22-05-2012, 11:01 AM
breast tissue classification using statistical feature extraction of mammograms
[attachment=22660]
INTRODUCTION
Texture segmentation has long been an important topic in
image processing. Basically, it aims at segmenting a textured
image into several regions with the same texture features. An
effective and efficient texture segmentation method will be very
useful in applications like the analysis of aerial images, biomedical
images and seismic images as well as the automation of industrial
applications [1]. Like the other segmentation problems, the
segmentation of textures requires the identification of proper
texture-specific features with good discriminative power.
Generally speaking, texture feature extraction methods can be
classified into three major categories, namely, statistical, structural
and spectral [2]
Statistical Approaches
In general, any image processing and analysis applications
would require a particular feature for classification /segmentation.
Mainly texture features and statistical features are of more significant
in pattern recognition area. A frequently used approach for texture
analysis is based on statistical properties of intensity histogram. One
such measures is based on statistical moments.
MATERIALS AND METHODS
The mathematical model for the moments to compute the
six texture features of a mammogram are as listed in
Matlab functions to implement the same have been developed
and tested over certain mammograms. The programs were tested
over ten selected images from mini-MIAS data base. [7]. The results
obtained are as tabulated in . The basic classification
based on the values of the texture parameters are as shown in
The original image and their corresponding histograms
for basic classifications are also shown in . From
the above results it can be inferred that the statistical features
extracted from the mammogram images are useful parameters for
tissue classification [8].
SIMULATED RESULTS
As the Mini MIAS database consist of 332 mammograms of
different categories, it has been selected for the testing of
performance of the proposed algorithms. As per the literature
of the above database the mammograms have been grouped
under only three categories like, fatty, glandular and dense.
According to the recent research results of University of
Calgary, (Biomedical Engineering Research Group), these have
been further classified based on some statistical features in to
the four classes as Uncompressed fatty, fatty, Nonuniform, and
high density [9]. This classification would help a radiologist to
determine the breast anatomy (fibroglandular tissue) affected
due to Estrogen secretion. when a patient is under harmone
replacement therapy (HRT) [10].
CONCLUSION
The method employed here has given better performance. The
results have been validated by visual inspection by an expert
radiologist. Certain mammogram images (60) of different
abnormalities randomly chosen from the data base have been
selected for our experiment and the algorithm applied over it have
classified them in accordance to the category as stated above.
Further the results of our experiments show very clearly the
appearance of any abnormality in the growth of breast tissue.
[attachment=22660]
INTRODUCTION
Texture segmentation has long been an important topic in
image processing. Basically, it aims at segmenting a textured
image into several regions with the same texture features. An
effective and efficient texture segmentation method will be very
useful in applications like the analysis of aerial images, biomedical
images and seismic images as well as the automation of industrial
applications [1]. Like the other segmentation problems, the
segmentation of textures requires the identification of proper
texture-specific features with good discriminative power.
Generally speaking, texture feature extraction methods can be
classified into three major categories, namely, statistical, structural
and spectral [2]
Statistical Approaches
In general, any image processing and analysis applications
would require a particular feature for classification /segmentation.
Mainly texture features and statistical features are of more significant
in pattern recognition area. A frequently used approach for texture
analysis is based on statistical properties of intensity histogram. One
such measures is based on statistical moments.
MATERIALS AND METHODS
The mathematical model for the moments to compute the
six texture features of a mammogram are as listed in
Matlab functions to implement the same have been developed
and tested over certain mammograms. The programs were tested
over ten selected images from mini-MIAS data base. [7]. The results
obtained are as tabulated in . The basic classification
based on the values of the texture parameters are as shown in
The original image and their corresponding histograms
for basic classifications are also shown in . From
the above results it can be inferred that the statistical features
extracted from the mammogram images are useful parameters for
tissue classification [8].
SIMULATED RESULTS
As the Mini MIAS database consist of 332 mammograms of
different categories, it has been selected for the testing of
performance of the proposed algorithms. As per the literature
of the above database the mammograms have been grouped
under only three categories like, fatty, glandular and dense.
According to the recent research results of University of
Calgary, (Biomedical Engineering Research Group), these have
been further classified based on some statistical features in to
the four classes as Uncompressed fatty, fatty, Nonuniform, and
high density [9]. This classification would help a radiologist to
determine the breast anatomy (fibroglandular tissue) affected
due to Estrogen secretion. when a patient is under harmone
replacement therapy (HRT) [10].
CONCLUSION
The method employed here has given better performance. The
results have been validated by visual inspection by an expert
radiologist. Certain mammogram images (60) of different
abnormalities randomly chosen from the data base have been
selected for our experiment and the algorithm applied over it have
classified them in accordance to the category as stated above.
Further the results of our experiments show very clearly the
appearance of any abnormality in the growth of breast tissue.