jump to navigation

Development of a free knowledge base to support the use of differential diagnosis information in PET-CT imaging. June 3, 2010

Posted by tomography in Conference, Nuclear Medicine, Radiology, SNM 2010.
add a comment

Here is the second project I am involved with that will be presented at SNM 2010 in Utah.

Development of a free knowledge base to support the use of
differential diagnosis information in PET-CT imaging.

Zoltán Tóth, Zoltán Barta, András Székely, Nikoletta Pócsi, Ildikó Garai.

OBJECTIVES

Our aim was to create an application integrated into the reporting
workflow to provide quick and easy access to relevant scientific
publications, facilitate the use of differential diagnosis information
and support evidence-based decision making. This project is part of a
larger project targeting the development of a structured reporting
system (ReportAssistant) for oncological whole-body PET-CT studies.
(See accompanying poster).

METHODS

Creation of a PET-CT gamut list

The gamut list was created by combining the results of two independent
methods. First we selected 160 representative reports from our
practice. These were processed sentence-by-sentence by two physicians
in order to extract the most common and less frequent findings in
whole-body PET-CT studies. Copies of the original report text were
modified inline: sentences describing relevant lesion types were
“tagged” with “location – lesion – diagnosis” triples. These tags were
collected and summarised as potential gamut items with the help of a
software script. Parallel with these, a second list was created by
processing and sytntetising various scientific literature sources.
After reviewing, these two lists were merged to form the first edition
of a PET-CT gamut list. These version of the list covers chest,
abdomen and pelvis related lesions and consists of 400 gamut items.

Collecting and tagging publications

Systematic literature searches using a comprehensive list of keyword
expressions distilled from the PET-CT gamut list were conducted. The
main target journals were Radiology, Radiographics (both official
journals of RSNA), Journal of Nuclear Medicine (the official journal
of SNM) and European Journal of Nuclear Medicine & Molecular imaging
(the official journal of EANM). A total of 745 articles were selected.
The URL addresses of the full text articles were recorded and manually
tagged with proper gamut triples. Further processing, including
downloading and indexing of PDF articles, was automatically done by
custom software written by one of the authors (Z Barta).

Software development

All software development was done by using the X4Tk toolkit. X4Tk is
an open source (LGPL), object-oriented MVC-framework for easy desktop
application development and more, including support for standard
enterprise messaging protocols, TCP/IP based server-client development
and various interprocess communication solutions. X4Tk is written in
Perl by one of the authors (Z Barta), with the first release in 2005,
it is actively maintained and field-tested.

RESULTS

Database of tagged publications

A comprehensive collection of high-quality, manually selected,
properly reviewed and tagged radiology and nuclear medicine
publications was generated by using the methods described earlier with
the help of our custom RA PaperManager software. This list can be
exported to an external file. By loading this list one can easily
rebuild the database on his own computer without violating copyright
restrictions.

RA PaperManager software

The main features of RA PaperManager are:

* extensive support for collecting scientific papers ranging
from dragging-and-dropping of files to direct web browser integration through
bookmarklets, respectively import of list of URLs pointing to full
text PDFs and scanning & monitoring of hard drive directories.

* “free” and gamut-list based tagging of papers

* full text search support

* easy linking & interprocess communication with other
applications (STOMP, Windows MMF, HTTP)

CONCLUSION

We have designed, implemented and tested a highly useable, general
purpose, point-n-click SR software. Our software system can serve for managing personal collections of scientific publications and it can be easily integrated into structured reporting softwares through it’s REST API. The
software is available under various licenses, including free licenses
for personal and academic institutional usage.

Look for Zoltán Barta and Zoltán Tóth on site for more details!

– András

ReportAssistant: a highly useable structured reporting system for PET-CT studies June 3, 2010

Posted by tomography in Conference, Nuclear Medicine, SNM 2010.
add a comment

Here is the first poster I am involved with:

ReportAssistant: a highly useable structured reporting system
for PET-CT studies

Zoltán Barta, Zoltán Tóth, András Székely, Nikoletta Pócsi, Nikol
Fedinecz, Bernadett Szűcs, Ildikó Garai

OBJECTIVES
Structured reporting (SR) solutions may have many practical benefits compared to conventional reporting. While SR systems are widely used in various diagnostic settings (in gastroenterology, cardiology and obstetrics), their use in general radiology and nuclear medicine is very limited due to the complexity of the domains and the lack of useable software products. The main objective of our project was to develop a practical SR system for whole-body oncological PET-CT studies, which can be easily adapted to general nuclear medicine and diagnostic CT imaging studies later on.

METHODS

Requirements

We defined the following main requirements:

* easy creation of multimedia reports

* context-sensitive and quick access to relevant decision
support materials

* user-friendly support of academic – research and education – activities

* flexible and simple integration with custom DICOM image
processing softwares

Concepts and prototyping

We used the concept of tagging as the basis of the design for our
softwares. We hypothetised that tagging – as used by many contemporary popular web services – can be the unifying component in interlinking lesion data, patient history, relevant scientific publications, etc. Our initial studies proved that
using composite tags of “location + lesion type + disease”
(the classic gamut formula), we can eliminate the deep hierarchical
navigational lists found in several commercial radiology point-n-click SR softwares. After experimenting with various decision support models, we concluded that a simple and cost-effective solution would be to use tagged, high quality, and freely accessible review and CME articles as decision support resources. (This work is presented in an accompanying poster). During the whole project we considered software usability as the most important problem to be solved. Therefore many throw-away prototypes were built and tested in order to find simple and effective UIs to streamline the workflow. We invested significant effort in reducing the time required to perform key repetitive tasks to the minimum. It is
obvious that even a two-second reduction in the completion time of a moderately frequent task can save many hours or days per person at a year scale.

Final design & architecture

The ReportAssistant platform consists of three independent, but
interconnected programs: an “editor” responsible for the creation of structured reports, a “viewer” application for archiving, teaching
file viewing and querying of the report database, and a custom
“papermanager” program, for collecting, full text indexing and tagging of scientific papers.

Software development

All software development was done by using the X4Tk toolkit. X4Tk is an open source (LGPL), object-oriented MVC-framework for easy desktop application development and more, including support for standard enterprise messaging protocols, TCP/IP based server-client development and various interprocess communication solutions. X4Tk is written in Perl by one of the authors (Z Barta), with the first release in 2005, it is actively maintained and field-tested.

RESULTS
RA Editor

RA Editor is the key application of the ReportAssistant platform: it
provides all the functionalities needed to create structured reports.
Other important features are:

* report template editing

* flexible, adaptable text macro subsystem

* study metadata and patient data viewing/importing/editing

* connection to our legacy RIS database

* communication with Mediso InterviewFusion image processing
software, through standard TCP/IP connection, by using a simple
XML-based declarative mini-language.

RA Viewer

RA Viewer is responsible for the viewing and searching of structured reports created with RA Editor. RA Viewer features include:

* search based on “free” and Dx tags, patient and study metadata

* export of ROI/VOI data combined with patient and study data
(MS Excel format)

* display of key images, report text, report summaries for quick reviewing

* annotation of images and export to PDF and MS PowerPoint formats

* digital signing and email distribution of the electronic
reports (in PDF format)

RA PaperManager

The main features of RA PaperManager are:

* extensive support for collecting scientific papers ranging
from dragging-and-dropping of files to direct web browser integration through bookmarklets, respectively import of list of URLs pointing to full text PDFs and scanning & monitoring of hard drive directories.

* “free” and gamut-list based tagging of papers

* indexing for full text search

* easy linking & interprocess communication with other
applications (STOMP, Windows MMF, HTTP)

CONCLUSION

We have designed, implemented and tested a highly useable, general purpose, point-n-click SR software. Our solution provides decision support directly integrated in the workflow. We believe (suggest) that access to “solved” cases and relevant evidence-based papers (reviews, CME articles from top journals) increases confidence in decision making, facilitates learning, deepens experience, which indirectly can lead to a better and more effective/reliable medical practice and is specially useful for resident students. Academic research and education activities are also supported by the software in the form of easy search & export of lesion data to spreadsheets, calculation of likelihood ratios, respectively a searchable structured report database of studies. Undergoing work will integrate speech-to-text solutions and will provide support for the conversion of the reports to DICOM SR (Suppl 23) standard format.

Look for Zoltán Barta and Zoltán Tóth on site for more details!

– András