Interesting new project

[Phil Hobbs]

Simon and I have just got signed up to build a trans-abdominal fetal
blood oxygenation sensor--i.e. a pulse-ox for the baby that works from
outside the mom's abdomen.

The current method uses ultrasound to look at how responsive the baby's
heart rate is to the mom's contractions, in order to detect fetal
respiratory distress.

It's a very sensitive measurement, but unfortunately it's also highly
sensitive to a lot of other things that have nothing to do with fetal
distress. That leads to a lot of unnecessary caesarean sections, and
hence to a lot of harm to the patients.

This one, which we're doing in cooperation with a start-up spun off from
UC Davis, does the usual pulse-ox thing with LEDs and photodiodes
sensing haemoglobin absorption modulated by the heart rate. Since the
mom's pulse is usually slower than the child's, in principle this is
unambiguous.

The issue as usual is getting enough photons from the right depth in the
tissue, and then not corrupting them with all sorts of circuit noise.
In a highly-scattering medium like tissue, the light intensity
distribution tends to follow Laplace's equation, so it falls off steeply
with depth. For the same reason, you want to separate the source and
detector by approximately the desired sensing depth--to get 100 mm into
the tissue, you want them 100 mm apart.

I'm expecting low picoamps of signal even with big fat photodiodes.
That'll be okay for the proof-of-concept version, but after they get an
external funding round they'll want to make the probe disposable,
meaning cheeeeep, like AliExpress cheap.

Doing amazing things(*) with nearly zero apparatus is my favourite sort
of project.

Fun stuff.

Cheers

Phil Hobbs

(*) For sufficiently-permissive definitions of 'amazing', of course. ;)

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com

[jla...@highlandsniptechnology.com]

How hard can you bang the source?

You could almost do time resolution of depth. Use a gain-switched
laser and a fast receiver. Maybe average one photon per shot and use
an avalanche-type detector. That might be worth thinking about. I
guess the velocity of light drops in a scattering medium like a baby.

Does pulse ox use multiple wavelengths?

It's nice to do stuff that helps people.



--

John Larkin Highland Technology, Inc

The best designs are necessarily accidental.

[Dieter Michel]

Hi Phil,

> [...] trans-abdominal fetal blood oxygenation sensor
> [...]

> The issue as usual is getting enough photons from [...]

> I'm expecting low picoamps of signal even with big fat

> photodiodes [...] they'll want to make the probe disposable

that sounds as if effect modulation (like in "lock-in
amplifier" ...), in addition to sensing the baby's
pulse frequency, could be useful.

Would that help? And be feasible using a low-cost sensor?

Good luck for that project!

Dieter Michel

[Phil Hobbs]

About a watt, peak. Don't want to fry the baby's eyeballs if it's
presenting face forward, or burn the mum's skin.

>
> You could almost do time resolution of depth. Use a gain-switched
> laser and a fast receiver.

You need a lot of detector area, which makes fast receivers hard unless
you use something like a microchannel plate PMT. The light is diffuse,
so you can't usefully play concentrator tricks to reduce the area.

> Maybe average one photon per shot and use
> an avalanche-type detector. That might be worth thinking about. I
> guess the velocity of light drops in a scattering medium like a baby.

It's very much analogous to carriers propagating by diffusion in
semiconductors--on the order of c*(mean free path/propagation
distance)**2, with a factor of order unity to take account of 3D
propagation.

> Does pulse ox use multiple wavelengths?

Usually two, often 600- and 900-nm-ish, depending on available LED
wavelengths. You can see the Hb/Hb02 absorption spectra at
<https://medicine.uiowa.edu/iowaprotocols/pulse-oximetry-basic-principles-and-interpretation>.

It also helps a lot with rejecting room lights and stuff. You
typically phase-lock to the pulse rate for noise rejection purposes.
(This requires some smarts because the jitter is very large.)

The blood spot detector we did some years ago also sensed haemoglobin,
but needed to reject a huge overlapping absorption from the brown-egg
pigment protoporphyrin--like 10,000x stronger than the detection
threshold, which is 80 dB electrical. (That's looking for a 1%
absorption signal in the presence of a two-AU background, i.e. 99% peak
absorption from the protoporphyrin.)

That one worked by picking another wavelength on the opposite shoulder
of the protoporphyrin peak, so that the two went up and down together as
a function of egg colour, and then take one minus the ratio. We didn't
invent the scheme, but we did come up with some new calibration and
post-processing wrinkles that allowed the wavelength tolerance of the
filters to be relaxed quite a lot, which helped reduce costs. (Just
getting rid of the usual flashlamps and PMTs was the main cost-reduction
freight.)

The challenge of the present project is more in the SNR and background
rejection areas.

>
> It's nice to do stuff that helps people.

Yup. Especially mums and babies. My younger daughter is planning to
get married in the fall, so grandchildren are a strong possibility in
the next few years.

Cheers

Phil Hobbs

[Phil Hobbs]

Dieter Michel wrote:
> Hi Phil,
>
> > [...] trans-abdominal fetal blood  oxygenation sensor
> > [...]
>
> > The issue as usual is getting enough photons from [...]
> > I'm expecting low picoamps of signal even with big fat
> > photodiodes [...] they'll want to make the probe disposable
>
> that sounds as if effect modulation (like in "lock-in
> amplifier" ...), in addition to sensing the baby's
> pulse frequency, could be useful.
>
> Would that help?

Chopping the source can help reject additive background, for sure.
We'll almost certainly do that, maybe with direct-sequence spread
spectrum to make it robust against fluorescent and LED lights, PWMed
displays, variable-frequency motor drives, and suchlike vaguely-periodic
interference.

What it can't reject is the shot noise of the background, so we'll need
a black elastomer sheet as part of the probe. The modulation also has
to be slowish, because big photodiodes have a lot of capacitance.
Capacitance differentiates the voltage noise of the front end, causing a
noise current that rises linearly with frequency. In a low-light
measurement that gets ugly in a hurry.

And be feasible using a low-cost sensor?

Illumination LEDs are pretty cheap per watt, IR ones less so. There'll
be a tradeoff of source cost vs. light transport cost, for sure. We
might wind up with a disposable skin rather than a disposable probe, the
way they do with fever thermometers.

> Good luck for that project!

Thanks! Praise God for mums and babies!

Cheers

Phil Hobbs

[Joerg]

On 5/1/21 11:57 AM, Phil Hobbs wrote:
> Dieter Michel wrote:
>> Hi Phil,
>>
>>  > [...] trans-abdominal fetal blood  oxygenation sensor
>>  > [...]
>>
>>  > The issue as usual is getting enough photons from [...]
>>  > I'm expecting low picoamps of signal even with big fat
>>  > photodiodes [...] they'll want to make the probe disposable
>>
>> that sounds as if effect modulation (like in "lock-in
>> amplifier" ...), in addition to sensing the baby's
>> pulse frequency, could be useful.
>>
>> Would that help?
>
> Chopping the source can help reject additive background, for sure. We'll
> almost certainly do that, maybe with direct-sequence spread spectrum to
> make it robust against fluorescent and LED lights, PWMed displays,
> variable-frequency motor drives, and suchlike vaguely-periodic
> interference.
>
> What it can't reject is the shot noise of the background, so we'll need
> a black elastomer sheet as part of the probe.  The modulation also has
> to be slowish, because big photodiodes have a lot of capacitance.
> Capacitance differentiates the voltage noise of the front end, causing a
> noise current that rises linearly with frequency.  In a low-light
> measurement that gets ugly in a hurry.
>

Modulating the source is always helpful. It doesn't have to be tens of
kHz but it would be best to avoid the usual suspect frequencies such as
the first five or ten odd harmonics of 60Hz (such as 180Hz) and for
international markets also those for 50Hz. This could be made adaptive
so that the system first ascertains which kinds of optical noise is
present and then automatically selects an operating frequency for the
LED and the photodiode RX. Icing on the cake would a wee bit of spread
spectrum behavior in there.

> And be feasible using a low-cost sensor?
>
> Illumination LEDs are pretty cheap per watt, IR ones less so.  There'll
> be a tradeoff of source cost vs. light transport cost, for sure.  We
> might wind up with a disposable skin rather than a disposable probe, the
> way they do with fever thermometers.
>
>> Good luck for that project!
>
> Thanks!  Praise God for mums and babies!
>

Yes! Especially those who then stay home with them when they grow up. I
am forever grateful to my mom that she did. Can also be dad who stays
home like it is with neighbors.

--
Regards, Joerg

http://www.analogconsultants.com/

[Phil Hobbs]

I think we'll probably wind up with a moderate-bandwidth direct-sequence
SS system, ideally designed to have holes where the well-known
interference sources live. I've never gone through an FDA certification
process, but I expect that making the system's behaviour vary with
interference conditions could make that a lot more complicated. Is that
likely to be so?

>> And be feasible using a low-cost sensor?
>>
>> Illumination LEDs are pretty cheap per watt, IR ones less so.
>> There'll be a tradeoff of source cost vs. light transport cost, for
>> sure.  We might wind up with a disposable skin rather than a
>> disposable probe, the way they do with fever thermometers.
>>
>>> Good luck for that project!
>>
>> Thanks!  Praise God for mums and babies!
>>
>
> Yes! Especially those who then stay home with them when they grow up. I
> am forever grateful to my mom that she did. Can also be dad who stays
> home like it is with neighbors.

Yup. Maybe harder to maintain spiritual leadership that way, but there
are lots of ways to succeed. Having children raised by strangers is
usually very sub-optimal, unless the parents are hopeless.

Cheers

Phil Hobbs

[Joerg]

Generally not, as long as you can describe the behavior and show that
the energy stays within a certain band. In this case it's all just
optical spectrum anyhow when it comes to tissue exposure. Modulation
would, from a regulatory POV, come into play if if was highly pulsed.
High peak energy versus low average, et cetera, but yours doesn't sound
like that. The FDA is mostly concerned about stuff such as maximum
permissible energy into tissue and with ultrasound the max levels were
very different between adult and fetal exposure. That has to be very
well documented. Also, if you are around the limit and must maintain a
max exposure time that timing cannot be done solely by software or
firmware. It has to have a hardware limiter.

It's been almost two decades for me but the golden path was always a
510(k) FDA process which basically claims that your device is very
similar to existing and already approved technology. "It's like a mouse
trap but ours is better". If there are no predicate devices then its the
full PMA process and at least in my time that used to be super-onerous
and almost insurmountable by a start-up, financially.

We often got certified in Europe or Japan first because that was less
difficult and faster. For that reason we did clinical trials a lot over
there as well.

>
>>> And be feasible using a low-cost sensor?
>>>
>>> Illumination LEDs are pretty cheap per watt, IR ones less so.
>>> There'll be a tradeoff of source cost vs. light transport cost, for
>>> sure.  We might wind up with a disposable skin rather than a
>>> disposable probe, the way they do with fever thermometers.
>>>
>>>> Good luck for that project!
>>>
>>> Thanks!  Praise God for mums and babies!
>>>
>>
>> Yes! Especially those who then stay home with them when they grow up.
>> I am forever grateful to my mom that she did. Can also be dad who
>> stays home like it is with neighbors.
>
> Yup.  Maybe harder to maintain spiritual leadership that way, but there
> are lots of ways to succeed.  Having children raised by strangers is
> usually very sub-optimal, unless the parents are hopeless.
>

+1

For spiritual leadership that's actually the best upbringing. By far.

[whit3rd]

On Saturday, May 1, 2021 at 11:57:47 AM UTC-7, Phil Hobbs wrote:

> Chopping the source can help reject additive background, for sure.
> We'll almost certainly do that, maybe with direct-sequence spread

> spectrum...

> What it can't reject is the shot noise of the background

A factor of two is available if you polarize the light source; that can
also diminish multiple-scattered light from non-target parts of
the intended light path.

How about using a pulsed source, and delay-line-amplification
to allow the receiver to have a slow pulse response, but still
get roughly one-pulse-worth of received signal while rejecting
the long tail? EMI crosstalk might be easier to design around
than electron recombination lifetime, at high pulse frequency.
So, you let the receiver be slow.

[boB]

On Sat, 1 May 2021 14:39:21 -0400, Phil Hobbs

>presenting face forward, or burn the mum's skin.''


Aren't fetuses liquid cooled ?

[Phil Hobbs]

Sure. So are you!

[Phil Hobbs]

whit3rd wrote:
> On Saturday, May 1, 2021 at 11:57:47 AM UTC-7, Phil Hobbs wrote:
>
>> Chopping the source can help reject additive background, for sure.
>> We'll almost certainly do that, maybe with direct-sequence spread
>> spectrum...
>
>> What it can't reject is the shot noise of the background
>
> A factor of two is available if you polarize the light source; that can
> also diminish multiple-scattered light from non-target parts of
> the intended light path.

Nah, highly-scattering media destroy polarization information. The mean
free path is tens of microns.

>
> How about using a pulsed source, and delay-line-amplification
> to allow the receiver to have a slow pulse response, but still
> get roughly one-pulse-worth of received signal while rejecting
> the long tail?

It wouldn't even be that complicated--we could make the pulse as short
as we like, and just keep the rep rate down and detect slowly. Thing
is, the allowable dose doesn't go quite as fast as 1/(duty cycle), so
you wind up losing SNR instead of gaining it.

> EMI crosstalk might be easier to design around
> than electron recombination lifetime, at high pulse frequency.
> So, you let the receiver be slow.

It's the e_N*C noise that's the main issue.

[Steve Wilson]

I have a technique that I invented long ago. I call it binary sampling. It
is similar to delta modulation, invented in France in the late 1940's. I
have searched for the patent but was unable to find it.

Delta modulation samples the waveform multiple times per cycle, and
generates a single bit to tell if the waveform has increased or decreased
since the last sample.

It is very efficient since it uses a single bit per sample instead of one
or more bytes. It was used in some old Sony compact discs, in some
telemetry applications, and on the space shuttle for voice communications.
It has the problem of establishing and maintaining the voltage reference.

Binary sampling is very similar, except instead of sampling multiple times
per cycle, it allows one or more cycles between samples.

The amazing thing is this rejects noise riding on the signal. It can follow
a signal that is buried in -30 dB of noise. Since you control the sample
time, you can sweep through the entire waveform and recover signals that
would be impossible any other way.

The system works. I have already built and tested it. I am having trouble
getting google drive and sabercat to work, so I can't post the schematic
and waveforms. However, if you are interested, I could email them to you.

I am too old, and after a number of strokes, I am too weak to pursue
starting a company to commercialize this technique. So I am making it
available to anyone who can use it.

If it can help you, I would be very happy.

BTW, it also has application in GPSDO's where it can help remove the jitter
caused by sawtooth error in the 1 pps timing. It operates in real time, so
you don't need the time delay and granulation noise of current methods.





--
The best designs occur in the theta state. - sw

[Phil Hobbs]

Sounds very interesting. I'd like to have a look at it.

[Joerg]

On 5/2/21 9:07 AM, Phil Hobbs wrote:
> boB wrote:

[...]

>>
>> Aren't fetuses liquid cooled ?
>>
>
> Sure.  So are you!
>

Except our liquids often have hops in there :-)

--
SCNR, Joerg

http://www.analogconsultants.com/

[Phil Hobbs]

Joerg wrote:
> On 5/2/21 9:07 AM, Phil Hobbs wrote:
>> boB wrote:
>
> [...]
>
>>>
>>> Aren't fetuses liquid cooled ?
>>>
>>
>> Sure.  So are you!
>>
>
> Except our liquids often have hops in there :-)
>

Mine mainly ice cubes.

[Steve Wilson]

Phil Hobbs <pcdhSpamM...@electrooptical.net> wrote:

[...]

> Sounds very interesting. I'd like to have a look at it.

> Cheers

> Phil Hobbs

OK, I'll go through the documentation and clean it up a bit. A lot of the
links have gone 404 - I'll see if I can find newer versions. I also have some
new information to add. I hope to be able to send it in the next day or so.

[whit3rd]

On Sunday, May 2, 2021 at 9:14:47 AM UTC-7, Phil Hobbs wrote:
> whit3rd wrote:
> > On Saturday, May 1, 2021 at 11:57:47 AM UTC-7, Phil Hobbs wrote:
> >

> >> Chopping the source can help reject additive background...

> > A factor of two is available if you polarize the light source; that can
> > also diminish multiple-scattered light from non-target parts of
> > the intended light path.

> Nah, highly-scattering media destroy polarization information. The mean
> free path is tens of microns.

Depends on your IR wavelength; about three thousand microns of my
skin is pretty transparent to a security IR camera (lookit all those veins!)
and skin is pigmented to stop light. Amniotic fluid and fetuses are
less opaque.

I'm thinking there's scattering, but mainly forward scattering.

[Phil Hobbs]

Try it at 30 mm. (It has to go through some significant amount of
subcutaneous fat, for a start.)

And the obvious veins in your hand are large compared with their depth,
so they're still visible despite the blurring due to diffusion. You sure
aren't seeing your biceps in that NIR image.

>
> I'm thinking there's scattering, but mainly forward scattering.
>

Well, you'd be wrong then. Forward vs. reverse depends on the size of
the scatterer--Rayleigh scatter by air molecules has a dipole pattern,
i.e. a single molecule scatters into 2 pi steradians.

Multiple scattering destroys polarization for various reasons, but one
inescapable one is topological phase. If you make light traverse a path
such that the k vector encloses a solid angle Omega, the polarization
rotates by 2 Omega. See e.g.

<https://electrooptical.net/www/sed/topologicalphase.pdf>

and the Crabtree reference at the bottom of the second page. This works
regardless of how it's done, including mirrors and optical fibres.
Doing that over an ensemble of 3D paths through a scattering medium
homogenizes the polarization pretty thoroughly.

[Steve Wilson]

Phil Hobbs <pcdhSpamM...@electrooptical.net> wrote:

> Steve Wilson wrote:

[...]

> Sounds very interesting. I'd like to have a look at it.
>
> Cheers
>
> Phil Hobbs

Did you get my emails? I sent them on Monday to your electrooptical.net
address. There were two. The second is more important.

Do you have a spam filter that rejects attachments or large files? I notice
you have no protection on your email address. You must get swamped with
offers from Nigerian princes:)

There are a number of ways to obfuscate your email address to block
spambots. Here's one that uses javascript, but leaves the "mailto" prefix
exposed:

http://www.email-obfuscator.com/

The problem with this is spambots also use javascript, and they search for
the "mailto" prefix.

I wrote a bayesian filter to block spam, but it never worked very well.
Spambots have an infinite variety of ways to get past these filters.

I like this approach much better since it also encodes the mailto:

https://www.hcidata.info/obfuscate-email-address.htm

After switching to email encoding, I never got a single piece of spam from
my online address in 20 years. Of course, that didn't help block spam from
vendors I purchased items from. So I had to use a different email address.

For those, I use disposable email addresses at

http://www.e4ward.com/

It never fails. If a vendor persists in sending you junk mail, just delete
the address you assigned to him.

Other email forwarders turned out to be unreliable, or do not allow you to
assign the email address yourself. E4ward. com has never lost an email, but
some sites reject the forwarding address and want you to give a real one.

Then it is a choice if I want to do business with them or not, and how to
persuade them to stop sending me junk advertisements. I usually regret
giving them my real address.

Let me know if you got my emails:)

[Phil Hobbs]

Yes I did, thanks. I'll look at them in the next day or so.

[Steve Wilson]

Phil Hobbs <pcdhSpamM...@electrooptical.net> wrote:

> Steve Wilson wrote:

[...]

>> Let me know if you got my emails:)

> Yes I did, thanks. I'll look at them in the next day or so.

> Cheers

> Phil Hobbs

Thanks for the reply. I wanted to get them to you as soon as possible in case
they mght help with your project.

The noise rejection is real. I don't know if the dribble-up is due to thermal
tail or skin effect, or both. It may have no effect on your short pulses.

[Tom Gardner]

On 04/05/21 16:45, Steve Wilson wrote:
> I wrote a bayesian filter to block spam, but it never worked very well.
> Spambots have an infinite variety of ways to get past these filters.

The filter in Thunderbird/Seamonkey is extremely effective:
very few false positives and negatives.

[Phil Hobbs]

Except that they seem to have gutted it lately--I used to rely on
"anywhere in headers", and that no longer exists. :(

[Tom Gardner]

On 05/05/21 15:06, Phil Hobbs wrote:
> Tom Gardner wrote:
>> On 04/05/21 16:45, Steve Wilson wrote:
>>> I wrote a bayesian filter to block spam, but it never worked very well.
>>> Spambots have an infinite variety of ways to get past these filters.
>>
>> The filter in Thunderbird/Seamonkey is extremely effective:
>> very few false positives and negatives.
>
> Except that they seem to have gutted it lately--I used to rely on "anywhere in
> headers", and that no longer exists. :(

I was thinking of the "automatic" Bayesian filter that is updated
when I declare an email is (or is not) junk by clicking the "Junk"
(or "Not Junk") button.

Apart from that, I run SeaMonkey/Linux 2.49.4, and that still has
email filters for "From To Cc BCC" and others. That's sufficient
to enable me to put messages in specified folders