MediGrid – Facilitating Semantic-Based processing of

Biomedical Data and Knowledge

Jan Vejvalka

Petr Lesný

, Tomáš Holeček

, Kryštof Slabý

, Adéla Jarolímková

and Helena Bouzková

Faculty hospital in Prague Motol

V Úvalu 84, 150 06, Prague 5 – Motol, Czech Republic

FMS FHS Charles University

U Kříže 10, 158 00 Praha 5 Jinonice, Czech Republic

CESNET, z.s.p.o.

Zikova 4, 160 00 Praha 6, Czech Republic

National Medical Library

Sokolská 54, 121 32 Praha 2, Czech Republic

Abstract. New ICT approaches promise better support to the traditional ways

that medicine handles information. Exploring ways to provide this support, we

formulated the principles of he MediGrid: (1) Data processed by biomedical

algorithms are (following the philosophical tradition of phenomenology) indi-

cators that can be transformed into other indicators. (2) Data and algorithms can

be shared across conceptual domains if documented semantic links exist to

support such interconnection. The need of explicit and detailed documentation

of semantics leads to the requirement of good documentation of computer pro-

cedures that implement the biomedical knowledge contained in scientifically

accepted algorithms. A proof-of-concept implementation of a system based on

these principles has been published on Sourceforge.

1 Introduction

Fast growing volume of scientific knowledge and resulting specialization in medicine

stresses the utilization of algorithms and guidelines, that objectify and standardize

healthcare processes. The idea of using ICT to support these algorithms and

guidelines is natural: execution of algorithms is what information technologies

usually do. Algorithms of care, however, are executed on a different level, and with

different data than computer algorithms. Proper use of ICT requires a proper

computer representation of these algorithms and data. Several projects aiming at

creating or documenting a collection of biomedical algorithms exist (such as

MEDAL, MedCalc etc.).

Vejvalka J., Lesný P., Hole

cek T., Slabý K., Jarolímková A. and Bouzková H. (2009).

MediGrid – Facilitating Semantic-Based processing of Biomedical Data and Knowledge.

In Proceedings of the 1st International Workshop on Open Source in European Health Care: The Time is Ripe, pages 18-21

DOI: 10.5220/0001828500180021

 SciTePress

A problem that has been complicating ICT support to algorithms in healthcare is the

nature of information that healthcare deals with, its non-trivial semantics and

traditional models of handling this information. To make automatic computerized

processing of biomedical information possible, its semantic value must be properly

represented (and evaluated) in a computer-comprehensible form. To facilitate correct

processing of biomedical data and their semantics, we have designed the MediGrid, a

working tool for representation and processing of domain ontologies in biomedicine.

2 Materials and Methods

MediGrid aims to facilitate automatic processing of biomedical data with

scientifically relevant knowledge transformed into computational algorithms, using

up-to-date information technologies.

MediGrid approach is based on representation of biomedical data and algorithms as

resources, and on sharing these resources in a grid-like environment. The natural way

of assigning biomedical algorithms to data (or vice versa) is by comparing their

semantic values; therefore MediGrid builds its mechanisms of sharing on capturing

and processing of semantic information. Results from previous projects (SMARTIE,

Growth) have been used as a base for our work: we used existing algorithms

(Growth, MEDAL) to verify our results and we used our experience from SMARTIE

to structure semantic information.

3 Results

Based on extensive research and on previous work, we have postulated the basis of

MediGrid as the working tool for analyzing algorithm-related ontologies in medicine

and for realizing (and practically implementing) their potential for data processing.

The key principles of MediGrid can be expressed as follows:

1. Data processed by biomedical algorithms are (following the philosophical

tradition of husserlian phenomenology) indicators that can be transformed

into other indicators and grouped into indicator classes by their roles in these

transformations.

2. Data and algorithms can be shared across conceptual domains if documented

semantic links exist to support such interconnection.

The emphasize on semantics of both algorithms and data, the need for its extensive

review and verification when linking data with algorithms, and also the stress on cor-

rect procedures (lege artis) in medicine bring practical requirements on design and

implementation of a knowledge-processing system based on these theoretical princi-

ples:

• Semantic information (meaning for the human user) of both indicator classes

and transformations must be explicitly described and available for assess-

ment and validation.

• Semantic information must be bound to the current scientific paradigm and

to evidence based medicine through extensive links to published and re-

viewed works.

• Mechanisms of procedural authority and trust must be implemented to sup-

port users' decisions about procedural and semantic values of individual

components.

We have identified several entities, which have to be documented in order to achieve

the aforementioned requirements. These entities can be set up in 4 layers:

• The source layer, which contains the description of MediGrid information:

author and cited work.

• The algorithm layer, which contains the description of the most commonly

used categories – the transformation (algorithm), indicator class (the data

entered or exchanged between the algorithms)

• The implementation layer, which contains information about the specific

implementation of transformations and of validations of indicator classes in

computer programs.

• The review/trust layer, containing the user reviews and trust statements.

Description of each entity contains four basic elements:

• human semantics – the collection of human readable description, which

enables user to understand the meaning of the entity

• metadata – pieces of computer recognizable data, which can be utilized e.g.

to construct the user interfaces

• relations – relations of this entity to other entities

• classifications – special case of relations that position the entity in external

classification systems and ontologies (UMLS, MeSH descriptors etc.).

Based on the described principles, we posted our first (proof of concept) open source

implementation of MediGrid to Sourceforge: http://medigrid.sourceforge.net

4 Conclusions

The need for explicit documentation of biomedical data and algorithms and their

semantics on several layers leads to the requirement of good documentation of

computer procedures that implement the biomedical knowledge contained in

scientifically accepted algorithms. Similarly, in order to be open to scientific scrutiny,

also the framework system for processing of biomedical information has to be

sufficiently documented at all levels.

Acknowledgements

Supported by Czech research projects MediGrid, 1ET202090537 and by VZ FNM,

MZO 00064203.

References

1. Krásničanová H., Lesný P.: Compendium of pediatric auxology 2005. Praha, Novo Nordisk

(2005).

2. Svirbely J, Iyengar MS.: Issues in the implementation of computer-based medical algo-

rithms. Technology and Health Care. 2005; 13(5); 438-439.

3. Smart Medical Applications Repository of Tools For Informed Expert (S.M.A.R.T.I.E.)

and MedNotes™. IST (2000). http://www.smartie-ist.org

4. Lipscomb C.E.: Medical Subject Headings (MeSH). Bull Med Libr Assoc (2000) 265-6.

5. Cote R.A., Robboy S.: Progress in medical information management. Systematized nomen-

clature of medicine (SNOMED). Jama (1980) 756-62.

6. Lindberg C.: The Unified Medical Language System (UMLS) of the National Library of

Medicine. J Am Med Rec Assoc (1990) 40-2.

7. Husserl E. Logical Investigations, Investigation I. (translated by J. N. Finday). Routledge,

London (2001) 184.

8. Kuba M., Krajíček O., Lesný P., Vejvalka J., Holeček T.: Grid empowered sharing of

medical expertise. Stud Health Technol Inform (2006) 273-82.

9. Bodenreider O.: Biomedical ontologies in action: role in knowledge management, data

integration and decision support. Yearb Med Inform. 2008:67-79.