The rationale for the HP47201's eight wavelengths is in their paper, but it sounds like a big part of that rationale is to force the earlobe of whatever skin tone into behaving in aggregate like a system where the assumptions of the Beer-Lambert law apply well, with the incandescent-lamp-and-color-wheel tech they were using at the time, which probably wasn't emitting very spectrally pure "monochrome" light
The earlobe is the wrong location, because there's so little arterial blood. Fingertips are the ticket.
Cheers
Phil Hobbs
Probably want to get some kind of computer scientist to think about the problem:
The rationale for that location seems to be that it was the least sensitive to patients in shock; this machine seemed to have been designed with trauma patients as the primary concern.
The "DSP" looks to have been done with PROM-based state machines and discrete logic as was the fashion at the time.
Find it hard to believe that a modern hospital doesn't have similar higher-accuracy equipment still available, that by rights they should've been using for diagnosis, but during Covid there probably wasn't enough of it to go around.
Hey, there's a cool amplitude-stabilized sine shaper on page 10, though it only works at a single frequency. Wonder if you could sub a shunt crystal for the series LC.
Kinda late to the party. I was going suggest outside /other (maybe acoustic) detection of the two pulses. (like using an external light source chopping wheel with sensor as the reference input on a Lockin.) Martin was there first. I was also confused how the notch worked on a signal with changing frequency. Then I had this silly idea, if you know when the mom beats are you could stretch or shrink the time base between pulses so that they all had the same period. Then FFT and subtract the harmonics. Then un-apply the time stretching. And then do the same for the babies pulse?
George h.
If you resample so that the samples are equally spaced in phase rather than in time, the frequency of the pulse isn't changing by the time the filter hits it.
That's more or less the algorithm I posted. ;)
The issue of how to handle the interaction between the two pulses when they're close to harmonically related is still open. Martin's suggestion of basically ignoring the intervals where the two systoles collide may work if the diastoles are long enough. (The pressure pulse isn't as sharp as the ECG signal, and smears out in the smaller arteries.)
Cheers
Phil Hobbs
Hey, Google, separate out these two heart rhythms.
And it is a bad approach. The heart pulse is just that - a pulse of blood that gets injected into the aorta, which has to stretch a bit to accomodate the extra volume of relatively fast-moving blood. The pressure wave - and the extra blood - propagate down the aorta as a pulse in a - rather dispersive - transmission line.
The pulse gets broader as you move deeper into the circulatory system and out to the extremities, but its velocity isn't going to be changed by the pulse rate.
You've got electrocardiogram signals that can tell you exactly when the pulse was put into the system. Getting them for the fetal heart is technically difficult, but it is done as a matter of routine.
Why bother? The fetal circulation is driven by the fetal heart.
Of course it isn't as sharp as the ECG signal - the nerve impulses are driving the muscular contraction that starts the pressure pulse, and the pressure pulse does get smeared out as it propagates out to the extremities, but it is the source of all the variation you are looking for and at.
The problem is separating the fetal pulse from the maternal pulse, which does dominate any signal you can pick up outside the mother's body.
It's a demanding problem with electrocardiogram signals, but one that has been solved.
The "smart speakers" are doing Doppler sensing on an 18 to 20kHz acoustic wave in air, and getting the fetal pulse from displacements of the surface of the maternal abdomen might not be all that easy.
OK good.
Hmm, how about if the gizmo checks for a harmonic relation, and throws that data out if there is one. Make another data run. Maybe tickle mom to change her heart rate?
George H.
That might work much of the time. The rub would be if it happened when the child is in respiratory distress.
A lot of this stuff is in the suck-it-and-see category, I expect. I'm trying to get some raw data from the customer to play around with. (They did some measurements on a sheep, where they could combine all sorts of fetal and maternal measurements--glad I wasn't part of that one.)
Cheers
Phil Hobbs
How are the two pulses supposed to "interact"?
Phil's problem seems to be that his optical scheme illuminates both maternal blood - where the level oxygenation varies just with the mothers pulse rate - and fetal blood where the mothers pulse rate will have some effect (mainly on the amount oxygen in the maternal blood in the placenta) but the fetal pulse rate will vary the oxygen levels of of the fetal blood as oxygenated blood collected from the placenta is pumped out into the fetus to get deoxygenated in it's extremities.
If he gets variations that correlate with the fetal pulse rate, he's looking at blood circulating in the fetus, which is presumably what's interesting
My feeling is that the mother's pulse rate is just noise superimposed on the fetal signal that you want. Pulling it out may be useful to get some idea of the amount of maternal blood getting into the placenta - I don't know what the clinicians are looking for. They may not either.
Have something to add? Share your thoughts — no account required.
Ask the community — no account required