Opportunities and Constraints
Vesselin Gueorguiev, Ivan Evg. Ivanov
Technical University Sofia, blvd. “Kliment Ohridski” 8, Sofia, Bulgaria
Desislava Georgieva
New Bulgarian University, Sofia, Bulgaria
Keywords: Image perception, Medical images, X-ray images.
Abstract: The aim of this paper is to present part of a research work oriented to selection and usage of low-cost
imaging devices for Telemedicine and e-Health. The results presented are oriented to increase the usability
of digital photo cameras as a device for creating and/or digitalizing medical images. All activities are part
of a DAPSEpro project “Medical data acquisition, processing and collection for e-Health solutions”.
The aim of this paper is to present current results of
the DAPSEpro project “Medical data acquisition,
processing and collection for e-Health solutions” in
the part oriented to selection and usage of low-cost
imaging devices and following image processing for
the needs of telemedicine and e-Health. The present
work is the result of the partnership between the
Technical University of Sofia and the Medical
University Sofia (clinics of nephrology and
Under the DAPSEpro project one of the initial tasks
was in the field of “digital libraries” and especially -
the need for digitalization and archiving of patients’
old X-rays. This provoked the development team to
design and implement a technology to solve this
Basic requirements and limitations for the
fulfilment of this task were:
the ability to obtain 'acceptable' images from all
size classical X-ray pictures such as old X-rays,
badly preserved X-rays, or ones made on a
calibrated older generation machines
very limited financial resources for new
a quick and inexpensive image digitalization
process (Marketing research envisaged similar
services in prices between 5-15 USD depending on
the size and quality due).
Two possible solutions were investigated:
usage of existing scanners for X-ray pictures or
scanners for semitransparent images
design of specialized low cost system for X-rays
images digitalization.
To use existing scanners was impossible for the
following reasons:
scanners, which can be used for all sizes of X-
rays at the required quality and speed, are too
reasonably priced scanners have a number of
o the scanning process is very-to-extremely slow
o any high quality scan needs unacceptably long
processing time
Gueorguiev V., Ivanov I. and Georgieva D..
DOI: 10.5220/0003192906080613
In Proceedings of the International Conference on Health Informatics (HEALTHINF-2011), pages 608-613
ISBN: 978-989-8425-34-8
2011 SCITEPRESS (Science and Technology Publications, Lda.)
o pre-view and real scan are two different phases
and both are of long duration
if the picture is bigger than 420x297 mm (format
A3) in any dimension the following problems
o The picture has to be scanned in sections. This
requires adjustment of the picture section by
section to the axes of the scanner for the needs of
the stitching process. In practice this is not
absolutely possible or is very time-consuming,
requires skills and experience. In this case
stitching is mandatory. Because the scan process
is not automated stitching cannot be automated
o In case when the top optical plane of the
scanner is under the top end of the scanner case
processing images bigger than A3 is impossible
because they have to be bent to follow the top
surface shape.
only few scanners include a light intensity
control function
a selected target zone of the picture cannot be
zoomed optically.
These observations directed our efforts to design
and implement a specialized scanning machine. The
first implementation used as a basis for new research
was a successful project implemented 10 year ago,
and it is still in operation in the Institute of
Orthopedics of the Medical University Sofia - a
system for handling X-ray images. Under this
project, together with colleagues from Sofia
University a system (a scanning device and software
for image processing) was designed and developed.
It is based on a digital video camera with high
optical zoom (20x) and black and white sensor with
an 800000 real pixels matrix. The success of this
solution proved the feasibility of low-cost solutions
based on modern digital photo- and video cameras.
The exploitation of the system revealed some
limitations, which were both the result of the quality
of the video cameras from that period, and the
peculiarities of the use by physicians:
Influence of the parameters of the optical system
and sensor matrix on the usability of the system,
differences in the perception of a digitalized X-ray
image and a classical X-ray film from medical point
of view; the limitations of the human eye and related
problems to achieve realism. These observations
accelerated our work significantly.
A major problem in using computers for image
visualization is the characteristics of the human
vision as a system for perceiving visual information:
in many aspects the human eye is an extremely
sensitive sensor, while for other things it is a very
imprecise one. This requires the use of special
computer visualization techniques targeted to the
application area: while with special effects for the
cinema the aim is to "trick" the eye while in
scientific visualization the goal is to increase the
perception and understanding of information.
One of the specifics of digital images as a basic
information source is that they have to be visualized
on many different types of output devices or media.
In present medical applications usually these are
different types of monitors and less often
reproduction on paper or film.
The sequence “image digitalization – image
visualization” generates new problems. Some of
them are:
The terms “color space” and “color model”: the
color space represents the set of possible colors to
reproduce; the color model represents a method of
presentation of colors by a group of primary colors.
The color space of the human eye (known as CIE
1931 x,y chromaticity space – see Figure 1) is much
greater than that of any input and output digital
device. In addition, the color space of the digital
photo and video cameras (known as Adobe RGB) is
also greater than that of computer monitors (known
as sRGB). Additional sources of errors are
imprecisely (and/or wrongly) calibrated devices
(white point temperature characteristic). The white
point is the point of the color space and its incorrect
value moves along the line in the chart, i.e. moves
the device color space within the human eye
perception area. This changes the perception of
purity and color saturation which is inadmissible
when the color is the main source of information in
the image.
One of the main problems in the image
perception is the perception of 'visual weight'. This
allows the perception of some parts and the
suppression of others, according to their location in
the image and the surrounding areas. For medical
- Opportunities and Constraints
imaging this is particularly important because black
and white images are widespread: X-rays,
ultrasound and similar. An example is shown in
Figure 2 – scanned X-ray image and its negative.
The human eye has an exponential law of
perception change depending on light intensity, i.e.
the sense of change for a color close to black and
another close to white is associated with a very big
difference in the change of intensity for both colors.
This could seriously change the perception of B/W
images particularly on the LCD-monitors, such as
laptops, as a function of the slope of the monitor.
For example, when one needs to observe white
details against a black background (figure 2a) the
optimal position is close to the perpendicular to the
view direction, while the opposite case (figure 2b)
requires a noticeably greater slope.
The property of human vision, called
"Approximate color consistency", is another source
of very serious difference between classical
examination and the use of digital images. In nature,
when a human perceives two light sources as
equally white and they illuminate both sides of one
and the same object, the perception is the same color
of the two object sides. Shooting the same object by
digital camera has a different result: the color of the
object depends on the temperature of the light
source. The property of human vision, called
"Approximate color consistency", is another source
of very serious differences between direct
observation and the use of a digital image. In
medical applications this problem can be seen best
in the imaging and visualization of human skin.
Figure 1: CIE color space.
Modern digital photo cameras have a very high
potential, but the selection of an appropriate
apparatus is not a trivial task. It is necessary to
examine several key features and their mutual
influence on the shooting process. It is not rare that
individual characteristics of a good, but not user-
friendly digital camera require professional
photographic skills. It should be noted that the basic
requirements according to target area - color images
obtained for imaging (e.g. for the purposes of
dermatological remote diagnostic activities) or
digitalization of existing images are different.
When using color images there are two essential
characteristics: white balance and camera color
distortion. White balance directly affects the color
change because the temperature of the light source
affects the location of the ‘white point’ on the
“Equal Energy Curve” in color space. An incorrect
setting can result in the following side effects:
decolouring, bluish-green or reddish areas or spots.
These effects increase when shooting human skin
because of natural phenomena known as 'subsurface
scattering' and 'caustics'.
Color distortion is a feature of the cameras,
showing how natural color is perceived by the
camera when lighting is correct, i.e. shows a
replacement of natural color with a color from the
camera color space:
digital cameras have completely different
characteristics, resulting from the combination of
lenses and an optic system, a sensor matrix and
embedded digital algorithms
there are no cameras with equal color distortion
for different shooting modes for both single
shooting and video
the setting for the film sensitivity (ISO value)
also affects the color distortion.
When the apparatus is used as a scanner the key
characteristics are: optical distortion (barrel or
pincushion); gray scale accuracy; sharpness, low
contrast details and resolution; embedded noise
reduction algorithms.
The availability of optical distortion above 1% is
a limitation in the following cases:
o the use of images for measurement of lengths,
angles, distances, curves;
HEALTHINF 2011 - International Conference on Health Informatics
o very big X-ray pictures and/or need to increase
the depth of the digital image requires scanning
in parts: in this case optical distortion prevents
the final stitch procedure;
o if a low level of distortion results from digital
processing by camera firmware no more than 2/3
of the image can be suitable for measurement or
stitch procedures.
Digital cameras use a 3-channel (R, G, B) sensor
matrix. For B/W images these devices use digital
algorithms converting R-, G- and B-values for each
pixel in the intensity (gray colors): the type and
quality of the embedded algorithms specify the
correctness of the gray palette. In different shooting
modes and under different conditions, the gray
palette accuracy varies because the sensitivity of the
matrix color channels is changed differently.
Digital cameras with excellent algorithms for
noise removal are suitable for laymen but from a
technical point of view their algorithms may distort
the realism of shooting images: some algorithms are
too aggressive, and as a result, small size details
disappear from the image.
Sharpness, low contrast details and resolution
determine the level of detail that can be achieved in
different shooting conditions.
As a result of investigation and analysis of many
digital cameras and experiments with a number of
led to resulted to the following set of requirements:
the camera should have a manual mode:
o After a stitch procedure the scanned image will
have the necessary quality and reliability only if
all parts are shot in the same mode and camera
o If color images are used to trace the growth of
some processes, the accuracy of the analysis and
results is determined by the camera settings for
each photo: different settings generate different
color distortion. As a result, one and the same
color in two images will result from two
different natural colors.
the camera must ensure an adequate optical
o scanning activities require minimum 10x
optical zoom, but the optimum is 12x or 14x
based on sensor matrix size
o if color images are used to trace the growth of
some processes the required optical zoom is
determined by the camera: one has to compare
the quality of images taken at close range with
no zoom, and those taken form a greater distance
and zoomed.
Cameras with less than 7 million effective pixels
are useless. The scanning process requires 10 to 12
million effective pixels. Current cameras with more
than 14 million pixels are too noisy.
Preferable cameras are those with least optical
distortion and chromatic aberration (“aberration”
means that some artefacts are added to the image).
Cameras with weight less than 300 g require
much cheaper carrier mechanics for some activities.
A significant advantage is the ability to control and
manage the camera firmware externally. This
provides much greater flexibility and expands the
usability of the apparatus for the purposes of
Telemedicine and e-Health.
Two classes of digital photo cameras are used for
the development of camera-based scanning devices:
“Compact Super Zoom” (manual mode, 12x optical
zoom, 12 M effective pixels, less than 1% optical
distortion, <300 g) and „Compact” (manual mode,
4x optical zoom, 7 M effective pixels, ~ 1% optical
distortion, ~ 220 g). Cameras of the “Compact Super
Zoom” class have provided much greater
opportunities for developing devices which will
substantially expand the potential applications of the
system. All results described below are based on the
use of devices with this type of camera.
When using a digital camera as 'camera'
(creating images from nature) some new possibilities
are discovered:
It is possible to create a model of image
transformation which allows the use of different
shooting settings when color images are used to
trace a growth. So far experiments have been
conducted only with the basic camera, but future
plans are to extend them to other digital photo
A method for using the captured image as a basis
to create a pseudo-3D image was developed. This
may improve the diagnostic possibilities when
additional information for texture is needed.
The ability to manage firmware extremely
enhances the use of the digital photo camera. An
example is the possibility for direct capturing of
static and dynamic stereoscopic images. This will
allow remote obtaining of real three-dimensional
- Opportunities and Constraints
The results of using cameras as a scanner device
demonstrate the usefulness of this approach and also
that scanning with a resolution below grain size of
X-ray film is no longer a problem. This increases the
useful information, allowing digital processing to
eliminate errors in the final images (figures 3 and 4).
For X-rays having one of the following defects
(see figure 5) the developed technology significantly
increases perception (figure 6):
X-ray images with overexposure
X-ray images with underexposure
Incorrectly treasured up X-ray images
A badly calibrated X-ray machine: overexposure
and underexposure areas in one and the same image.
The images prepared and used for the present paper
were the ones used for diagnostic purposes and
served to confirm diagnoses made by other means.
Additionally it was confirmed that personnel with
lower qualification can easily interpret images.
Finally it was confirmed that images with very low
quality can be processed and interpreted. This
demonstrates the great potential of digital photo
cameras for Telemedicine and e-Health applications.
This work is funded by Bulgarian NSF under
D002/113-2008 and DRNF02/3-2009 projects.
Dallas, William J. 1990 A digital prescription for X-ray
overload, IEEE Spectrum
Preim,B., Bartz, D., 2007 Visualization in medicine,
Morgan Kaufmann, ISBN 978-0123705969
Gallagher, Richard S., Computer Visualization: Graphics
Techniques for Engineering and Scientific Analysis,
Solomon Press, CRC Press LLC, ISBN 849390508
Stavitskii R.V. at al., 2006 An approach to decreasing
dose load in prophylactic X-ray examination,
Biomedical Engineering, Volume 40, Number 4 / July,
2006, pp. 191-193
a) positive (native) image
b) negative image
Figure 2: The “visual weight” can substantially change the
perception of details.
a) b)
c) d)
Figure 3: Some examples of image quality of scanned by
digital photo camera X-ray image.
HEALTHINF 2011 - International Conference on Health Informatics
Figure 4: Different levels of details for scanned with
digital photo camera X-rays.
Figure 5: a) X-ray image with overexposure. b) Badly
calibrated X-ray machine.
Figure 6: a) Scanned X-ray image. b) details that can
be obtained after processing.
- Opportunities and Constraints