You are on page 1of 6

Object-Relational Data Model Assessment for Medical Imaging Repositories Content

Search

Baquiran,Velorie D.
School of Electronics, Electrical and Computer Engineering
Mapua Institute of Technology
Muralla St., Intramuros, Manila, Philippines
veldbaquiran@gmail.com

Abstract- As part of the progress in Discussed in this paper, the limitations of


medical technology, data or information the proposed Data Model.
repositories of medical images
Keywords- Medical Imaging, Object-
contributed greatly regarding medical
Relational Data Model
understanding. These records are being
used together with open-source software 1. Introduction
with different platforms in assisting
medical institutions playing a vital role Variety of data models are being used
during medical decisions and treatment in medical imaging practices throughout
procedures. Through propagation of the years. Medical imaging processes big
digital modalities, medical images data data due to its different resolutions in
grew leading to the need of having a support for more effective medical
bigger repository for safe-keeping that services and practices in clinical analysis.
would be able to accommodate more with [1] Laboratories that performs digital
acceptable performance. Different imaging deals with bulks of data which
database management systems supports includes the medical data of a person as it
each data model in order to fully grasp increases over a period of time. Such data
the its own structures. These databases aids medical doctors in decision-making
includes the use of Picture Archiving and regarding the patients diagnosis. In order
Communication System (PACS)-used in to support these data, advanced medical
order to take in the medical images with imaging equipment has been propagated.
Big Data usage set-ups and PostgreSQL Therefore medical repositories data opt to
for the object-relational database have a more flexible data model and
management, as well as the Dicoogle management systems to be used for
open source software and extensions. attaining a better performance. Necessary
hardware, software, distribution, storage
For this paper, the possible use of Object-
and analysis of medical images have to be
Relational Data Model would be explored
taken into consideration which have been
in enhancing the process of query and
answered with the use of PACS. The
retrieval of medical images in compliant
object-relational data will then use a novel
with the medical imaging standards.
likelihood-ratio of comparison for In this article, implementation of a
probabilistic models.[2] newer data model-which is the object-
One of the main data models used for relational data model- would be focused
data that are structured is the object- on a bigger volume for medical images
relational data model wherein developers storage with an acceptable quality of
integrates databases with their data types service will be allowed upon enhancement
and methods. By providing an addition of of the data model.
a newer and extensive object storage
capability, a field containing a data with 2. Background
specific attributes would be grouped 2.1 OBJECT-RELATIONAL DATA
together for easier database management. MODEL
By focusing on the optimization of Object-Relational Model provides a
querying services and targeting PACS relational database management which is
content discovery services a bigger storage essentially a relational model where users
for medical images could be allowed.[1] integrate object-oriented features. It has
The DICOM standard (Digital Imaging greater power and flexibility with better
and Communications in Medicine) was performance, and greater data integrity
created in order to properly handle the than the previous models. Data in this
medical images between various devices model is conceptually viewed as an
and vendors. orderly arrangement of data into a relation
Relational Database Management but with a featured contained object.
System (RDMS) has been used in PACS in Values are scalar, therefore at any given
handling repositories. NoSQL databases position in the relation there would only be
are not properly exploited to its full one value per se. it was developed in order
capabilities with regards to storage to meet the objective of combining
characteristics. Attributes that are concepts or relational data model together
characterized by name, data type and with object-oriented programming style.
domain sets values for parameters. As the Vendors nowadays implement OR
rate of data production increases, an concepts into their relational databases as
approach to where a single unified the ISO integrated it into the new SQL
network which allows collaborative standard referred to as SQL3 or SQL99.
situation should be presented to OR data model offers an extended
encapsulate the medical images as well as capability of database servers wherein new
to maintain a proper PACS solution with data types, as well as user-defined patterns
better performance database system. could be defined. One or more current
With the use of an existing dataset, existing data types could be combined into
utilization of the given variables and complex types in order to attain more
equations from previous studies that would flexibility in data organization of
be incorporated for this paper, structures composing of columns and
determination whether object-relational tables. As the user is able to define objects
data model would be fitting to be or types and tables with properties of other
implemented with Medical Imaging objects, addition of new properties that are
Content repositories searching that would specific to the said object that has been
provide a more accurate result. defined is also made possible.
Incorporation of such features aids in developing performance capabilities of the
managing conventional fielded data, more database system itself.
complex objects and others. Such
Indexing plugins organize data for
approach takes on the robust operation and
an easier access for the process of data
performance management as developers
retrieval and storage. This plugin works as
can work with more familiar tabular
the index tasks are called upon arrival of
structures and data definition that hold
new DICOM images for storage to the
more power and greater capabilities.
repositories. After the execution, reports
Query and procedural languages and the
would be sent back to Dicoogle regarding
call interfaces that are used in object
the elapsed time or the number of the files
relational database management systems
that were indexed or its errors. The Query
are familiar for the information system
plugins deals with the content search and
developer.
In order to properly support the retrieval. Upon request, queries are
classic features of an object oriented converted into an internal query language
programming, the relational model need to that would then be forwarded to the
be modified drastically. Thus creating plugins[1].
specific characteristics only for the OR 2.2.2 PostgreSQL
model such as having Base Data type
extension, support on complex objects, PostgreSQl is another Open source
inheritance and rule systems. As a result, Relational Database Management System.
functionality and performance for this It comes with the RDBMS together with
model are optimized giving it a massive other free of charge open sources that are
scalability property.[2] used in server/client model
But to be able to fully implement implementation that which supports the
its features and have its advantages, both standard language of ANSI:SQL2011
object-oriented and relational concepts database. The client and server are
should be understood. There are also cases separated in PostgreSQL resulting to the
wherein upon implementation of OR clients library to become lighter[5]. It is
concepts, object inheritance was not ACID compliant which has full support of
supported. various keys, trigger, etc. and can store
procedures in multiple language.
2.2 Database Management Systems PostgreSQLs features includes point in
and Open Sources time recover, tablespaces, asynchronous
2.2.1 Dicoogle replication, nested transactions, online
backups, complex query optimizer, etc.
Dicoogle is an open source widely Multibyte character encoding is also
used in setting up experimental supported by this open source database.
environments. With the use of its plugins Due to its scalability, developer friendly
and A Software Develeopment Kit (SDK) environment wherein procedures are run
for modules wherein developers are with the use of various programming
allowed to work with their deployment of languages- and its customizations,
functionalities. Plugins such as storage, PostgreSQL is used for object-relational
indexing, query and interfaces helps in database system.
2.3 Related Works using large datasets of medical images,
they were able to set guidelines for the
With newer innovations and new
correct usage of analyzed databases.
technologies that are being used in the
medical field, medical imaging has been 3. Methods
improving. Medical imaging repositories
This particular article aims to assess
performance are given focus of several
whether object-relational data model
studies in order to be able to attain the
would be a better alternative in medical
ideal operation of such technologies. With
imaging content search for relational data
newer technologies, new strategies and
model as conducted previously by other
bigger content storage are necessary and
researchers. Dicoogle was used by the
were targeted. Interaction of different
chosen references and it is due to its
control systems and logic were also
feature of having a Software Development
incorporated to the data models in
Kit (SDK) in order to implement a variety
quantifying relationship of the attributes of
of plugins. It provides storage, index for
the parameters in the database used.
allowing indexing of metadata information
Rascovsky et al, developed a solution of DICOM files and a query which allows
in order to support a medical imaging searching over the information system and
archive which is a CouchDB based retrieval or search results. With these
solution. With their study, CouchDB is features, developing new PACS
highly suitable in the process of storing, repositories was simplified.
retrieving and querying of DICOM files.
First datasets were gathered from a
Document-based databases have said to
relational database and an object-relational
not have limitation of the RDBMS
database produced by the use of Dicoogle
databases.
of the various previous researches.
Langer et al. showed that cloud Indexing plugins and query plugins were
computing helps solve problems faced by developed as it was given focus on for the
medical imaging researchers that are in medical imaging content search. The
line in data mining and processing of process of how the data flows with these
DICOM data[1]. different data models varied. Relational
Database Model made use of specific keys
There are also other existing
that were then used to relate the medical
researches which targeted different
images to keywords that lead to the needed
importance of optimizing the performance
data. But there were errors contained in
of PACS functions and operations with
files leading to creating the secondary
distributed environments. Others proposed
keys. Object-Relational Mapping was used
mechanisms with an efficient cache
wherein data were sorted in a hierarchical
technology that targets the delays in
model that were then retrieved from an
medical imaging retrieval.
already instantiated structure[1]. Indexes
Alves et al. made use of the relational were then created for an easier access a
data model in order to optimize the certain row in lookup tables-therefore
processes of storage, discovery and leading to a faster retrieval of data. In
distribution of medical imaging data. By order to insert queries, the implementation
of relational database it made use of a vendors are able to provide a more
translation to SQL. structured dataset for a better performing
system due to the data models flexibility.
On the other database study, due to the
flexibility of an OR model different 5. Conclusion
methods and datasets were included.
Relational Data Model already
Dicoogle was also used for indexing and it
provides a great database model in
was mixed with different algorithm that
supporting medical imaging services with
employs an iterative deepening strategy.
controlled experiments. It can be used in
As a set of parameter values represent a
setting guidelines with regards to the
class-level association and another
proper usage of databases with big data
different set to a new association a single
medical imaging developments. The use of
feature component could be produced
Object-Relational Model increases the
upon comparing two parametrized models.
performance of the system by having a
The features acquired were then treated as
more structured hierarchy of datasets. By
independent but with a correlation
having a link or relation of a key to a
component that measures the deviation in
certain object, it shows the possible
the features mutual information[2].
relations with a different class. A data
4. Results and Discussion could then be managed collectively
according to their parameters that then
Comparing both datasets and the use of
would lead to a single feature component.
both models and how the data flows
Therefore harnessing a faster way in
through the system resulted into the
obtaining medical imaging content.
realization of the connection between the
Having an object that is related to another
two data models. The process of
object through a mutual information while
acquisitions was mostly the same due to
having a more manageable system would
the use of Dicoogle and its plugins.
simplify the process per se due to its
Dicoogle played the most important role
scalability.
for both data models wherein the medical
images are being processed for storage, In conclusion, Object-Relational Data
indexing and retrieval. The use of Model could be used as an alternative for
Relational Data Model provided the Relational Data Model in Medical Imaging
dataset verifying that when compared to a content search that might provide more
NoSQL database solution, performance power to the system and a bigger storage
was better. But since it uses An OR due to its extensions and a better support
Mapping, it can be said that it has not yet for complex objects.
shown the best possible solution there is.
It is recommended that further study
An Object-Relational Data Model for the use of the Object-Relational Data
make use of queries can be manipulated Model in medical imaging repository
collectively and includes both the search should be done due to its
advantages of better performance of a complexity and increase in cost. A
relational data model and an object- different database should also be done
oriented data model. The developer or alongside with a Relational Data Model to
further verify the feasibility of using an Data, 2015 IEEE Symposium Series on
OR Data Model for medical imaging for Computational Intelligence. IEEE, sept
future references. 2015, pp. 1590-1598
Acknowledgements [3] S. Jandoubi, A. Bahri, N.Yacoubi-
Ayadi, S. Chakhar, A. Labib, Enchanced
Fuzzy Object-Relational Database Model
This paper would not be pushed for Efficient Implementation of FSM,
through without the supervision of my 2015 IEEE International Conference on
professor Engr. Jaye Balbin of Mapua Fuzzy Systems (FUZZ-IEEE). IEEE, aug
Institute of Technology, under the Faculty 2015, pp.1-8
of School of Electronics, Electrical and
[4] A. Kozmirchuk, A. Kokorev, V
Computer Engineering.
Nesterov, E. Mikhailova, PostgreSQL
Service with Backup and Recovery for
Cloud Foundry, Social Media and Web
References (ISMW FRUCT), 2016 International
[1] A.P.Alves, T.M. Godinho, C. Costa, FRUCT Conference on Intelligence. IEEE,
Assessing the Relational Database Model oct 2016 pp.1-6
for Optimization of Content Discovery [5] M. Jung, S. Youn, J. Bae, Y. Choi, A
Services in Medical Imaging Study on Data Input and Output
Repositories, 2016 IEEE 18th Performance Comparison of MongoDB
International Conference on e-Health and PostgreSQL in the Big Data
Networking, Applications and Services Environment, 2015 8th International
(Healthcom). IEEE, sept 2016, pp. 1-6. Conference on Database Theory and
[2] F. Riahi, O Schulte, Model-based Application. IEEE march 2016 pp 1-4
Outlier Detection for Object-Relational

You might also like