A lawsuit alleges Apple Watch's pulse oximeter is racist

Apparently, Massimo has done that research and verified that, while their SET technology does give a statistically-significant different result for patients with darker skin tones, that difference is not clinically significant and below the level of precision of the display (0.15% difference on a display that shows measurements in 1% increments).

5 Likes

Unlikely. Massimo’s sensor hardware costs single-digit $ as a disposable in hospitals, and the SET algorithm would likely have a negligible impact on cost of an Apple Watch that costs hundreds of dollars and has millions of units sold per year.

Again, it’s not a matter of whether they can it’s a matter of whether they care.

5 Likes

Are these sensor bits designed to plug into any thing, or are they entirely self contained? Is massimo selling 15 dollar razors blades with thousand dollar handles?

2 Likes

Nope. That’s not how the medical device market works. Depending on the company and their position in the market, they can sell $$$$ capital equipment with $$$ disposables, $ capital equipment with $$$$ disposables, or free capital equipment with contract to buy $$$$$ disposables.

2 Likes
1 Like

If Apple stated “your mileage may vary” or more specifically, “this doesn’t work in all cases for all people” would that satisfy the argument? Like so many things, the pulse ox is a cool feature that may or may not work for everyone. If they don’t make straps big or small enough for every single wrist, should we sue? There are many caveats like this that don’t make for lawsuits. They make for an understanding that people come in all varieties and one size never fits all. I’m 6’3" tall but have never thought of suing anyone because their product (jeans, car, airline seat, hotel bed) doesn’t fit me. I just move on (annoyed) and make different choices.

Still no. Historical and systemic anti-tallism isn’t a thing. Systemic racism most certainly is.

5 Likes

I suspect (again not an expert, so :salt:) that it may not be solvable with the discount sensors and limited CPU that the Watch is using.

The problem with smartphones and watches right now is that they are increasingly desperate for new applications. They solved email, messaging, cameras, etc. All the basic stuff that is pretty easy for a computer to do anyway. Then they got really good LEDs and managed to add Flashlight to the list. However now they’re getting into sophisticated sensing and niche verticals like medical diagnostics. They’re getting out of their league here, but they think they can solve it by hurling big piles of machine learning at it to compensate for the cheap hardware and shitty signal-to-noise ratio. It’s not really working, but consumers assume it does work or they wouldn’t sell it, which is a problem.

5 Likes

Apple has an article on the pulse oximeter with all sorts of details.

4 Likes

Apple Watches are commercial products people pay money for

Apple literally owes every customer every advertised feature

That’s how commerce, and “owing,” work

8 Likes

The reflectance measurement is more difficult than the transmission measurement, but the sensor components are just wavelength-paired LEDs-photodiodes. They are pennies each at volume.

You have a good point about the processor. Basic pulse oximetry can be run on an Eprom (I did that in college) but the advanced algorithms that Massimo runs for SET might be more hungry.

3 Likes

The cpu in the apple Watch is pretty hefty.

It’s hefty but also pretty busy. Can it run those algorithms in the background, without affecting all it’s other functions, 24/7? Maybe.

3 Likes

Here’s the text of the lawsuit btw.

It seems less than precise, but IANAL.

Fingertip pulse oximetry places the light source on one side of the finger and the sensor on another. As I understand it the SET trick involves using radial polarized light to avoid some of the extra absorption due to skin tone. Wrist pulseox (which I understand far less) are based on reflectance so the solution could be quite different. Also wrist pulse ox is challenging, sensors move around, ambient light gets in etc… Even good (consumer) ones seem to have a >5% error, so trends may be more useful than the inaccurate absolute values.

That being said…
I have a garmin with pulse ox and I turned it off. It just eats battery while providing minimal usefull info. I have heard it may have some application for athletes going through altitude aclimation. But for most of us, we can get far more usefull training info from the many other stats our sports focused trackers record. Personally this feels like a tool with shiny marketing in search of a use case…

If you have a medical concern a validated medical device with skilled interpretation should really be your first choice in most situations.

(Edit in the 3 hours it took me to hit post I see you mention absorbption vs reflectance above. I did want to add that the radius disposable sensor system is attached to a larger wrist device which then tslks to a larger post mounted monitor. where the processing occurs isn’t clear, and I have no idea if pulseox destroys watch battery life from processing or just running the LED/sensor)

3 Likes

The fact that this technology works better on light skin than dark skin isn’t racist in itself, but the fact that a more reliable method hasn’t been invented or become the standard yet is at least partly because of systemic racism in the medical industry, and the fact that Apple doesn’t bother to clearly note the technology’s limitations so that dark-skinned customers can take that into account when buying their product is evidence of their systemic racism. Just because it’s impossible to make everything one size fits all doesn’t mean it’s okay to mislead customers. “Pulse oximeters don’t work well on dark skin” isn’t the same level of common sense as “these jeans aren’t my size”. It’s a form of bait-and-switch, and while I wouldn’t go so far as to say it deliberately targets people of color, it ends up affecting primarily them in a way that’s passively racist.

6 Likes

shocked philip j fry GIF

maybe it’s time to stop treating white dudes as the default human, and figure out a way to be more inclusive on stuff like this.

Super Troopers Yes GIF by Searchlight Pictures

It’s getting really tedious having people swoop in and explain how the lived experiences of millions of people and actual facts are not “real”… Some people seriously need to shut up and listen.

Anthony Bourdain Yes GIF by Ovation TV

I don’t think this stuff is a mystery. Just fucking calibrate these things on a wide variety of people, instead of just some random white dudes. It’s not particular difficult. and never has been. but as long as people see white dudes as default, THIS IS GOING TO KEEP BEING AN ISSUE. And YES, THAT itself IS part of systemic racism.

It’s just racist. Let’s stop qualifying this shit and call it what it is.

4 Likes

It’s also heavily throttled for battery life reasons. It may have a lot of horsepower on tap, but it’s not going to get to use it for any substantial amount of time with a battery the size of a penny.

1 Like

I’ve read one of the papers on this topic and I know some people who worked on this technology for decades. There are two different types of sensor configurations: transmissive and reflective. From what I read in the paper, nearly all of the evidence proposing that skin color is a significant factor in accuracy were specific to the reflective type rather than the transmissive type. The paper, however, doesn’t acknowledge that distinction and lumps both approaches together.

I’ve never seen an Apple watch with this feature, but I assume they use reflectance.

It’s is harder for a transmissive sensor to get a strong signal through darker skin, as it attenuates both the red and infrared wavelengths. But the measurement comes down the ratio of the amount of each wavelength transmitted, so, in theory, the skin color should not bias the result, though it may be more prone to noise if the device cannot get a strong enough signal.

For actual medical devices, the FDA has long required manufacturers to include a spectrum of skin colors in their clinical trials. Consumer grade oximeters aren’t very good in general.

SET doesn’t use radial polarized light (that’s an emerging technology). SET just uses advanced signal processing.

The Radical disposable sensor uses standard light sources and photodiodes. The reusable wrist adapter converts the SET signal so it can be used as an SpO2 source for non-Masimo monitors, so the signal processing has to be done at the level of the adapter (though that might be an inadequate term for it).

Masimo has a version of SET for their consumer watch product, but it’s different than the 2 LED, 2 sensor setup that the Apple Watch and Garmins have. The back of the watch looks like the lighting setup for a rave…

2 Likes