Transcript Prototype Development - Université Libre de Bruxelles

Black Skin Dermatology Online :
Contribution to teledermatology
Prime Minister's office
Federal office for scientific, technical and cultural affairs
Multiannual information society support programme 2001-2008
1
Contents







General objective and partnership
Project management
Feasibility study
Content acquisition
Platform Prototype
Demo
Second phase
2
General objective
 To establish an experimental web platform
of teledermatology on black skin, multilingual and
open, integrating image and voice recognition
technologies.
 Phase 1
 Development of a prototype
 Build a representative corpus in dermatology on
black skin
 Phase 2
 Development of telematics services related to the
database.
 Evaluation of the prototype, including its usability
3
Partnership

Five partners




Contents and medical expertise
Definition of the functionalities of the platform
Guarantee scientific reliability of the database
Testing






Centre hospitalier régional de Kaolack (Sénégal)
Hôpital Erasme
Institut de médecine tropicale Prince Léopold
Universiteit Gent
Universiteit Antwerpen (UIA)
A coordinator


Realize data-processing analysis and developments
Plan partner ’s action in the study of the needs, the functional
analysis and the evaluation of the prototype

Université Libre de Bruxelles
4
Advisory committee
 Dermatologists
 Expert in voice technology
 Expert in information technology
 Feedback on various issues
5
Project management (1)
 Successful follow-up of tasks during the
project
 State of art (months 1-6)
 Study of needs (months 1-6)
 Feasibility study (months 6-12)
 Development of a prototype (months 12-24)
 Content acquisition and alpha-tests (months 124)
 Project management (months 1-24)
 Supervision of four activity reports
 Creation and regular updates of a Web site
(http://www.bib.ulb.ac.be/RDIB/Blackskin/index.htm)
6
Project management (2)
 Organization of twelve working meetings among
partners
 Organization
of
two
advisory
committee
meetings
 Conclusion of a subcontract with the expert in
vocal recognition for a state of the art and a
feasibility study
7
Feasibility study
 Covers:
 Study of needs
 Metadata of the images
 Software overview and choices
 To efficiently store and retrieve the images
 To deliver content to the end-user
 Voice recognition
8
> Feasibility Study
Study of needs (1)
 Phase 1 :
 Online image database
 Comprehensive description of images
 Image submission through Web interface,
with appropriate authentication mechanisms
 Interoperable
 Data protection mechanisms
 Integration of voice recognition technologies,
if feasible
 Content management system
 Free and open-source software
9
> Feasibility Study
Study of needs (2)
 Phase 2 :
 Image database
 Continuous development of image collections
 Possibility to add comments on already existing images
 Collection export onto self-contained CD-ROM for offline
browsing


Prototype evaluation
Added-value services in the CMS:
 Expert’s forum
 « Ask a specialist » service
 Teaching support (didactic quizzes, integration in ecourses, …)


Technological watch
Collaboration
with
other
existing
telemedecine and teledermatology
projects
in
10
> Feasibility Study
Metadata of the images (1)
 based on other existing atlases
 Dermis/DOIA (Dermatology Online atlases)
 Dermatology image bank (“Knowledge
weavers” University of Utah)
 Knowledge Lumen University of Chicago
 Dermatlas (Johns Hopkins University)
 Atlas Tunisien de Dermatologie online
 ULB image library
 MS-Access prototype to determine
required fields
11
> Feasibility Study
Metadata of the images (2)
12
> Feasibility Study
Metadata of the images (3)
13
> Feasibility Study
Metadata of the images (4)
 Thesaurus
 Study of MESH, ICD-10, CISP-2, UMLS
 Based
on
ICD-10
(WHO’s
International
Classification of the Diseases): 10th version of
the nomenclature created in 1893.
 Descriptive structure of pathologies in 5 levels.
 Contains more than 2000 terms coming from all
the chapters from the standard.
 Translation in NL, FR, EN and DE.
 Interoperability with other systems that use
PubMed classification (Dermis, Dermatlas, …)
and with the CISP-2, UMLS
and MESH
standards.
14
> Feasibility Study
Metadata of the images (5)
15
> Feasibility Study
Software
 Content management system
 Digital library software
 Characteristics
 Open-source
 Interoperable
 Maintainable
16
> Feasibility study
Content Management System (1)
 Definition
“A content management system (CMS) supports the creation, management, distribution, publishing, and
discovery of corporate information.
It covers the complete lifecycle of the pages on your site, from providing simple tools to create the content,
through to publishing, and finally to archiving.
It also provides the ability to manage the structure of the site, the appearance of the published pages, and
the navigation provided to the users. […]”
(http://www.steptwo.com.au/papers/kmc_what/)
 Lots of competitors:
 350+ CMS on http://www.cmsreview.com
 Requirements
 Open-source
 Localization  multilingual
 Plug in API
17
> Feasibility study
Content Management System (2)
 Comparison of 4 CMS:
 Magnolia/OpenCMS/Mambo/Typo3
 Our Choice: Typo3
 Rapid prototyping
 Hundreds of available plug-ins
 Forum
 News management
 News letter
 Quiz
 Active community (+3000 registered sites)
 Rich documentation
18
> Feasibility study
Content Management System (3)
19
> Feasibility Study
Image Database (1)
 Dspace
 Java
 Workflow Management
 Indexed fields are hard-coded
 Greenstone 2.52
 lots of desirable features
 lack of powerful communication interface for our CMS
 building process requires the whole collection to be rebuilt
each time
 Home-made solution
 blackskin specific
 cost
 seriously investigated before announcement of…
20
> Feasibility Study
Image Database (2)
 Greenstone 3
 complete redesign of Greenstone 2 structured
as a network of independent modules that
communicate with XML
 collections are backward compatible
 written in Java → servlet
 new incremental collection building process
 quicker
 more flexible
 SOAP interface
 still alpha release
21
> Feasibility Study
Image Database (3)
 Interoperability





Major concern
SOAP interface
Open Archives Initiative
Z39.50
Seamless integration in existing digital library systems
22
> Feasibility Study
Voice Recognition (1)
 Experts recommended to adopt an already
existing software solution rather than a
home-made one.
 Cost
 Integration problems
 2 widely available software solutions were
benchmarked
 IBM ViaVoice vs Dragon Naturally Speaking
 Dragon was available in all 3 languages
required (FR, NL & EN)
 Dragon proved to be the clearest and the
most user friendly.
23
> Feasibility Study
Voice Recognition (2)
 Experts’ general conclusions:
 Integration of voice recognition techniques
in Blackskin not as easy as it first seemed.
 Voice
recognition
is
definitely
not
appropriate to fill in forms such as those
requested by Blackskin
 Much slower encoding
 Discouragement of users
 But we wanted to confirm the experts’
conclusions by conducting empirical tests.
24
> Feasibility Study
Legal aspects
 No specific Belgian
databases but:


legal
directives
for
image
Law of 8 december 1992 with respect to “the protection
of privacy in processing data of a person”.
Ordre des Médecins: «Recommandations relatives à la
tenue de bases de données médicales contenant des
données
nominatives
ou
identifiables»
http://www.ordomedic.be/web-Fr/fr/a97/a097008f.htm
 Blackskin does not contain any personal information
of the patients (except for age and gender).
 Draft of license agreement in which submitter gives
away all his rights over the images to Blackskin
25
Content acquisition
 Scanning of existing slides
 400 slides selected and scanned
 from which 300 are completely
available in the prototype
described
and
 Picture acquisition in Senegal
 1800 pictures
 from which 600 are suitable for inclusion in the
prototype
 300 completely described and ready for upload
26
Prototype Development





Software architecture
Web site structure
Typo3 plugin
Greenstone3 configuration
Backend
27
> Prototype development
Software Architecture
 Open-source components
 Typo3 plug-in and GSDL3 backend
28
> Prototype development
Web site structure (1)
29
> Prototype development
Web site structure (2)
 Home
 Welcome message
 Short presentation
 # of pictures
30
> Prototype development
Web site structure (3)
 Search pages
 Browsing tools
 ICD10
 Description
 Body site
 (Homunculus)
 (New Images)
 Site search
31
> Prototype development
Web site structure (4)
 Information about:





Project
ICD10 classification
Staff
Contributors
Disclaimer
32
> Prototype development
Web site structure (5)
 Tools
Submit an image
Contact maintainers
Forums
Newsletter
subscription
 Links page




33
> Prototype development
Web site structure (6)
 Teaching
 Ask a specialist
 Pedagogic
cases
and quizzes
 Bibliography
 CD-ROM order
34
> Prototype development
Web site structure (7)
 Help



Technical help
FAQ
Sitemap
35
> Prototype development
Typo3 plug-in (1)
 Handles communication between Typo3 and
Greenstone3 through SOAP calls
 Fully configurable
 Configuration files
 TypoScript
 Problem
 Lack of documentation about available messages
and options
 Solution: use of tcpTunnelGui and demo
collection to analyze messages
36
> Prototype development
Typo3 plug-in (2)
 GSDL3 can query remote installations
37
> Prototype development
Typo3 plug-in (3)
 Features:




Search a given index
Browse hierarchical classifiers
Multilingual support
Look and feel can be adjusted using CSS
 Flexible but could be more generic
38
> Prototype development
Greenstone 3 (1)
 Using latest version from CVS
 Problem
 Lack of up-to-date documentation for
installation
 Alpha software
 Solution: Live with it!
39
> Prototype development
Greenstone 3 (2)
 Collection configuration
 One collection for each partner
 Inherent to GSDL building process
 One supercollection
 Indexes for searching
 For each searchable field
 Hierarchical classifiers
 Body site





Used by homunculus
Contributor
Diagnostic
Description
ICD10
40
> Prototype development
Greenstone 3 (3)
 1 supercollection
 1 collection / partner
41
> Prototype development
Configuration files (1)
 Used to populate drop down menus and to configure
classifiers
 Automatically generated from Excel files

Easily maintainable
 Depending on their nature, generated by:


Java application
XSLT
42
> Prototype development
Configuration files (2)
43
> Prototype development
Configuration files (3)
44
> Prototype development
Configuration files (4)
45
> Prototype development
Configuration files (5)
46
> Prototype development
Backend (1)
 New user-friendly GSDL3 backend
 J2EE servlets and JSP
 Motivation
 2 existing means
 Collector
 Web-based interface
 Too limited
 Greenstone Librarian Interface
 Java application
 Effectively allows users to upload, modify and build
collections
47
> Prototype development
Backend (2)
 GLI screenshot
48
> Prototype development
Backend (3)
 Motivation
 Why?
 Must be run on same computer
 Solution: applet but... not always available
 No preview mechanism
 need for a second application
 Difficult speech recognition integration
 Graphical layout harder to modify than
HTML
49
> Prototype development
Backend (4)
50
> Prototype development
Backend (5)
 Features








Authentication
Upload
Metadata Enriching
Image preview
Modify & delete
Collection building
Experimental speech recognition interface
Multilingual
51
> Prototype development
Backend (6)
52
> Prototype development
Backend (7)
53
> Prototype development
Backend (8)
54
> Prototype development
Backend (9)
55
> Prototype development
Backend (10)
56
> Prototype development
Backend (11)
57
> Prototype development
Collection building (1)
 We want more than just display a picture
 Need to show metadata too
 Solution: create a folder with HTML file
 Collection building process
 Collection preparation
 Create folder for each jpg file
 Copy thumbnail created at upload time
 Use template file to generate HTML with values from
metadata file

more flexibility
 Then, we can
mechanisms
use
the
regular
GSDL
building
58
> Prototype development
Collection building (2)
59
> Prototype development
Voice recognition (1)
 Followed expert’s advices
 Alternative interfaces were created
 Drop down menus replaced by lists of clickable links
 Intermediary steps to reduce on -screen choices
 Results were deceiving
 Similar terms
 Software not finely tuned for this field of application
 Solution
 Only speak an identifying number
 Better recognition but still slower than keyboard and
mouse
60
> Prototype development
Voice recognition (2)
61
> Prototype development
Voice recognition (3)
62
> Prototype development
Bulk import (1)
 Motivation
 To
not
slow
down
dermatologists
classification and description of pictures
with
 Excel file
 Use of XSLT to convert it to a metadata.xml
file
 no validation made at this stage
 erases any record metadata that could be
present in the collection
 Allowed import of ± 300 pictures
63
> Prototype development
Bulk import (2)
64
Demo
http://localhost:8080/blackskin
65
Concluding remarks

Technical platform put in place, taking into account users
needs for both phases of project




Extensibility through Typo3 plug-ins
Interoperability (OAI, Z39.50)  can be easily integrated
into existing or new-to-come networks
Open source and free  reusability
Maintainability in mind

Voice recognition : not mature enough

Deliverables OK



Development of a prototype
Build a representative corpus in dermatology on black skin
Robust basis for the added value services planned in the
second phase
66
Phase 2

Hand-over of scientific management of project


head librarian of ULB medicine library
Collection development



continue collection development of project partners
creation of additional collections: old dermatological
atlases, medical documents (e-books),…
development of partnerships : other on-line atlases,
dermatological sites, foreign institutions, individual
dermatologist in African countries, …
67
Phase 2
 Added value services






users forum to discuss clinical cases.
forum for experts: discussion of complex cases
front-end for submitting cases and images diagnosed by
the users
medical cases
didactic functions: quizzes, …
collection export onto a CD-ROM with a complete search
and visualization system
68
Phase 2
 Interfaces
 Backend : testing procedure of the encoding
interfaces for the experts and search interfaces
for the users in collaboration with specialists in
ergonomics. A generic procedure that could be used
for the evaluation of all databanks of medical images
in general.
 Front-end
improvements
thanks
to
users
feedback.
 French, German and Dutch versions of the
Website, encoding interface and search pages.
69
Phase 2
 Technical


Development of a registration office to handle the
administration of new and existing partners
Security
and
authentication
issues
e.g.
watermarking solution implementation and rights.

A more generic Typo3
Greenstone collection
plug-in
for
accessing

Technical follow-up of the platform (upgrades,
potentially useful developments [e.g. the new GSDL3
collection building process])
 Business plan
70