that may be classified
as tumor. For that issue volume of interest need to be created according to volume. This
information supported by local and  global data histogram processing may be useful in
recognizing tissues of interest. Segmentation algorithm proposed followed by recognizing
procedure.
The contrast injected into patient before MRI scanning changes intensity response within
acoustic schwannoma tissue. However the intensity does not increase enough to create in a
histogram a clear separate peak describ- ing the mentioned tumor. This peak lies slightly
beyond second peak that describes brain This observation allows to divide the
segmentation process into 2 steps. In the first step the peak of brain, tumor and other
objects possessing the same intensity range, are treated as a single peak. The global
histogram (Fig.4a) can be treated as bimodal. Under such assumption the best solution
would be dynamically adapting threshold value to every histogram.
This demand is fulfilled by Otsu method [7]. The effect of thersholding The voxels of interest
that are used to build new histogram are a part of binary mask and are placed inside a
sphere with a centre and a radius defined empirically. Radius of a sphere delimities a
position in relation In the second segmentation stage the histogram shows bimodal structure
also. Such bimodal character of local histogram emerges when huge background peak is
removed. Huge peak seen in lower intensities represents brain and the smaller one
represents tumor and other structures covering the same intensity range. Character of local
histogram allows usage of Otsu method for second stage segmentation threshold. An
exemplary MRI image consists of tumor segmented structures is presented below (Fig.6a):
 Local histogram thresholding operation finalizes fully automated inner ear tumo
segmentation process. The next step is interactive and requires intervention.


Puppies for sale

http://www.articlesbase.com/internet-articles/inner-ear-tumor-segmentation-procedure-5067545.html

No comments:

Post a Comment