Medical Image Processing - Medical Image Processing: Advanced Fuzzy Set Theoretic Techniques - page 33

The main reason to carry out research using advanced fuzzy set theory is
to obtain better and accurate results for better diagnosis.
2.7 Some Applications of Advanced Fuzzy
Set in Medical Image Processing
Fuzzy image analysis finds applications in many computer vision systems
including robot vision, object recognition, biomedical image processing and
remotely sensed scene analysis. With the introduction of intuitionistic fuzzy
set and Type II fuzzy set, researches are carried out on medical images for
obtaining better results. However, applying a single image analysis method
may not always yield reliable results. It is observed that when an ensemble
of image processing techniques is applied to an image, appropriate results
are likely to be obtained. The raw medical image that is initially obtained
does not have better contrast. So, the image is required to enhance before
going to another process. Next, segmentation is required before analysis.
Segmentation divides the abnormal lesions or haemorrhage/clots or blood
vessels and different types of leukocytes. After segmentation, analysis
is done.
A few applications of advanced fuzzy set theory in medical image process-
ing are discussed:
1. Image contrast enhancement : There are many image enhancement
methods, and the most common is histogram equalization. But
this method may not always yield good results on medical images,
since pixel grey levels in an image are imprecise. To deal with such
kind of images, many authors suggested fuzzy methods to handle
the inexactness of grey values. The most common operator is the
contrast intensification (INT) operator, which is applied globally
to modify the membership and increase the contrast of the image.
This approach transforms the higher membership values to much
higher values or lower membership values to much lower values in
a non-linear manner. But this method has some limitation, which
is improved using the NINT operator that uses Gaussian-type
fuzzification function. In many cases, global histogram approaches
to enhancement fail to achieve satisfactory results as these meth-
ods consider the whole image in its totality, so local enhancement
techniques work better. Enhancement using fuzzy methods per-
forms well as it considers the vagueness in the image, but in some
cases, proper enhancement is not achieved using the fuzzy method.
Then intuitionistic and Type II fuzzy set theories are used that use
more uncertainties or different types of uncertainty. This is done to