An angiogram is an X-ray image that uses fluoroscopy to take pictures
of the blood flow within an artery. Angiogram images are of great significance
in explaining blood vessel problems such as aneurysms, narrowed areas
and blockage caused by blood clots or the buildup of fat and calcium deposits.
However, the acquired angiogram images are inadequate to explain vessels
with low contrast. The contrast problem with an Angiogram image is due
to bodily tissues, such as bones, that create shadows on X-ray image.
The main limitations on acquired angiogram images are the overlap of non-vascular
structures and this explains the fact that small vessels with low contrast
are hardly visible (Fig. 1). For facilitating, diagnose
arterial disease and angioplasty or bypass surgery it is essential to
enhance these angiogram images (Alejandro et al., 1998) to an acceptable
Conventionally grayscale domain is used to enhance there angiogram images.
Grayscale images are typically composed of shades of gray, varying from
black at the weakest intensity to white at the strongest. Human beings
can only distinguish a few dozen gray level values (Abidi et al.,
2006), therefore, grayscale enhancement of the angiogram image pose an
On the other hand, human beings can differentiate thousands of colors;
so we claim that the use of color for human interpretation could improve
the identification of different objects. Pseudo coloring of grayscale
images is a typical process used to increase the visual quality of a given
image. A pseudo-colored image is derived from a grayscale image by mapping
each pixel value to a color according to a table or function.
||An ordinary angiogram image
Pseudo color processing is an enhancement
technique that accentuates certain features that are essential for a given application
but hidden with low contrast otherwise. However, this whole process is feasible
once we can identify some characteristic associated with those hidden features.
Then exploiting these characteristic, one can provide pseudo-colors. In present
angiogram case, we identify the width of the vessel as our parameter for providing
false colors. Therefore, a pre-processing step is adopted to capture various
vessels with variety of width. Once captured, the vessels with a range of width
are further enhanced by providing them with suitable colors. The resulting angiogram
image provides enhanced perceptual range.
One of the results is shown in Fig. 2 which clearly
shows that pseudo color processing has been successful in enhancing low
An Angiogram image enhanced using pseudo coloring clearly
increases the visual quality of low contrast vessels
Pseudo coloring of grayscale images can significantly improve the ability
to detect weak features, structures and patterns in an image by providing
image details that would not be noticed otherwise (Troccaz et al.,1997).
A better qualitative overview of complex data sets will then assist in
identifying regions of interest for more focused quantitative analysis,
by making similarly joined areas in the scene more distinguishable (Abidi
et al., 2005). The problem in pseudo color processing is assigning
a color to a particular gray level intensity. To extract information present
in the angiogram images different techniques are followed. Some of them
work at a fixed scale and some of them use the non linear combinations
of finite differences.
Intensity level slicing: Intensity level slicing is a particular
kind of grey level mapping where specific ranges on the input image are
mapped to easily seen values in the output image as shown in Fig.
Grey level transformation: Grey level slicing is another way of
coloring grey level image. This approach performs three independent transformations
on the gray level of any input pixel (Czerwinski et al., 1999).
The three results are then fed separately into the red, green and blue
channels of a color television monitor. This method produces a composite
image, whose color content is modulated by the nature of the transformations
on the gray-level values of an image. The results of grey level transformation
are in shown in Fig. 4.
One of the major drawbacks associated with intensity slicing is the simple
fact that it entirely ignores in the contents present in the image. However,
Pseudo color processing is primarily done to enhance certain features
in the image.
||Angiogram image using intensity slicing false contouring effect
An Angiogram image using grey level transformation absorbs
low contrast vessels due to blind color transformation
In Grey level transformation the vessels are blindly colored.
Hence, the low vessel structures are absorbed. The important features
present in angiogram images from medical point of view are vessels. Logically
the first step should be to locate these vessels and then color them accordingly
to get an enhanced angiogram image.
We applied pseudo color to an angiogram image on the basis of scale.
Vessels are colored in such away that they are differentiated well and
the visual quality of an angiogram becomes better then the gray scale
enhancing technique. While if the angiogram image is colored blindly,
then the vessels are not differentiated well. Therefore, some criteria
must be defined on the basis of which we give the colors, so that these
vessels could be enhanced to better differentiation.
Angiogram image colored using HSI model. The details
of low vessel structures are more precise
As by Abidi et
al. (2005), they have applied HSI model on X-ray luggage images and found
that it is superior as compare to RGB-based coloring technique. Hence,
we also claim that if HSI model is efficient in coloring X-ray luggage
images, it will also be adequate in coloring angiogram images to locate
the small vessels (Fig. 5).
In HSI model each color is represented by hue, saturation and intensity
component. Hue is a color attribute that describes a particular wavelength
in the spectrum. Saturation is the measure of the degree to which a pure
color is diluted by a white light. Intensity represents how little black
has been added to the color. The intensity component is decoupled from
the color information in the image and the hue and saturation components
are intimately related to how humans perceive color. These features make
the HSI model suitable for developing image processing algorithms for
human interpretation (Fig. 6).
In the approach followed by Alejandro et al. (1998), they have
conceived vessel enhancement as a filtering process that searches for
geometrical structures. They have analyzed the local behavior of an image
L, by its Taylor expansion.
This expansion approximates the structure of an image up to second order.
∇0,s and H0,s are the gradient vector and
Hessian matrix of the image computed in x0 at scale s.
||(a) The second order derivative of a Gaussian kernel
of the range (-s, s). (b) The second order ellipsoid where λ1
(a) image extracted when s = 1, (b) image formed when the value
of s = 3. It shows different vessels from the same image for different
values of s
By analyzing the second order information, we can say Hessian justifies
the perspective of vessel detection. By using the second derivative of
scale s (Alejandro et al., 1998) generates a kernel that measures
the length and the width in the region (-s, s) Fig. 7
describes the length of the ellipsoid while λ2 describes
(a) An original angiogram image, (b) Angiogram image
colored using cosine color and HSI model. (c) and (d) angiogram image
colored using transparent colors in addition to HSI model. The results
retrieved in (c) and (d) are more detailed and precise as compared
to (a) and (b)
Eigen-value analysis is used to extract the principle directions that
can be used to decompose the local second order structure of the image
(Alejandro et al., 1998). On various s values we get different
images within vessel that matches the value of s (Fig. 8).
These vessels are then colored using the following formula:
For further mathematical detail see (Alejandro et al., 1998).
But there are some overlapping vessels inside other vessels in Angiogram
images. The information about these vessels might be lost even in HSI
model (Czerwinski et al., 1999). Therefore, some other mechanism
must be applied with HSI Model in order to retrieve all information. To
extract hidden information we have three different method Cosine color
maps, Exponential color maps and transparent color maps (Rafael et
al., 2004). Amongst all these the transparent color maps produce better
results. All these results are discussed in the next section. We apply
transparent colors in addition to HSI. The results retrieved from the
combination of HSI and Transparent colors were more effective and detailed
as shown in Fig. 9.
In present study, basically we extract the vessel structure and then
on the basis of different scale we apply different color transforms, to
the angiogram images which will further enhance the angiogram image. Further
we use the transparent colors for best visualization of the overlapping
vessel structures. In the future we will expand our database by applying
this algorithm on furthermore images to make our model more robust.