|
PROJECT IST 1999–20226 DERMA
Prototypal system – based on image acquisition, 3D and chromatic
reconstruction –
for the objective monitoring of skin lesions
Consortium Composition
- ENEA – Ente per le Nuove tecnologie,
l’Energia e l’Ambiente, UDA Agenzia per lo Sviluppo Sostenibile
- CGS –
Sas di Coluccia Michele & C. Sas – Sett. Ricerca Scientifica
Avanzata
- SIDI – Sistemi Digitali Srl
- DEPI – Department of Dermatology,
University of Pisa
- DECO – Copenhagen Wound Healing
Center
- DEOX –
Oxford Wound Healing Institute
- DEDE –
Unit of Health Economics and Technology Assessment in Health Care, Center
for Public Affairs Studies, Hungary
Contact information
Dr Marco Romanelli
Department of Dermatology
University of Pisa
Via Roma, 67
56126 Pisa, Italy
Tel: +39 050 992436
Fax: +39 050 551124
E-mail: m.romanelli@med.unipi.it
Introduction
Over the past few years, a number of highly developed objective and non-invasive
techniques for wound assessment have been brought into use. These techniques
are important research tools in investigating the different phases of
wound healing and in determining therapeutic effects. Since non-invasive
measurements can be performed repeatedly on the same wound site, changes
during wound healing may be monitored. Our knowledge about the physical
properties of acute and chronic wounds is likely to increase in the near
future, thanks to this greater interaction with the engineering field.
Because of its importance in wound healing, the use of non invasive measurements
can be expected to progress with new techniques and methods to be applied
in research and in clinical practice.
The DERMA system enables users to accurately acquire three-dimensional
digital models of various types of skin wounds and also to collect (together
with geometrical data) colour information. The main goal is to monitor
wound evolution, through the use of a series of numerical acquisitions
and objective measurements. The system therefore does not simply manage
digital acquisition, but is equipped with various measuring/analysis tools
and is able to compare scanning.
The DERMA system is equipped with a MINOLTA Vivid 900 laser scanner, which
is used for digitalizing (or scanning) wound shape and colour. All scanner
interface and control functions have been developed within the DERMA system,
in order to have user-friendly equipment (hardware and software). This
has led to a more complex system implementation, but has facilitated use,
since the management of an extremely complex tool, such as a 3D scanner,
has been solidly integrated within the system interface. Since the final
users will be doctors and not computer experts, a user-friendly system
is believed to be a fundamental parameter for its success. Having a single
software tool that intuitively runs both scanner control functions and
acquisition management/elaboration functions is highly important.
1. DERMA Features
1.1 Volume calculation
The perimeter, area and volume of a wound may be measured on the current
scan through a direct manipulation interface. For this reason, suitable
calculation procedures have been defined (which calculate linear measurements,
surfaces and volumes directly on the digital 3D model), as well as interactive
routines that enable the user to specify or delineate the areas where
measurements must be taken.
The part of the interface that visualizes the acquired surface enables
the user to define the wound edge with a semi-automatic procedure: the
operator makes an approximate drawing of the wound perimeter and this
initial detail is refined by the DERMA system through automatic algorithms,
which try to make the perimeter defined by the operator more faithful
to the digitized image profile.
The initial wound perimeter definition is iteratively controlled by the
interface: the interface allows the user to draw a broken line around
the edge by selecting the single points through which the line must pass;
by selecting again the first point given, the edge (polyline) is closed.
It is then possible to manually refine the result achieved by moving the
points or by adding further ones. The first, approximate edge is then
refined through automatic algorithms capable of individuating the wound
edge area, and adding new points to the polyline and/or moving existing
ones.
Once the wound perimeter has been digitally defined, it is possible to
calculate the internal wound surface by using the acquired 3D data. Since
this measurement is carried out on the real wound surface, it also takes
into consideration all creases, protrusions and small fractures inside
the wound itself. For this reason, the measurement may change from one
scan to another although the external wound perimeter remains exactly
the same. This type of information therefore may be considered as a new,
numerically accurate parameter to be assessed during the life of a wound.
With regard to the calculation of the ‘external’ surface and
volume of a wound, it is necessary to assess its original shape in order
to create the missing volume virtually. Since information on the shape
of the skin before the beginning of the pathology is missing, the virtual
reconstruction of the original wound surface must be as user-friendly
and easy to carry out as possible. The system, based on the shape of the
surface immediately outside the perimeter, proposes an interpolating surface
that is continuously connected to the surface outside the wound and to
that covering it. The physician, through the use of simple, mouse-guided
interactive tools, can control the main parameters that guide the construction
of the interpolating surface by moving it, for example, further away from
or nearer to the center (or other points) of the wound.
1.2 Colour segmentation of lesion area
- the user selects a point (seed) inside the
previously defined wound edge and corresponding to the colour to classify
within the 3D model. Automatically, a region-growing algorithm
with absolute colour classification criteria delimits the corresponding
area. The operator may use an automatic extension function that searches
for other seeds that are compatible with the average colour of the first
region by iteratively applying the region-growing algorithm; alternatively,
he/she may manually select another seed in an area thought to have similar
colour features (also in this case, the region-growing algorithm
is applied);
- the user may control the ‘tolerance’ of
the region-growing algorithm step by step (the final step must include
the overall wound);
- the user may repeat the procedure described above
for 5 classifications at most. The next application of the region-growing
algorithm avoids the possible overlapping of areas associated to different
colours by positioning the edge on the point that is chromatically equidistant
between two ‘competing’ seeds. The user may freely assign
a name to each classification obtained – the environment automatically
calculates perimeter, area and percentage with respect to the wound
surface of each classification;
- the classification results are visualized with ‘false
colours’ (corresponding to the average colour of the classified
region) on the 3D wound model (the area outside the edge of the region
simultaneously assumes a neutral colour).
Italian DERMA team visits Oxford for patient evaluation.
Above: DERMA equipment in the Oxford Clinic.
Above: Dr Antonio Magliaro, Dr Francesco Rizzello, Dr George
Cherry
and Dr Georgio Meini outside the Oxford Wound Healing Clinic.
2. User Interface
The graphic interface of the programme has two windows: the largest, on
the right, is used for 2D/3D visualization; the smallest, on the left,
is a dialogue window for managing the greater part of all expected operations.
Programme interaction has a linear structure: the user is guided through
clear ‘operative’ phases, each associated to a clear logical
action (e.g., scan, measurement, data analysis). This division helps systematically
group the various operations within the programme.
By following this setup, the left part of the interface is divided into
different ‘tabs’, each corresponding to a working data or
mode interaction phase.
2.1 Patient selection
The user may consult the list of patients grouped in the database, add
or delete names and correct data already entered. In order to proceed
towards the next step, the user must select a patient he/she wishes to
work on, who will become the ‘current patient’.
2.2 Scanning
After selecting the patient, the user has access to all the previous scanning.
During this phase, the user controls the scanner and may proceed with
a new scan by using the control tools made available by the graphic interface.
The scan may be rejected or, if important, inserted among the list of
the patient’s scans.
During this phase it is possible to recall from the scanning list the
‘current scan’, upon which the user may operate during the
course of subsequent working modes.
An immediate measurement function permits the user to measure distances
quickly and easily (just by clicking on two points); this function is
useful for immediately controlling the acquired area and for extemporaneous
measurements on library scans.
2.3 Analysis
A study of the colorimetric data acquired by the scanner may be carried
out on the current scan: the analysis permits the user to semi-automatically
individuate areas within the wound surface with different skin features
and to carry out measurements. The process is conducted by using automatic
segmentation algorithms that, starting from the image, individuate homogenous
areas with colours similar to a number of basic shades chosen by the user.
Once the areas have been individuated, the real 3D surface extension may
be measured, thanks to the biunique correspondence between each pixel
of the image acquired by the scanner and the 3D surface.
2.4 Comparisons
The operator may compare different scans (and the data associated to them)
carried out on a current patient thanks to the simultaneous visualization
of a number of scans (usually carried out at different times). The interface
enables the user to select two or four scans, which are visualized side
by side so that the wound’s evolution may be monitored over time.
For greater clarity, the associated measurements previously acquired by
the operator may be visualized on the selected 3D models.
It should be pointed out that it is not always possible to go from one
phase to another (for example, it is not possible to carry out a measurement
without selecting the ‘current’ scan); the following diagram
indicates the possible passages between the various system modes.
3. Database Interface
In order to facilitate system use, the database interface is completely
hidden to the final user.
When the programme is started, connections to the scanner and to the database
are made and from this point onwards the whole system is managed by specific
control and interface modules.
4. 3D Scanner Interface
The scanner is easily controlled, so that it may be used by non-specialized
personnel. In most cases the scanner can find the right parameters on
its own, leaving the operator to simply individuate the points.
If necessary, it is possible to control the scanning laser intensity and
focusing distance by hand.
On the right part of the interface the following are displayed:
- black and white scanning image (monitor mode), which
enables users to orientate the scanner and frame the area to be acquired;
- 3D surfaces acquired during scanning;
- colour 2D image acquired during scanning and needed
for dividing the lesion into its various component parts.
5. 3D Graphics
3D data visualization is carried out using Open GL libraries (hardware
accelerated by most video cards). This guarantees good drawing speed for
large-sized 3D models, such as those obtained with high resolution 3D
scanning.
The surfaces acquired are visualized as triangular and illuminated mesh;
visualization may be with or without texture mapping and colour rendering:
the former is useful for analyzing geometrical features, which are weakened
by the colour on the mesh. A simple mouse-run trackball enables the user
to rotate, move and reduce the 3D model allowing an accurate inspection
of the area.
|
