Image Processing Reference
invasive, non-contact, passive, radiation-free technique that can also be used in combination
with anatomical investigations based on x-rays and three-dimensional scanning techniques
such as CT and MRI and often reveals problems when the anatomy is otherwise normal. It
is well known that the radiance from human skin is an exponential function of the surface
temperature which in turn is influenced by the level of blood perfusion in the skin. Thermal
imaging is hence well suited to pick up changes in blood perfusion which might occur due
to inflammation, angiogenesis or other causes. Asymmetrical temperature distributions as
well as the presence of hot and cold cold are known to be strong indicators of an underlying
dysfunction (Uematsu, 1985).
Despite earlier, less encouraging studies, which were based on low capability and poorly
calibrated equipment infrared imaging has been shown to be well suited for task of detecting
breast cancer, in particular when the tumor is in its early stages or in dense tissue (Anbar,
Milescu, Naumov, Brown, Button, Carly, and AlDulaimi, 2001; Head, Wang, Lipari, and
Elliott, 2000). Early detection is important as it provides significantly higher chances of
survival (Ng and Sudarshan, 2001) and in this respect infrared imaging outperforms the
standard method of mammography which can detect tumors only once they exceed a certain
size. On the other hand, tumors that are small in size can be identified using thermography.
As cancer cells have a high metabolic activity this leads to an increase in local temperature
which can be picked up in the infrared.
In this section, we perform breast cancer detection based on thermography, using a series
of statistical features extracted from the thermograms coupled with a fuzzy rule-based
classification system for diagnosis. The features stem from a comparison of left and right
breast areas and quantify the bilateral differences encountered. Following this asymmetry
analysis the features are fed to a fuzzy classification system. The approach presented in
section 3.1 is used for generation fuzzy If-Then rules based on a training set of known
cases. Experimental results on a set of nearly 150 cases show the proposed system to work
well accurately classifying about 80% of cases, a performance that is comparable to other
imaging modalities such as mammography.
Thermograms for breast cancer diagnosis are usually taken based on a frontal view and/or
some lateral views. In our work we restrict out attention to frontal view images (we show
an example in Fig. 6.8 ). As has been shown earlier an effective approach to automatically
detect cancer cases is to study the symmetry between the left and right breast (Qi, Snyder,
Head, and Elliott, 2000). In the case of cancer presence the tumor will recruit blood vessels
resulting in hot spots and a change in vascular pattern and hence an asymmetry between the
temperature distributions of the two breast. On the other hand, symmetry typically identify
healthy subjects. We therefore follow this approach and segment the areas corresponding
to the left and right breast from the thermograms. We then convert the breast regions to
a polar co-ordinates representation as it simplifies the calculation of several of the features
that we employ. A series of statistical features is then calculated all of which are aimed to
provide indications of symmetry between the regions of interest (i.e. the two breasts). In
the following we briefly characterise the features we employ.
Basic statistical features
Clearly the simplest feature to describe a temperature distribution such as those encoun-
tered in thermograms is to calculate its statistical mean. As we are interested in symmetry
features we calculate the mean for both sides and use the absolute value of the difference
of the two. Similarly we calculate the standard temperature deviation and use the absolute
difference as a feature. Furthermore we employ the absolute differences of the median tem-
perature and the 90-percentile as further descriptors.