| Resources | Top 10 Lists | Open Source | PACS Consultants | Links |

Open Source FAQs

A Radiology Open Source FAQ
from July 2006 SIIM News
Steve Langer, PhD

What is open source? A lot has been written and discussed about this sometimes-controversial topic. This article uses a question-and-answer format to present certain challenge questions concerning the suitability of open source (OS) in radiology and responses to those questions.

Q: Isn't open source just the latest computing fad?

A: Actually, OS has been around for 30 years or more. There are thousands of OS applications/languages and tools that cover the gamut from office needs, file/print/web/mail/compute servers, science, accounting, mapping, programming, education, games, etc. Some of the most important applications that make the World Wide Web work are listed below.

Common Open Source Applications:

• Sendmail (1980): Responsible for routing about 60% of all email;

• Apache (1993): 64% of all web server traffic (Netcraft 2002);

• BIND (1986): Used by over 80% of Internet Name lookups;

• BSD UNIX variants and Linux (both around 1993): Combined (including Mac OS X) account for 44% of the OS server market (Netcraft 2001);

• PERL (1987): An enormously popular scripting language in wide use in bio-informatics;

• PostGresSQL (1985): One of the two most popular OS databases with 9% market share (2004 London Times);

• mySQL (1995): The most popular OS database with 33.3% market share (2004 London Times);

• Globus (2002): Originally conceived by high-energy physicists to promote distributed processing among heterogeneous networks, the Globus toolkit is now a key element in the caBIG (cancer bioinformatics grid) services architecture;

• Firefox browser (2004; traces to 1993 Mosaic): Market share 43% (Information Week 2005).

The following applications are not “backbones” of the Web, but may be more familiar, and are of interest to those in radiology:

• ITK (1999): NIH/NLM 3 year contract to develop an image processing toolkit, released under BSD license;

• OpenOffice (1999-2000): An office suite targeted as a drop in replacement for Microsoft Office on Windows, Mac and UNIX desktops;

• MIRC (2001): Created by RSNA, the Medical Imaging Resource Center is an extensible architecture for sharing teaching files (http://mirc.rsna.org);

• ImageJ (1997): Another gift from the NIH, this is an extensible, Java based DICOM image viewer (http://rsb.info.nih.gov/ij/);

• caBIG (2004): An NCI initiative to create multi-center cancer trials with a strong imaging component that is based on OS tools [Kakazu].

Q: I already have some free software -- that's the same as open source, isn't it?

A: No. People in the OS community have a saying, "Open-source is like free speech as opposed to free beer." What this means is, if someone gives you free beer for a week and then decides to start charging you for it, all you have at the end of that week is a slight headache. Without the recipe, you cannot choose to make the beer for yourself. Similarly, when you have free but closed source software, if the supplier chooses to start charging for that software you have no recourse but to pay (think EFilm) or go elsewhere. Further, you cannot choose to extend the functionality of the software yourself if you don't have that source.

In a nutshell, software distributed under an OS license implies that particular piece of code will always be distributed with its source code open to all eyes. As we shall see below, the particular license may or may not encumber derived works in a similar way.

Q: I've heard that if you use OS licensed software as a base for new products, those products can't be sold because the OS license on the original code taints everything derived from it to be OS as well.

A: There is not a single open source license, but several. The only constraint that they must all comply with is, “Open source promotes software reliability and quality by supporting independent peer review and rapid evolution of source code. To be OSI certified, the software must be distributed under a license that guarantees the right to read, redistribute, modify, and use the software freely” [Open Source Initiative].

Specific licenses abound. The one referred to in the question is the Gnu Public License (GPL), which does in fact require derived works to be distributed with their source [GPL]. However, the BSD, MIT and other licenses are more business (read profit) friendly [BSD, MIT].

Q: If one builds mission-critical systems on OS, there is no one to call if PACS goes down in the middle of the night?

A: This argument assumes that the decision makers at the medical center:

a) Took a suicidal leap of faith and changed products without support; or

b) That the market will not evolve solutions like Levanta - LinuxCare Data Center Automation (formerly LinuxCare) to support the OS products that are given away (http://www.levanta.com).

Obviously, neither of these events is likely. It may take some time before OS projects replace FDA approved clinical systems in private sector U.S. medical centers. However, the U.S. Veterans Administration Hospital System already uses OS clinically [Conn]. Similarly, other countries are using OS in mission critical health care [Uribe].

Q: If everyone starts giving away software, won’t commercial vendors go out of business?

A: To believe this you have to think one or more of the following:

a) A group of experienced professionals with years of experience cannot beat a bunch of (by definition) unpaid amateurs;

b) That current customers are so dissatisfied with their commercial product and/or its service that they would risk the exposures of Q4 to go it alone; or

c) That said professionals cannot out-innovate the amateur team.

To be blunt, if any of the above is true, perhaps the company so indicted should go out of business. This concern also forgets past history, such as the fact that the current commercial PACS vendors almost universally based their early products on the OS Central Test Node code released by Mallinckrodt at the 1993 RSNA. Or that IHE members benefit from the OS Mesa test libraries today.

Q: Isn't OS just socialism?

A: Perhaps in the same way that the scientific method is. Science done right is reported in respectable journals where informed peers attempt to reproduce the author’s results. Reproducible results add to the human body of knowledge, while irreproducible results go down in history as Piltdown Man or Cold Fusion. The key point is science advances with the open, transparent review of peers. Software advances the same way.

Q: What role makes sense for OS?

A: A good way to appreciate where OS makes sense is to see the types of projects where it is currently employed. As the list below makes clear, openness applies to more then just programs. It is arguably even more important in standardizing the ways in which programs talk to each other (protocols).

Applications:

• File, web and print servers

• Database servers

• Desktop/productivity applications

Protocols:

• Networking (TCP/IP, HTTP, XML)

• Database ( SQL, JDBC, ODBC)

• Health informatics (HL7, DICOM)

Languages:

• ANSI compliant Gnu C/C++

• Perl

• Python

Operating Systems:

• Linux

• BSD variants

Further candidates for OS projects are areas where there are strong standards (i.e. IHE) with well-known requirements and use cases such as PACS, RIS and basic display stations [Carr]. The value adds of OS in these cases are numerous:

• Prevents reinventing the wheel (if the project is core infrastructure that everyone needs, why code or buy it over and over);

• Leads to better quality code, more eyes, more testing, etc.;

• Leads to more stable code (i.e. if the code is in the Commons, no one vendor can discontinue it); and

• Frees talented people to work on more challenging problems where the solutions are not widely, or even, known (i.e. research).

OS projects are less well suited for areas where there is not broad consensus for the requirements or solution of a given problem. In fact, if the author of such a project attempts to “open it” it is unlikely many others in the community will embrace it specifically because there is not a wide consensus on the appropriateness of the solution. Alternately, if a vendor has solved a problem that gives them a significant competitive advantage, it is unlikely they will be willing to share it before the patent expires.

In sum, OS projects succeed best where a body of knowledge is commonly known, the solution requirements have broad consensus, and effort can be shared among users to “commoditize” that solution into community infrastructure (i.e. the intellectual Commons).

Open Source Resource List

http://www.oshca.org

http://www.openhealth.org

http://www.opensource.org

http://www.linuxmednews.com

http://www.openrad.com

http://Sourceforge.net

http://IATR.org

http://iDoImaging.com

References

BSD License information: http://www.opensource.org/licenses/bsd-license.php

Carr CD. “IHE: a model for driving adoption of standards,” Comput Med Imaging Graph JAN-2003; 27(2-3): 137-46

Conn J. Vista catches fire abroad; Providers from Egypt to Mexico are eager to put VA's IT system to work. Mod Healthc - 20-JUN-2005; 35(25): 30-1.

Feregrino C and Melendez AR. Implementing a Low Cost PACS + RIS in Mexico: First Experiences. SCAR Scientific Abstract Book 2004 Vancouver, pp. 89-90.

GPL License information: http://www.gnu.org/copyleft/gpl.html

Kakazu KK. The Cancer Biomedical Informatics Grid (caBIG): pioneering an expansive network of information and tools for collaborative cancer research. Hawaii Med J - 01-SEP-2004; 63(9): 273-5.

MIT License information: http://www.opensource.org/licenses/mit-license.php

Open Source Initiative: http://www.opensource.org/projects

Note: SCAR 2006 Annual Meeting presentations focusing on open source are available for online viewing or download.

Dr. Langer is Associate Professor of Diagnostic Physics at the Mayo Clinic, Rochester.