ORGANIC COMPUTING FOR HEALTH CARE SYSTEMS
Possible Benefits and Challenges
Florian Nafz, Frank Ortmeier, Hella Seebach and Wolfgang Reif
Lehrstuhl f¨ur Softwaretechnik und Programmiersprachen, Unversit¨at Augsburg
Universit¨atsstr. 14, 86135 Universit¨at Augsburg, Germany
Keywords:
Organic Computing, multi-agent systems, health-care applications, security.
Abstract:
Todays health care institutions will undergo major changes in the next two decades. The reason for this is the
change of ageing structure in many industrialized countries. In Germany statistics indicate that the costs for
health care systems will at least double per person while the number of contributing, working citizens will
significantly lower. At the same time average life expectation will rise above 80 years. To cope up with this
development adaptations to organization and process of health care are necessary. Typically tasks in stationary
health care can be divided in two groups: task which incorporate direct interaction with the patient (care tasks)
and tasks which focus on logistics and organization (background tasks). In health care it is not desirable and
feasible to reduce efforts in care tasks. So costs and efforts must be reduced within the second group of tasks.
This is possible if new paradigms both in organization and underlying software architecture are applied.
One such paradigm is organic computing. Organic computing aims at systems, which are self-organizing,
self-adapting to new challenges and self-optimize during runtime. Such systems can take away a lot of organi-
zatorial work form the staff and thus allow for more and better care without rising budgets. The paper outlines
the idea of organic computing as well as opportunities and challenges for applying it in the health care context.
1 INTRODUCTION
It is a widely accepted fact, that cost pressure will
make changes in todays health care institutions nec-
essary. Reasons for this development are the chang-
ing ageing structure and the rising life expectation of
many industrial countries’ population. In 2005 there
were in average 31.7 persons older than 65 years for
every 100 persons in working age. This percentage
will rise to 54.2 persons in 2030 and to 70.9 persons
in the year 2050 (Statistisches Bundesamt, 2006). At
the same time the average per year cost for health care
for persons older than 65 years is four times the aver-
age cost for younger people (Statistisches Bundesamt,
2007). These two facts together show that costs will
dramatically rise while the income will stay almost
constant.
It is not feasible to simply cut health care benefits
to reduce costs. Therefore it is necessary to provide
health care more cost efficient. Humans play and
will always play the most important role in caring
for others. So it is not possible and wanted to “re-
move” human staff from the caring process. To keep
the health care systems affordable and survivable it
will be of uttermost importance to support the staff in
all tasks, which are not directly related to the patients
as much as possible.
This can only be done by providing intelligent in-
frastructure and background systems, which support
the medical staff autonomously whenever possible.
A paradigm for construction of such systems is Or-
ganic Computing (VDE/ITG/GI, 2003). It allows for
constructing flexible systems, which dynamically (re-
)configure and autonomously self-adapt to changing
tasks and requirements.
In this paper a brief overview on Organic Com-
puting is given in Sect. 2. Sect. 3 shows chances and
possibilities of applying Organic Computing in health
care on a little example. Open questions for research
and challenges are discussed in Sect. 4. Sect. 5 con-
cludes this position paper.
2 ORGANIC COMPUTING
TECHNIQUES
The complexity of modern systems has grown im-
mensely in the last years. In particular the trend to
replace hardware solutions with software components
has made the software become (a) more and more im-
286
Nafz F., Ortmeier F., Seebach H. and Reif W. (2008).
ORGANIC COMPUTING FOR HEALTH CARE SYSTEMS - Possible Benefits and Challenges.
In Proceedings of the First International Conference on Health Informatics, pages 286-290
Copyright
c
SciTePress
portant for the functionality and (b) more and more
complex to develop. A good example is a modern
(digi-)cam. Five years ago most cameras where using
analogous media to record pictures. Extra function-
alities like (visual) noise suppression or image sta-
bilization where implemented by using specific me-
dia or by adding extra hardware components. Today
such functions are usually implemented in software
and can relatively easy be transfered from one cam-
era to another. This led to very complex software and
this process is now reaching a saturation level. To
further enhance the capabilities of both the system
in general and the controlling software in particular –
new approaches are necessary to deal with the grow-
ing complexity. One such approach is Organic Com-
puting (M¨uller-Schloer, 2004).
The Organic Computing Paradigm. Or-
ganic Computing (M¨uller-Schloer et al., 2004;
VDE/ITG/GI, 2003) is an extension to Ubiquitous
and Autonomic Computing (Kephart and Chess,
2003). The core goal of Organic Computing is,
that future (computing) systems should be able
to dynamically adapt to changes in requirements,
to automatically detect and neutralize component
failures and to continuously optimize themselves for
better performance. Such capabilities are called self-
adaptation, self-healing and self-optimization. The
hope is, that systems with self-x capabilities are by
far superior to standard systems in terms of possible
functionalities, availability, reliability and effort for
maintenance. At the same time self-organization
mechanisms allow for easier construction. As not
every single scenario has to be anticipated at design
time. New scenarios and requirements will be
detected by the system during runtime and self-
adaptation mechanisms will reconfigure the system to
this new challenges. So the systems behave from an
external point of view very much like living beings.
They monitor their environment, reflect upon changes
and adapt to new situations. Therefore such systems
are called “Organic Computing” systems.
It is often useful and possible to cope mecha-
nisms and techniques from biology/sociology for the
design of self-X algorithms and systems so called
bio-inspired algorithms. One example is a (self-
)protection algorithm which function like the human
immune system (Pietzowski et al., 2006). But there
also exists a variety of other algorithms, which are
based on more traditional approaches.
Technically, implementations of Organic Comput-
ing systems often contain one part which is respon-
sible for delivering the wanted/intended functionality
and one part which constantly monitors the environ-
ment and – if necessary controls/changes the func-
tional part. There already exists a broad variety of
methods (Seebach et al., 2007; Richter et al., 2006),
middlewares (Trumler et al., 2004; Trumler, 2006)
and analysis tools (G¨udemannet al., 2007; G¨udemann
et al., 2006) for design, analysis and construction of
such systems.
Application Domains. Organic Computing is typ-
ically applied to software-intensive embedded sys-
tems. Example domains are large networks of sen-
sors, traffic control systems or production automa-
tion. One example could be a vision of the next
generation of production cells. Assume a produc-
tion cell which should process workpieces followinga
given specification. The functional part of this system
may be a set of robots which perform certain tasks
e.g. drilling holes, inserting screws and tightening
screws on workpieces and a set of transport units,
which transport workpieces from station to station. If
new types of workpiece are to be processed, then the
robots must be reprogrammed (or these new work-
pieces must have already been anticipated during sys-
tem design). If a single robot/component of the sys-
tem fails, then production will often come to a stand
still.
If in contrast this system is designed in an or-
ganic way, then each robot will be enhanced with
a supervising (software) component. This software
not only monitors the supervised robot, but also de-
cides/plans what tasks the robot will perform an from
where workpieces are taken resp. where workpieces
are to be placed next. If new workpieces – with a new
specification about how they are to be processed – are
given to the production cell, then the supervisors try
together to find a reconfiguration to achieve the new
goal.
Assume for example a cell that consists of three
robots, which are all capable of drilling holes, in-
serting screws and tightening them and a set of au-
tonomous carts for transportation, then an (initial)
configuration could state that one robot only drills,
one only inserts screw and one only tightens the
screws. The workpieces will be transported in badges
from robot 1 to robot 2 to robot 3. The supervisors
constantly monitor the functionality of the robots and
the type of workpieces which are to be processed. If
now for example the drill of robot 1 breaks, then the
supervisors will search for a new configuration to sub-
sume production. One solution is to let robot 2 do the
drilling, assign robot 1 to inserting screws and change
the transportation routes accordingly. This is called
self-healing. If a a new type of workpiece is to be pro-
cessed; e.g. workpieces now only need holes (and no
screws at all). Then the supervisorwill recognizes this
and reconfigure all robots to drill (self-adaptation)
1
.
The technical basis for such effects is (i) the ca-
pability of the supervisors to reflect their controlled
subsystems capabilities, (ii) to interact/communicate
with each other and to (iii) use a common language
about goals and capabilities. In the last years there
have been big advances in building Organic Comput-
ing systems. There exists now a number of working
middlewares, organic architectures and intelligent re-
configuration algorithms. They have also been suc-
cessfully applied to various technical scenarios. But
there exist hardly any application to systems in which
privacy and individual user-trust play an important
role.
3 ORGANIC COMPUTING IN
HEALTH CARE
How can Organic Computing help for health care?
What are the challenges, which have to be met to
allow for using Organic Computing in health care?
These question can be best answered, if an example
scenario is taken into account.
A possible example scenario may be support and
comfort functionality in future rehabilitation centers.
This scenario is only one selected example of a broad
class of background support systems both inside and
outside the domain of health care.
An Example. The scenario assumes, that during the
next decade patients as well as medical staff and doc-
tors will be equipped with mobile devices, which can
form ad-hoc-networks. The topology of these net-
works will be continuously changing as the device
are carried around by their possessors. Stationary de-
vices like ergo-meters or central services will also be
equipped with wireless communication and can also
participate in the network.
Typically patients in a rehabilitation clinic arrive
with a fixed treatment concept. This concept states
a number of treatments (e.g. physio, massage, etc.)
which have to be applied to the patient in certain
intervals (e.g. twice a week). Periodically often
every two weeks these concepts may be changed
or updated according to the physical state of the pa-
tient. Using the treatment concept and the capacities
of the staff and installations as input data, a central
administration service creates a treatment plan for the
patients. This plan schedules treatments to certain
1
A more detailed description of this example may be
found in (G¨udemann et al., 2006).
time slots and staff/installations. Similarly work plans
for the staff are created. This process is relatively
complex and requires intensive planning. It is usu-
ally done on a weekly basis. The plans are typically
printed and handed out to patients and staff.
This process is performing adequately while there
are no unpredicted events. In reality there typically
occur a lot of disturbances like patients coming to
late to their treatments, illness of staff or malfunction
of some equipment. The consequence is that the plans
must be adapted. This is often not possible or only if
the disturbance can be foreseen (like illness of staff).
As a consequence in most cases treatments will ei-
ther be canceled or at least postponed for several days.
The root of this problem lies in the static nature of this
process and the central planning.
Benefits of Organic Computing in Health Care.
Better results can be achieved if an Organic Comput-
ing systems is implemented for this task. In an or-
ganic approach, the treatment plan will not be calcu-
lated a priori by a central administration. It will rather
evolve by interaction of all agents with each others.
This allows for dynamic reconfiguration and continu-
ous adaptation for optimal solutions. The agents can
much better locally decide if a schedule for treatments
is possible, impossible or borderline. Agents are in
this context the mobile devices of all relavant paritci-
pants. An automatic replanning algorithm could for
example first collect localization information of the
patient and the next treatment‘s position. It will then
consider the physical state of patients or delays of the
treatment staff (for example because of meetings or
emergencies). Together this information will often al-
low for dynamic re-organization of treatment sched-
ules. The consequence is that many disturbances can
be compensated by a little reorganization. The big
benefit is, that this can be done “on-the-fly” without
the need of a central coordination.
An example: assume patient A will not make it
to his next treatment in time, while patient B, who is
scheduled right after A, is already close to the treat-
ment. It is then possible to inform A, B and the
staff and to dynamically switch the patients‘ treat-
ment. Note, that this simple scenario could also be
possible without a computer aided planing system.
But it is clear that the same method can also be used
to make more complex re-schedulings, which may in-
volve multiple changes. This is better than traditional
approaches, because firstly re-planning starts as soon
as a disturbance is detected and secondly it is not nec-
essary for all agents to physically meet at one place
but rather share and exchange information through
the ad-hoc network. Similar algorithms can compen-
sate for delays of staff, broken equipment and many
other disturbances. Furthermore the system could be
used to automatically integrate new patients into the
system by calculating treatment schedules for them.
As a result of later automatic optimization some re-
scheduling of existing schedules could make sense to
improve the overall quality of the planning system.
All the properties described above are typical
properties of Organic Computing systems. In Organic
Computing these effects are called self-organizing
(e.g. autonomously integrating new patients and cal-
culating schedules), self-optimizing (e.g. minimizing
waiting times), self-healing (e.g. compensating for
broken equipment or illness of staff) and self-adapting
(e.g. adapting to changing treatment requirements
and health situation of a patient). So its only a log-
ical step to apply techniques, tools and middlewares
of Organic Computing to the domain of health care.
4 DOMAIN SPECIFIC
CHALLENGES
What are the specific challenges to implementing an
Organic Computing system in health care? The sad
point is: traditional Organic Computing techniques
are not directly applicable to this domain. The prob-
lem is, that most Organic Computing approaches
make two assumptions of the systems and its com-
ponents agents. The first assumption is, that all par-
ticipants/agents share a common goal. The second
one is, that there are no “malicious” parties involved.
This means, all agents work for common interests and
no agent tries to gain only benefits for himself while
causing major drawbacks for all others. This arises a
number of challenge, which must be mastered for suc-
cessfully integrating Organic Computing in the do-
main of health care:
1) Individual Goals
Whenever agents of an Organic Computing system
are personalized to individual persons, they in general
won‘t share a common goal. In the example described
above, agents owned by patients, medical staff and
administration will have different goals. Patients are
interested in treatment schedules, which are compat-
ible with their leisure times, medical employees are
interested in compact working times and the clinic‘s
administration is interested in maximum profits i.e.
no canceled treatments. These conflicting goals have
to be balanced and weighted in a meaningful manner.
It must also be assured, that security mechanisms pro-
hibit individual persons from getting benefits on the
cost of drawbacks for multiple others. So it will be
a challenge for the next generation of Organic Com-
puting algorithms to be able to cope with individual
goals and requirements.
2) Sensitive Data
Personal data is sensitive. Medical data is even more
sensitive. Therefore privacy and security is a prime
requirement for every system, which handles data in
a medical care scenario. On the other hand, the Or-
ganic Computing paradigm strongly relies on inter-
action and exchange of data between system com-
ponents/agents. New and optimal configurations are
computed at runtime jointly by all agents. This is only
possible, because of information exchange and reflec-
tion of the exchanged data. For health care it will be
a challenge to adapt Organic Computing algorithms
such that only for reconfiguration and planning nec-
essary data is exchanged and that sensitive data can
be kept private wherever and whenever possible.
3) Behavioral Guarantees
Health care is a highly critical domain. Whenever a
system is allowed to autonomously make decisions,
which have effects on the medical treatment of a pa-
tient, it must be assured that these decisions may
never pose a thread to the health of the patient. This
problem can be solved by applying formal methods
for analyzing the systems. They allow for rigorously
proving, that the system will always fulfill some be-
havioral guarantees. For Organic Computing such
analysis is much more difficult, because it can often
not be anticipated at design time in which environ-
ments/scenarios the system will eventually be asked
to (self-)adapt to. Therefore new analysis methods
must be developed, which allow for giving behavioral
guarantees for Organic Computing systems.
4) User Trust
This is possibly the most important challenge. Or-
ganic Computing systems can achieve a lot of bene-
fits from a global point of view. The benefits result
from their ability to self-organize and self-adapt. Im-
provements become typically visible on system wide
performance metrics. For local agents, (re-) configu-
ration is not necessarily beneficial or even traceable.
This will also affect the user, who own the agents. If
users are often confronted with decisions/planningsof
the system, which they can not understand (and from
which they don‘t benefit), then they will likely not ac-
cept the system. This must not happen. So it will be a
challenge to provide algorithms, which provide users
with enough information to understand and accept the
decision of the system (without violating privacy con-
straints).
5 CONCLUSIONS
Organic Computing is a very promising new method
for construction of modern software systems. It has
provento be an superior architecture in many domains
like traffic control, sensor networks or production au-
tomation. Organic Computing is also very promising
for a variety of user-intensive scenarios, where indi-
vidual users own individual agents. The purpose of
these agents can be to assist, support and/or guide
their users. One scenario in health care is a plan-
ning/scheduling system in a rehabilitation clinic. An
organic system can perform in this context much bet-
ter than traditional approaches. On the other hand
constraints, which arise in this domain like con-
current/individual goals of users, privacy, behavioral
guarantees and user-trust require new OrganicCom-
puting algorithms and techniques. Developing such
algorithms and methods can have a significant im-
pact on many domains and open new opportunities
and functionalities.
REFERENCES
G¨udemann, M., Ortmeier, F., and Reif, W. (2007). Using
deductive cause consequence analysis (DCCA) with
scade. In Proceedings of SAFECOMP 2007. Springer
LNCS. accepted at SAFECOMP07.
G¨udemann, M., Ortmeier, F., and Reif, W. (2006). Formal
modeling and verification of systems with self-x prop-
erties. In Yang, L. T., Jin, H., Ma, J., and Ungerer, T.,
editors, Proceedings of the Third International Con-
ference on Autonomic and Trusted Computing (ATC-
06), volume 4158 of Lecture Notes in Computer Sci-
ence, pages 38–47, Berlin/Heidelberg. Springer.
G¨udemann, M., Ortmeier, F., and Reif, W. (2006). Safety
and dependability analysis of self-adaptive systems.
In Proceedings of ISoLA 2006, 2nd Symposium on
Leveraging Applications of Formal Methods, Verifica-
tion and Validation. IEEE CS Press.
Kephart, J. O. and Chess, D. M. (2003). The vision of auto-
nomic computing. Computer, 36(1):41–50.
M¨uller-Schloer, C. (2004). Organic computing: on the fea-
sibility of controlled emergence. In CODES+ISSS
’04, NY, USA. ACM Press.
M¨uller-Schloer, C., von der Malsburg, C., and W¨urtz, R. P.
(2004). Organic computing. Informatik Spektrum,
27(4):332–336.
Pietzowski, A., Satzger, B., Trumler, W., and Ungerer, T.,
editors (2006). A Bio-Inspired Approach for Self-
Protecting an Organic Middleware with Artificial An-
tibodies, volume LNCS 4124. Passau, Germany.
Richter, U., Mnif, M., Branke, J., M¨uller-Schloer, C.,
and Schmeck, H. (2006). Towards a generic ob-
server/controller architecture for organic computing.
In Hochberger, C. and Liskowsky, R., editors, INFOR-
MATIK 2006 – Informatik f¨ur Menschen, GI-Edition
Lecture Notes in Informatics, pages 112–119, Bonn,
Germany. K¨ollen Verlag.
Seebach, H., Ortmeier, F., and Reif, W. (2007). Design
and Construction of Organic Computing Systems. In
Proceedings of the IEEE Congress on Evolutionary
Computation 2007. IEEE Computer Society Press. ac-
cepted for publication.
Statistisches Bundesamt (2006). 11. koordinierte
bev¨olkerungsvorausberechnung annahmen und
ergebnisse.
Statistisches Bundesamt (2007). Gesundheit ausgaben
2005.
Trumler, W. (2006). Organic Ubiquitous Middleware. PhD
thesis, Universit¨at Augsburg, Eichleitnerstr. 30.
Trumler, W., Bagci, F., Petzold, J., and Ungerer, T. (Septem-
ber 20-24 2004). Towards an Organic Middleware for
the Smart Doorplate Project. In Workshop on Or-
ganic Computing, INFORMATIK 2004 - Informatik
verbindet, pages 626–630, Ulm, Germany.
VDE/ITG/GI (2003). Organic Computing: Computer-
und Systemarchitektur im Jahr 2010. http://www.gi-
ev.de/download/VDE-ITG-GI-Positionspapier Or-
ganic Computing.pdf.