DISTRIBUTED MULTIAGENT APPROACH FOR

HYDROCEPHALUS TREATMENT AND MANAGEMENT USING

ELECTRONIC SHUNTING

N. Al-Zu’bi, A. Al-kharabsheh, L. Momani and W. Al-Nuaimy

Department of Electrical Engineering and Electronics, University of Liverpool, Brownlow Hill, Liverpool, U.K.

Keywords:

Hydrocephalus, Electronic shunt, Multi-agent.

Abstract:

Hydrocephalus is a common chronic condition that results in excessive accumulation of cerebrospinal ﬂuid

(CSF) inside the skull, often leading to brain damage. The treatment and management of hydrocephalus remain

a challenging issue, especially for diagnosis, improving current shunt treatment, and predicting shunt success.

Current diagnosis procedure depends mainly on the surgeons’ observation of the clinical symptoms, neuroim-

ages and instantaneous of intracranial pressure recording. These lack accuracy in diagnosis and predicting

the outcome. Dominant treatment relies on passive mechanical shunts; these also exhibit virous problems.

Adding to that, the lack of communication between the community surgeons and a limited understanding of

the hydrodynamics of this disease have limited the effectiveness of hydrocephalus treatment.

This paper proposes a new approach to improve the treatment and management of hydrocephalus through a

multiagent cognitive system over a distributed network of hydrocephalus patients with intelligent shunting

system. This approach will not only develop autonomous treatment method for hydrocephalus, but also it

deﬁnes a method for information acquiring and analysis to better understanding hydrocephalus and assess

shunt functionality.

1 INTRODUCTION

Hydrocephalus is an excessive accumulation of cere-

brospinal ﬂuid (CSF) inside the skull due to im-

balance of the production and absorbtion of the

CSF, and without treatment has a 50-60% death

rate (A.D.A.M., 2002). Even though there is no ac-

ceptable statistical information of hydrocephalus pa-

tients, estimates show that approximately 1 in every

500 children are affected by hydrocephalus and this

rate is increasing rapidly (NINDS, 2008).

Despite more than 40 years of shunt development,

hydrocephalus remains a challenging issue particu-

lary in the three critical aspects of diagnosis, treat-

ment and management. This paper seeks to allow the

uniﬁcation of these aspects by facilitating communi-

cation at all levels.

Nowadays, hydrocephalus is characterised and di-

agnosed by clinical symptoms, like dementia, uri-

nary incontinence, and gait disturbance, and anal-

ysis of neuroimaging either by computed tomog-

raphy (CT) or magnetic resonance imaging (MRI)

coupled with mean value of intracranial pressure

(ICP) (W. Pﬁsterer, 2007). However, these meth-

ods failed to achieve accuracy in diagnosing hydro-

cephalus and selecting patients that would beneﬁt

form shunting (A. Marmarou, 2005a). In addition to

that, there is no acceptable standard for diagnosis and

treatment of hydrocephalus (A. Marmarou, 2005b).

Currently, the dominant treatment for hydro-

cephalus is to divertthe CSF from the ventricles in the

brain to another part of the body by means of an im-

planted shunt. The currently used shunts are differen-

tial pressure shunts that depend on mechanical valves

for operation. Even with the latests advances in shunt

technology allowing non-invasive adjustment of the

pressure setting, the resulting treatment is far from el-

egant, lacking responsiveness and autonomy, and un-

able to communicate. Several studies have veriﬁed

serious problems and shortcomings relating to cur-

rent shunting techniques, such as mechanical prob-

lems and shunt blockage, excessive and insufﬁcient

CSF draining (over-drainageand under-draining),and

others. Shunt failure rate currently stands at 35-

40% (Metzemaekers, 1998). These problems have

been exacerbated by the lack of a viable and meaning-

503

Al-Zu’bi N., Al-kharabsheh A., Momani L. and Al-Nuaimy W. (2009).

DISTRIBUTED MULTIAGENT APPROACH FOR HYDROCEPHALUS TREATMENT AND MANAGEMENT USING ELECTRONIC SHUNTING.

In Proceedings of the International Conference on Health Informatics, pages 503-507

DOI: 10.5220/0001779205030507

 SciTePress

ful communicationchannel between practicing neuro-

surgeons, and the lack of a detailed understanding of

the hydrodynamics of the disease.

The increaseing number of patients suffering from

this disease, and the associated costs of their treat-

ment, have together highlighted the deﬁciencies of

the current modes of treatment, and has stimulated re-

search towards the next generation of hydrocephalus

shunts. What is beginning to emerge is a totally new

approach towards the treatment and management of

the disease. This approach responds to the needs of

individual patients in an autonomous way, and forms

a network among those patients to establish a man-

agement and learning protocols to better understand

aspects of hydrocephalus diagnosis and treatment.

In order to achieve this goal, the ”old fashioned”

mechanical shunt valve must ﬁrst be upgraded to

an electromechanical valve, controlled by software.

This type of shunt will have sensory inputs to al-

low it to process and analyse the patient ICP directly,

and base the patients’ treatement regime on the val-

ues of selected parameters derived therefrom. It has

been demonstrated that certain parameters extracted

from the ICP signal are more meaningful than the

mean or instantaneous ICP in terms of providing in-

formation for diagnosis and predicting clinical out-

come (M. Czosnyka, 2007). Such a system would

be responsive to patient feedback, provind the shunt

with a means of evaluate its own performance, cou-

pled with other derived numeric performance indica-

tors. This can be achieved through communication

between the implanted shunt and a hand-held device,

which could be a normal mobile phone.

This framework of communication, with the abil-

ity of the hand-held device connecting to the inter-

net, gives the ability to connect these shunts in a dis-

tributed network of agents, which could enable shunts

to share their information and disseminate successful

treatment regimes. Moreover, this approach will pro-

vide a platform for classiﬁcation of ICP signals, hy-

drocephalus patients, and treatment regimes.

2 APPROACH SPECIFICATION

2.1 Intelligent eShunt System

This approach proposes a software-driven,

electronically-controlled shunt replacing the passive

mechanical one. This shunt consists of an electronic

valve, ICP sensor, microcontroller with software, and

a transceiver which provides the shunt two-way to

communicate with the outside word. This modiﬁ-

cation of shunt characteristic by adding autonomy

and intelligence allow us to overcome some of the

problems in the mechanical shunts, farther enables

the system to detect malfunctioning.

2.2 ICP Signal Analysis and Parameters

Depending on symptoms, MRI and mean ICP, for di-

agnosis of Hydrocephalus achieves low accuracy, add

to that cost of treatments and shunt revisions which

exceed $1 Billion in year 2000 (Vacca, 2007). Fur-

thermore, these factors do not reliably predict re-

sponse to treatment or clinical outcome after shunt-

ing (W. Pﬁsterer, 2007), which leads to emphasis on

discovering new trends and parameters in ICP wave-

form, which carry valuable information about patient

state (M. Czosnyka, 2007).

This approach, by extracting a set of representa-

tive parameters from the ICP signal, enables the elec-

tronic shunt to respond to the dynamics of the ICP sig-

nal and provide a way of controlling the valve on de-

mand. These parameters, such as Pulse amplitude of

the ICP (AMP), RAP index, mean ICP, and mean ICP

wave amplitude and latency, (M. Czosnyka, 2007),

proves efﬁciency in treatment.

2.3 Distributed Multi-agent

Intelligent agents are an innovativetechnology for de-

veloping complex and distributed systems. These in-

telligent agents have, in an autonomous way, a ﬂex-

ibility in performing actions to achieve their goals.

This ﬂexibility includes reactivity, pro-activity , and

social ability (Weiss, 1999).

Multiagent systems are widely used in healthcare,

medicine and other domains (Moreno, 2003), where

here several novel aspect need to be highlighted:

1. Clustering technique, for clustering patients into

groups.

2. Local and general classiﬁcation technique for ICP

pattern classiﬁcation.

3. Performance evaluation technique, for evaluat-

ing agent’s performance, and assigning successful

credit for each case.

4. Using communication language and identiﬁcation

ontology for agents to announce about their self

and their experiences.

5. Knowledge sharing and negotiation techniques.

This technology provides a suitable choice in this

case, where distributed electronic shunts need to com-

municate and exchange their successful experiences.

These shunts represent a distributed sources of infor-

mation and decision around multiple patients, where

HEALTHINF 2009 - International Conference on Health Informatics

504

single patient (i.e. electronic shunt) does not have

a complete view of the hydrocephalus environment

and does not encounter adequate cases of ICP pat-

terns to develop a general classiﬁcation and manage-

ment protocol. By sharing data and treatment expe-

rience among electronic shunts in different patients,

more effective treatment and management system can

be developed.

Distributed agents will beneﬁt from the semantic

web technology which will be used for exchanging

information, and interacting and reasoning about their

knowledge (Hendler, 2001), with a deﬁned ontology

domain which deﬁne a set of terms and messaging

vocabulary to be used by agents in the communication

language.

While agents are using their knowledge about

their cases and their performance in treatment and

information from other agents, they will be able to

reason about and recommend appropriate treatment

scheme for particular case.

3 APPROACH ARCHITECTURE

3.1 Functionality Overview

The functionality of the approach, goes through the

following steps:

• ICP sensing, the eShunt agent senses the ICP sig-

nal through the pressure sensor and store it in the

database of the eShunt.

• Preprocessing, the ICP signal is preprocessed to

extract its parameters, like pulse amplitude of the

ICP (AMP), RAP index, mean ICP, true ICP, and

mean ICP wave amplitude and mean ICP wave la-

tency.

• Formatting form, the extracted parameters, the

real ICP and other data related to the patient such

as disease history, patient feedback, surgeon ad-

vice and diagnosis, are putted down in a common

form.

• Processing, then based on the information in the

form, the eShunt agent tries to discovernew trends

like new thresholds and values of the ICP pa-

rameters with corresponding clinical cases, such

as shunt malfunctioning, patient clinical outcome,

ICP production rate state.

• ICP classiﬁcation, then the eShunt agent, based

on different ICP parameters and processed data,

classify the ICP signal into categories.

• Testing, these classiﬁers are tested using appropri-

ate cases from the distributed eShunt agents.

• Patient clustering, including ICP classiﬁcation,

patient history and patient general information,

patients are clustered into different categories.

The eShunt agent attempts to diagnose its patient

for whom ICP data and other information are avail-

able, uses web service to establish a record of the pa-

tient case information into the system’s blackboard,

in anonymous form. The system via different eShunt

agents may suggest classiﬁcation and diagnosis based

on their data for similar cases, or the enquiring eShunt

may ask via the net whether certain classiﬁcation and

diagnosis are appropriate. Then the performance of

the suggested classiﬁcations and diagnosis becomes a

negotiation process among the system, and when they

agreed on a satisﬁed level of performance of the clas-

siﬁer and diagnosis, the classiﬁer is published.

3.2 Agents Needed

The functionality of the system is fulﬁlled through the

following agents:

• ICP sensing agent to sense the ICP signal and

store it in the database of the eShunt.

• Preprocessing agent to extract the speciﬁed pa-

rameters from the ICP signal.

• Processing agent to discover new trends and

thresholds of the parameters with the correspond-

ing clinical cases.

• Classiﬁcation agent that classiﬁes ICP cases based

on different parameters.

• Testing agent that test the locally classiﬁed and

diagnosed cases through the distributed system.

• Clustering agent that cluster other eShunts in

groups based on similarities.

• Communication agent that identify other agents

and establish a communication link with them.

3.3 Layered Framework

Since the system should be both reactive and proac-

tive, reactive to instantaneous ICP changing and do-

ing simple action based on that, and proactive to sup-

port in a timely fashion ICP parameter extraction and

analysis, and negotiation of appropriate case diagno-

sis and corresponding proper treatment, and in addi-

tion that this system is wrapping the eShunt system

which locally control the valve, the system adopts

multi-layer framework. It is composed of the follow-

ing layers:

• Inner control layer, represents the eShunt behavior

locally which is limited to controlling the valve in

an instantaneous manner.

DISTRIBUTED MULTIAGENT APPROACH FOR HYDROCEPHALUS TREATMENT AND MANAGEMENT USING

ELECTRONIC SHUNTING

505

Figure 1: System high-level overview.

• Processing layer, where ICP analysis is happened.

• Database interface layer, it locally logs ICP data,

treatment regimes and stores patient information

in the database.

• Processing layer, which mines the data provid-

ing the main functionality of the approach such

as ICP classiﬁcation, ICP thresholds ﬁndings, and

patients clusterings.

• Testing layer, where evaluation process occurred

and cases veriﬁed through the system.

• Communication layer, which provide environ-

ment for agent communication via the web.

Figure 2: Framework layers.

4 CONCLUSIONS

This paper has presented a new methodology for the

integration of hydrocephalus biomedical data and ap-

plying knowledge discovery techniques. It speciﬁ-

cally addresses the use of the valuable information in

the ICP signal coupled with patient feedback and sur-

geon examination and enforced by the eShunt agent,

to improve understanding and management of hydro-

cephalus.

A requirement of a considerable amount of ICP

analysis and treatment data is necessary to build a

self-learning and robust classiﬁcation system for ICP

waveforms and hydrocephalus patients. This ap-

proach will tackle challenges in analysing, collecting

and managing ICP data, by providing a distributed

system of eShunt agents that manage patients au-

tonomously, and share information between them. It

is envisaged that such an approach will typify the next

generation of hydrocephalus management and treat-

ment; it will undoubtedly reduce treatment costs dra-

matically, and can potentially save lives.

HEALTHINF 2009 - International Conference on Health Informatics

506

REFERENCES

A. Marmarou, H. Y. e. a. (2005a). Diagnosis and manage-

ment of idiopathic normal-pressure hydrocephalus: a

prospective study in 151 patients. In J Neurosurg, vol-

ume 102, page 987997.

A. Marmarou, M. B. e. a. (2005b). Development of guide-

lines for idiopathic normal-pressure hydrocephalus:

introduction. In Neurosurgery, volume 57, pages S1–

S3.

A.D.A.M., I. (2002). Hydrocephalus.

http://www.avera.org/amck/ adam/1/001571.adam.

Hendler, J. (2001). Agents and the semantic web. IEEE

Intelligent Systems, 16(2):30 – 37.

M. Czosnyka, P. S. e. a. (2007). Intracranial pressure: More

than a number. In Neurosurgical Focus, page E10 17.

Metzemaekers, J. (1998). Hydrocephalus Shunts: A clinical

and laboratory study. PhD thesis, University Gronin-

gen.

Moreno, A. (2003). Medical applications of multi-agent

systems.

NINDS (2008). Hydrocephalus fact sheet.

http://www.ninds.nih.gov/disorders/hydrocephalus/

detail hydrocephalus.htm.

Vacca, V. (2007). Diagnosis and treatment of idiopathic nor-

mal pressure hydrocephalus. In The Journal of neuro-

science nursing, volume 39, page 107.

W. Pﬁsterer, F. A.-E. e. a. (2007). Continuous intraven-

tricular pressure monitoring for diagnosis of normal-

pressure hydrocephalus. Acta Neurochirurgica, Vol-

ume 149(Number 10):983–990.

Weiss, G. (1999). Multiagent systems : a modern approach

to distributed artiﬁcial intelligence. Cambridge, Mass

: MIT Press.

DISTRIBUTED MULTIAGENT APPROACH FOR HYDROCEPHALUS TREATMENT AND MANAGEMENT USING

ELECTRONIC SHUNTING

507