PULAB

Computational-Intelligence Aided Management, Diagnosis,

Teleassistance and e-Learning of Pressure Ulcers

Laura Morente

Escuela Universitaria de Enfermería, Diputación Provincial de Málaga, Málaga, Spain

Francisco J. Veredas, Héctor Mesa, Enrique Morris

Departamento de Lenguajes y Ciencias de la Computación, Universidad de Málaga, Málaga, Spain

Keywords: Pressure Ulcer, Computational Intelligence, Computer Vision, Teleassistance, e-Learning, Nursing

Informatics, Software, Collaborative Diagnosis.

Abstract: The pressure ulcer is a clinical pathology with high prevalence rates, which involve high costs for the

Health systems. The health promotion carried on these lesions, as well as the prevention, suitable evaluation

and correct treatment, have become effective indicators of the quality of health assistance. PULAB

(Pressure Ulcer LABoratory) is a computational tool that enables remote management, diagnosis and

monitoring of pressure ulcers, which include digital images of the wounds. This teleassistance software

gives support to the collaborative work of multiple clinical experts to concurrently evaluate the pressure

ulcers by reaching consensus on each particular case, based on the effective analysis of automatically

segmented and tissue-labeled images of the wounds. In the current phase of our research project an e-

learning module for pressure ulcer diagnosis education is being designed, which will turn this software into

a valuable pedagogical tool for pressure-ulcer-management training for undergraduate students and

professional clinicians.

1 INTRODUCTION

The European Pressure Ulcer Advisory Panel

(EPUAP) defines a pressure ulcer (PU) as an area of

localized damage to the skin or its underlying tissue

caused by pressure, shear, friction or a combination

of these factors (EPUAP, 1999; Gawlitta et al.,

2007; Tsuji et al., 2005). The prevention, care and

treatment of the PU pathology involve high costs for

private or state health systems and have important

consequences for the health of the population,

especially for elderly citizens. PU prevalence rates

vary significantly among different environments of

health assistance. Several studies carried on

populations of elderly patients with home assistance

have shown prevalence rates that fluctuate between

12.7% and 15.1% (Bours et al., 1999; Woodbury &

Houghton, 2004); on the other hand, in acute units

the prevalence data found are even more variable

and range from 7% to 33% of the population

analyzed (Gunningberg, 2004; Melotti et al., 2003;

Tannen et al., 2004); finally, higher prevalence rates

were observed in studies carried out on units of

long-term hospitalized patients (Woodbury &

Houghton, 2004; Horn et al., 2002).

The precise evaluation of PUs is a fundamental

task for diagnosis, monitoring of healing evolution

and making decisions on care and pharmacological

treatment interventions. Precise evaluation and

monitoring of the PU could be achieved whether all

the tissues present in the wound or surrounding areas

are accurately measured and precisely registered

(see (Edsberg, 2007) for a complete and systematic

review of PU histology). Following this strategy, in

(Veredas et al., 2010) the same authors of this

current paper presented a computational tool for

automatic segmentation and tissue detection on PU

digital images. This software was based on a hybrid

system which uses Computational Intelligence (CI)

techniques, neural networks and Bayesian classifiers

to be precise, to automatically classify significant

394

Morente L., Veredas F., Mesa H. and Morris E..

PULAB - Computational-Intelligence Aided Management, Diagnosis, Teleassistance and e-Learning of Pressure Ulcers .

DOI: 10.5220/0003275003940398

In Proceedings of the International Conference on Health Informatics (HEALTHINF-2011), pages 394-398

ISBN: 978-989-8425-34-8

 2011 SCITEPRESS (Science and Technology Publications, Lda.)

B C

Figure 1: Automatic tissue recognition on a PU digital image. Picture A shows the results from the automatic region

segmentation of a PU image using the mean-shift procedure and region growing. For B, a group of expert clinicians have

labeled each one of the regions resulting from the segmented image in A. Picture C shows the automatic labeling done by

the CI system based on neural networks, Bayesian classifiers and heuristics. The tissues shown in B and C have been

labeled with a number that represents: 1: skin; 2: healing tissue; 3: granulation tissue; 4: devitalized tissue; 5: necrotic

tissue. Regions labeled with a same tissue type have been given a same pseudocolor for the sake of clarity. (Figures

included here with permission of (Veredas et al., 2010)).

regions from segmented wound images, obtaining

high precision rates as results.

These CI techniques for automatic tissue

detection have been incorporated into the core of

PULAB tool, which has been designed for the

registering, monitoring, the collaborative expert

evaluation, teleassistance and continuing education

on PUs. At present, PULAB tool is being enriched

with the incorporation of a new e-learning module

designed for education on

PU diagnosis for

undergraduate students as well as for continuing

education for professional clinicians. This e-learning

software will be validated during the next few

months by using it in education and competence

acquisition of Nursing university students.

2 TISSUE RECOGNITION

PULAB tool for management, collaborative

evaluation and teleassisance of PUs is internally

composed of an integrated core based on the CI

strategies and machine learning techniques that have

been developed by the same authors of this current

paper. This software enables automatic segmentation

and precise tissue detection on PU images (Veredas

et al., 2010). Image segmentation on this PU images

is arranged by means of the mean-shift segmentation

method (Comaniciu & Meer, 2002). In figure 1-A, a

typical PU image has been segmented by means of

the mean-shift procedure.

Table 1: Efficiency rates from automatic tissue detection

on 113 PU images. (Data obtained, with permission, from

Table VI in (Veredas et al., 2010)).

Sensitivity Speci-

ficity

Accuracy

Necrotic 86.3 % 98.5 % 98.2 %

Devitalized 67.4 % 95.7 % 93.3 %

Granulation 82.7 % 94.7 % 92.6 %

Healing 59.9 % 91.1 % 85.4 %

Skin 85.2 % 91.0 % 87.9%

GLOBAL 78.7 % 94.7 % 91.5 %

In the table 1, efficiency rates are shown from the

results obtained in the automatic classification of the

tissues present in a set of 113 testing PU images (not

previously “seen” by the machine learning system).

As can be deduced by observing this table, the

automatic classification system based on CI

techniques and used by PULAB for PU evaluation

shows high efficiency rates, not only in the

classification of each particular tissue type, but also

in global terms.

3 THE PULAB TOOL

PULAB is a multiuser teleassistance tool that makes

possible the recording of clinical data and the

collaborative evaluation of PUs. This software tool

has the main purpose of increasing the accuracy of

the diagnosis and the effectiveness of care and

treatment interventions. Moreover, this software

enables the management of each particular wound

case by means of registering digital pictures in the

system and storing contextual clinical data

PULAB - Computational-Intelligence Aided Management, Diagnosis, Teleassistance and e-Learning of Pressure Ulcers

395

Figure 2: PULAB’s graphical user interface for manual tissue labeling of a registered PU image.

associated with the PU. The system’s users, i.e. the

clinicians and health professionals in general, can

also use PULAB to manage PU series of wounds

grouped by different criteria. Once a PU image has

been uploaded to the system (with its clinical

associated information enclosed), it is immediately

processed by the CI subsystem for region

segmentation and automatic tissue detection. From

that moment on, the user counts on an initial

automatic diagnosis and can look up this automatic

evaluation done by the system to do manual

adjustments on the tissue classification in order to

get a final refined diagnosis. These manual

improvements on the tissue classification are done

with the aid of a friendly graphical user interface

that facilitates the task of region-of-interest selection

and tissue identification. Finally, the system

provides a useful module to efficiently create and

manage collaborative work groups of clinicians who

can share their opinions and reach agreements on

diagnosis of each particular PU case.

PULAB has been designed following a

methodology that is based on a client/server model.

PULAB’s user interface has been developed in Java

and is accessible at the url https://itaca.lcc.uma.es/

ulceras/pulab/launch.jnlp by means of Java Web

Star®.

The main modules of PULAB are the following:

• User-authentication and session-control module:

provides the control of authenticated users,

giving the basis for the management of

collaborative work groups.

• PU-series module: makes possible the creation

and management of series of PUs grouped by

heterogeneous criteria (temporal series of a

same patient, grouping by grade or presence of

different tissue types, grouping by anatomic

location, etc.).

• Collaborative-group module: enables the

management of groups of users to share PU

series, evaluations, diagnosis or decisions on

interventions on the wounds.

• System-record module: provides the

management and controls notification of actions

launched by the systems, such as registrations

and deletions of users, groups, series, etc.;

termination of processes of image segmentation

or tissue detection by the system, etc.

• PU-visualization module: enables the

interaction between the user and the system to

visualize the PU images and navigate on the

segmented regions and classified tissues.

• Labeling module: provides the tools for manual

labeling of the regions resulting from the

automatic segmentation of the images; this

module supplies the user with the necessary

HEALTHINF 2011 - International Conference on Health Informatics

396

tools to manually classify, in a easy and friendly

manner, the segmented regions into the different

tissue types (see the figure 2 for an screenshot

of the user’s interface of the labeling module on

an example of a real PU image).

3 DIAGNOSIS E-LEARNING

An effective strategy to reduce the use of

pharmacological treatments of not-validated benefits

and to homogenize clinical interventions could be to

improve the education of both Health undergraduate

students and professional clinicians. Traditional

education on PU pathology suffers from some

weaknesses that could put the efficacy of the

learning process at risk: on one hand, students

behave usually as merely passive actors during the

learning process, and their interests and motivations

are usually very poor; on the other hand, in clinical

practice, a high variability in the learning procedure

is usually generated. Considering these two issues

above, and as a strategy to improve the learning

process and also guaranty its efficacy and

homogeneity, the current development of PULAB

tool has the main objective of introducing

Information and Communication Technologies

(ICT) to education on PUs for Health undergraduate

students and professionals (i.e. continuing

education).

Very few studies exist on evaluating educational

experiences with ICT-based tools in the specific

field of education on PUs for professional clinicians.

However, some authors conclude that the

development of tools that make e-learning possible

increases the efficacy of the educational processes

since it reduces the time consumed in the learning

process and improves the accessibility of the student

to that learning (Bolwell, 1993). Furthermore, a

recent study by Beeckman et al. (Beeckman et al.,

2008), could be pointed out which deals with

improving the ability of students in the classification

of different PU types and their differentiation from

those other wounds produced as an effect of skin

humidity.

Considering the main goal of minimizing the

user’s system requirements, PULAB’s e-learning

module is being currently developed using web-

based technologies, which will allow using a simple

web browser to have full access to the complete

functionalities offered by this software tool.

Moreover, for the sake of usability, AJAX

technology is being used in the designing of the

graphical user interface with the major objectives of

building complex controls in the forms, minimizing

the data communication and facilitating the

interaction between the user and the system.

PULAB’s e-learning module is being designed as an

adaptive learning interface, which will enable the

students to receive their education in an manner

adapted to his or her particular background-

knowledge level, this way starting from initial

simpler diagnosis cases and progressively going to

more complex examples of PU evaluations and

diagnosis. Experts and professional clinicians will be

included in PULAB with the profile of “teacher” and

will be able to continuously add new evaluated PU

cases in the database, which can be adaptively

included in the sets of PU samples available for their

students. The teachers will be provided with tools

for designing and managing tests for their students,

in order to evaluate their educational progressions.

Both, the teachers and the students, will be supplied

with statistics tools to monitor the learning evolution

and progressions.

Once the PULAB’s e-learning module had been

developed, these authors will proceed to the

validation of this software as an efficient educational

tool, by means of comparing the educational results

obtained from the application of PULAB with those

outcomes coming from the application of traditional

teaching classes. The initial proposed hypothesis for

this validation phase establishes that the PULAB

tutoring system, as an educational software designed

specifically for adaptive education on PU

management, diagnosis and treatment, would

increase the underlying knowledge and improve the

aptitudes for diagnosis, classification, tissue

differentiation and therapeutic decision-making for

undergraduate Health currently students, in

comparison with those results obtained with

traditional teaching methods on PUs management.

4 CONCLUSIONS

PULAB tool has been developed to make possible

the objective evaluation of pressure ulcers. This

software enables teleassitance as well as the

collaborative work of professional clinicians.

PULAB tool consists of an internal core, hosted in

the application-server, that uses computational

intelligence techniques for image segmentation and

tissue detection, which have demonstrated recently

high efficiency rates when applied to real pressure

ulcer images. Finally, the incorporation of an e-

learning module into PULAB tool will make

possible to have an educational tool which will

PULAB - Computational-Intelligence Aided Management, Diagnosis, Teleassistance and e-Learning of Pressure Ulcers

397

increase the efficacy of the learning process on

students or professionals on pressure ulcer

evaluation, diagnosis and care or pharmacological

interventions. The development and validation of

this e-learning module will be concluded in a few

months. Finally, a remarkable issue to be considered

is the possibility of using this same technology

implemented in PULAB for the evaluation and

diagnosis of other sort of skin wounds that require a

teleassistance management similar to the one used

with pressure ulcers and implemented in PULAB.

That could be the case of burn wounds or even

different types of melanomas.

ACKNOWLEDGEMENTS

This project has been supported by the Consejería

de Salud, Servicio Andaluz de Salud, of the Junta de

Andalucía, project id. PI-0502/2009.

REFERENCES

EPUAP, European Pressure Ulcer Advisory Panel., 1999.

Guidelines on treatment of pressure ulcers. EPUAP

Review, 1, 31-33.

Beeckman, D., Schoonhoven, L., Boucque, H., Van

Maele, G., & Defloor, T., 2008. Pressure ulcers: E-

learning to improve classification by nurses and

nursing students. Journal of Clinical Nursing, 17(13),

1697-1707.

Bolwell, C., 1993. Using computers as instructional

technology in the pressure ulcer field. Decubitus, 6(4),

20-25.

Bours, G. J., Halfens, R. J., Lubbers, M., & Haalboom, J.

R., 1999. The development of a national registration

form to measure the prevalence of malnutrition in the

netherlands. Ostomy Wound Manage, 45, 28-40.

Comaniciu, D., & Meer, P., 2002. Mean shift: A robust

approach toward feature space anaylsis. IEEE

Trans.Pattern Anal.Mach.Intell., 24, 603-619.

Edsberg, L. E., 2007. Pressure ulcer tissue histology: An

appraisal of current knowledge. Ostomy/wound

Management, 53(10), 40-49.

Gawlitta, D., Li, W., Oomens, C. W. J., Bader,Frank

P.T.Baaijens and Dan L., & Bouten, C. V. C., 2007.

The relative contributions of compression and hypoxia

to development of muscle tissue damage: An in vitro

study. Annals of Biomedical Engineering, 35(2), 273-

284.

Gunningberg, L., 2004. Risk, prevalence and prevention of

pressure ulcers in three swedish. Journal of Wound

Care, 13(7), 286–-290.

Horn, S. D., Bender, S. A., Bergstrom, N., Cook, A. S.,

Ferguson, M. L., Rimmasch, H. L., et al., 2002.

Description of the national pressure ulcer long-term

care study. Journal of the American Geriatrics

Society, 50(11), 1816-1825.

Melotti, R. M., Fortuna, D., Chiari, P., & Cavicchioli, A.,

2003. Prevalence and prevention and treatment

modalities for pressure sores. Epidemiologia e

Prevenzione, 27(3), 141-146.

Tannen, A., Dassen, T., Bours, G., & Halfens, R., 2004. A

comparison of pressure ulcer prevalence: Concerted

data collection in the netherlands and germany.

International Journal of Nursing Studies, 41(6), 607-

612.

Tsuji, S., Ichioka, S., Sekiya, N., & Nakatsuka, T., 2005.

Analysis of ischemia-reperfusion injury in a

microcirculatory model of pressure ulcers. Wound

Repair and Regeneration, 13(2), 209-215.

Veredas, F., Mesa, H., & Morente, L., 2010. Binary tissue

classification on wound images with neural networks

and bayesian classifiers. Medical Imaging, IEEE

Transactions on, 29(2), 410-427.

Woodbury, M. G., & Houghton, P. E., 2004. Prevalence of

pressure ulcers in canadian healthcare settings. Ostomy

Wound Management, 50(10), 22-38.

HEALTHINF 2011 - International Conference on Health Informatics

398