X-Message-Number: 33257
Date: Thu, 20 Jan 2011 19:23:20 -0500
From: "Perry E. Metzger" <>
Subject: Re: Automated Data Collection
References: <>

This gets a bit afield of the actual cryonics, but then again, so much
here does. Hopefully it isn't entirely worthless to discuss.

> From: 
> > 
> > Mike Darwin writes:
> > > MD: I'll leave this question to Brian,  because he can discourse
> > > at length on the magic that is differential  scanning calorimetery
> > > (DSC). DSC can detect  the minutest of  phase changes in a sample,
> > > and if your numbers all sum out   right, then you can be pretty
> > > sure "what's what" after cooling to any  given  temperature. I
> > > was still at 21CM when Brian began this  work, and it was
> > > horrible  - the kind of thing that would drive  me barking mad.
> > > Basically you crimp a tiny  volume of solution  'just so' into
> > > sealed metal pans, and put them in the device  and  cool them. Of
> > > course, the catch is, that you must do this thousands  and
> > > thousands of times to build a picture of how different  solutions
> > > behave under  different regimens of cooling and  re-warming. It
> > > is boring, repetitive  and truly dull   work.>>
> >
> > These days, this is what one builds small robots to do.  (That is
> > what has made much of recent science possible. No one could handle
> > the precision and boredom associated with many modern data
> > collection tasks -- automation is why they are possible at all.
> > Luckily, technology to do things like that has gotten dirt cheap in
> > the  last decade or so.)
[...]

Now, I want to be clear, I'm not actually proposing any of this is
important enough to do...

> Preparing the sample for analysis in the DSC requires (at least)
> these steps:
>
> 1) Someone has to mix up the various solutions that are to be
> evaluated. These solutions will 'one-offs' that are made up just
> for the purpose of evaluation. It's not like you are taking samples
> from various batches of medicine, soda, beer, etc., that being
> produced en masse for commerce.

There are inexpensive pump systems these days that can quite precisely
move a small, well measured quantity of fluid from a reservoir to an
output location.

> 2) The DSC 'pans' are very small, indeed tiny little metal
> containers - smaller than most hearing aid batteries. They must be
> filled with care and precision: no air bubbles can be present, and
> care must be taken to avoid contamination of the sample with water
> from the atmosphere in the case of hygroscopic CPAs (i.e., most of
> them).

Glove box + insulin needles at the ends of the tubing from the
reservoirs will probably work nicely. It is straightforward these days
to buy or build a stage that will move quite precisely that one can
put one's containers onto -- drive motors, screw shafts and the
associated electronics will cost only a couple hundred dollars for
that. Once a precision molded plastic grid to hold the containers
would have cost a bunch of time and money to produce, but now we have
rapid prototype machines that will print such things in plastic in a
few minutes given a straightforward CAD design.

Given this, one could easily place a few hundred open "pans" in
precision locations on a movable stage and fill them all with varying
mixtures very, very rapidly.

> 3) The pan must be very artfully closed and crimped, and only then
> is it place in the DSC.

That would be a bit of a challenge -- possibly an insurmountable
one. Probably the only reasonable way to deal with that is to alter
the "pan" design. I know little about this stuff and it may not be
possible at all, of course.

> The whole thing is a nightmare of careful and precise labor. And
> of course, you can't mix up 1 mL  of your putative CPA mixtures,
> because you  can't measure that precisely or repeatedly. [Actually,
> you can, but microchesmistry has its own headaches and is only
> economical if the reagents  cost a fiortune.]

It isn't so bad if you use the right equipment at this point. A
machine will happily dispense sub-milliliter quantities of several
liquids easily, and such machines are no longer expensive to construct.

> Could a machine do all this? Absolutely! Just be prepared to shell
> out $1-2 million and to spend a couple of years ironing out the
> bugs.

Oh, I doubt it. We're in an era where the controller electronics,
servos, etc. to build something like a Makerbot are now so cheap the
something like that -- which is a much more precise device than what
we're talking about here -- can cost well under $1000, even made more
or less from scratch with little in the way of economies of scale. I
know people who've built "bar bots" -- robots designed to precision
mix drinks -- as stunts for parties, on quite low (<$2000) budgets.

I suspect doing something like this correctly is a couple months work,
and the main cost would be labor, with the hardware in the noise. I'd
guess we're talking tens of thousands, not millions, of dollars.

Now, ten or twenty years ago, I'd grant you that things would have
been quite different, but they've changed a lot. Why is this possible
now when it wasn't a few years ago?

Well, several reasons. First, the cost of prepackaged microcontroller kits
capable of managing the hardware has gone through the floor -- Arduino
and similar systems now cost well under $100 and are easy even for
non-experts to program. Similar devices twenty years ago would have
cost tens of thousands of dollars if they existed without custom
design and fabrication at all.

Families of mechanical components capable of serving these markets
have become common and easy to purchase from suppliers that produce
them in heavy volume. These make the construction of simple mechanical
stages, "arms" with very simple motions, simple but precise pumps,
etc., inexpensive.

Also, the expertise needed to build simple mechanical devices suitable
for lightweight automation of this kind has become more and more
widespread.

The world does indeed make progress with time.

> As someone who has had direct experience trying to automate some
> comparatively simple things, such as cardiopulmonary bypass (CPB)

Not clear what this involves here -- surely you don't refer to
automation of the surgical procedure itself?

> (which  is very simple to automated compared to some of the
> unbelievably complex and  exactly manufacturing processes I've seen
> automated), it can be surprisingly difficult, even if you have you
> have experts, and enormous computing power at  your disposal. CPB
> is arguably no more complex than flying and landing a jetliner -
> and that process has been fully automated for at least  a decade.
> Yes, you still need a pilot, and you would still need a
> perfusionist in automated CPB, but such automation provides
> enormous added  safety (used wisely) and it also may allow for the
> use of knowledgeable, but  less skilled personnel, in emergent
> situations - such as in-field CPB for  refractory cardiac arrest.

Ah, well, here, remember you face the FDA. There are folks who do
research in one corner of my lab on medical device automation, and I
get the distinct impression that no matter how simple the device, the
cost of actually getting anything to market is astonishingly large.

The world of medicine is filled with pain for those who propose to
improve it. Even now, most doctors I deal with have poor automation of
simple business practices, and that's stuff that doesn't even require
the FDA to fix.

> [You can have knowledge without reflexes!] However, such automation
> is challenging, and even if achieved, there has to be a market for
> it. Due mostly to regulatory constraints, that market is not there
> for automated CPB. That means you have no economies of scale and
> that further drives up the price and drives down the reliability of
> any system  you do develop. Wide use = robustness and reliability.
> As to automating CPA  evaluation, well, that is so far from a
> priority for this civilization it is  impossible to put into
> words.  At least not words I can post here ;-0.

Not surprised, sadly.

Perry
--
Perry E. Metzger		

Rate This Message: http://www.cryonet.org/cgi-bin/rate.cgi?msg=33257