[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Status of uw-prism packaging for Debian

Hi Andreas,

Answers to your extremely important questions below.

On 08/18/2014 11:38 PM, Andreas Tille wrote:
Hi Ira,

On Mon, Aug 18, 2014 at 01:42:51PM -0700, Ira Kalet wrote:

I had the good fortune at a recent social event to meet an attorney
for one of the big multinational corporations in the medical device
business, whose job is to look after FDA regulatory issues.  He even
was familiar with RTP systems.  He agrees with the view that if the
Prism software is not represented as a ready-to-use device or
system, and is only distributed as a source code library, then no
FDA involvement should be necessary.

Thanks for your continuous effort into this.

The way Common Lisp code is packaged in Debian appears to be most
suitable to this anyway.

I have to admit that I do not have the slightest idea how Common Lisp is
handled in Debian and that I feel incompetent to give a sensible answer
to your question.  So please take my answer with a grain of salt.  It
might be best to contact the Lisp developers list[1] about Lisp details.

Building a runnable binary is highly Lisp
vendor specific and would be difficult to automate.  But treating it
as a library from which someone really knowledgeable could build a
binary on their own would be relatively easy.  I would support this.

In general a Debian package tries to deliver a ready to run program (if
this is the nature of the software and if it is not just a library).
However, I'm not only Lisp-illiterate I have only a very vague
imagination what Prism finally is and how use cases of this software
really look like.  So it might perfectly be that the term "library" fits
better than a ready to run program.

BTW, in the consequence of our discussion another idea popped up in my
mind.  I have no idea whether this is possible but are there chances
that a patient suspects some mistake in calculating radiations and might
go to court and sue the person / hospital claiming they might have
caused some harm to him.  Is there any chance that this will fire back
to Debian / the Debian maintainer in some way?  If there would be a
chance for this we should be quite carefully in wording the description
for a potential prism package.

For instance to declare the program (library?) only fit for "educational
use" or something like this.  In case we should decide for such a means
I guess we would be done also with FDA issues since we are not shipping
a device but rather a program to demonstrate something for educational
purposes.  (If people might use it for real calculations anyway - that's
their problem.)

What do you think about this?

I know way more than I would like about this. I recently served as an expert to two law firms, each acting as attorney for the estate(s) and families of several people who were in fact killed by radiation misadministration. Errors in the design and configuration of RTP software and related products were part of the problem. In my opinion the manufacturers of the software were at least partly responsible, along with the hospitals and medical physicists who purchased and operated these systems. These kinds of incidents made international news back in the early 1990s, with injuries and deaths due to faulty software in a computer controlled radiation machine, the Therac-25, manufactured by a now defunct Canadian company, AECL. The story is told in a book by Nancy Leveson, "Safeware: System Safety and Computers", (1995) and is also discussed in chapter 6 of my book, "Principles of Biomedical Informatics", 2nd edition (2013). Despite the fact that these kinds of issues were understood 20 years ago, manufacturers are still not doing a good job with these products. Regulation by the FDA has had minimal impact. The devices that I reviewed in the lawsuits all had FDA 510K clearance. If I had been an FDA staffer, they would not have been approved, but I am just a poor physicist...

Although Prism has been in use for 20 years at the UW and we have never had a misadministration due to errors in Prism, that does not guarantee it won't happen. So, yes, the safe way to proceed is to make no claims as to fitness for clinical use, and to only claim that the program is for educational purposes.

Sorry to have taken the time of so many people to sort this out.

Well, sometimes it needs careful discussion to create a system which can
be used in health care.  I do not think that this was extraordinary long
discussion compared to the complexity of the problem.

If this source only strategy is OK with you all, what are the next steps?

As far as I understood previous mails of yours we need some
preconditions for prism in Debian anyway.  What about simply starting
with packaging these preconditions.  Even if we might finally decide
that we do not deliver a prism binary package (but just provide
packaging stuff to make it simple creating a Debian package with very
less effort) we can turn Debian into a system where it is brain dead
easy to install Prism onto.  Perhaps it is a good idea to contact the
Lisp developers via their mailing list[1] how to start with these Lisp

I would not want to make it "brain dead easy" to install Prism. That is exactly how a disaster can happen. It is like handing a loaded gun to a small child. I have written some relatively simple instructions for building a Prism system from source code, as part of the User Reference Manual, but they presume you understand how Common Lisp programs work, and also how RTP systems work. This software is not a drawing or CAD tool or email client, or word processor. It is complex and dangerous. Even having a turnkey running system still requires creating some extensive supporting data files, that have to be in the right format, with the right data, which depends on the particular radiation therapy machinery you are simulating. Using the system requires substantial knowledge of the process and concepts of radiation treatment planning.

I'll start with one of the components of Prism, the user interface toolkit library, SLIK, which is benign, has nothing to do with medicine, and which should be an easy way to learn what the packaging requirements are. After that, the technical issues of packaging Prism should be relatively clear.

Hope this helps and thanks for your effort


You are welcome.

[1] http://lists.alioth.debian.org/pipermail/pkg-common-lisp-devel/


Ira J. Kalet, Ph.D., FACMI
Professor Emeritus, Radiation Oncology
Professor Emeritus, Biomedical Informatics and Medical Education

Email: ikalet@uw.edu
Web: http://faculty.washington.edu/ikalet/

Reply to: