alt Hacker NewsFunctional ultrasound through the skull
Comments
“The thing that nobody tells you is that you can buy a real human skull online (shoutout to skullsunlimited.com). We did that, and then CT scanned it.”
This is an A+
This is fun, and the modeling is cool for sure, but it's well known that ultrasound can be used with surgical precision in the human brain.
Focused ultrasound is already used for non-invasive neuromodulation. Raag Airan's lab at Stanford does this for example using ultrasound uncaging.
https://www.frontiersin.org/journals/neuroscience/articles/1...
https://www.sciencedirect.com/science/article/pii/S089662731...
Also see the work by Urvi Vyas, eg
https://pubmed.ncbi.nlm.nih.gov/27587047/
I don't mean to discount the cool imaging-related reconstruction of a point spread function, but rather to say that ultrasound attenuation through the skull an soft tissue has already been well characterized and it's not a surprise that it is viable to pass through.
Also check out their other post: https://brainhack.vercel.app/ae
They are planning to locally change the electrical conductivity of brain tissue by focused ultrasound, modulate that with at few hundred kHz and do a lock-in (EEG) measurement to deduce electrical activity at that spot on the scale of 1mm. Pretty wild if that actually works.
'Previous work showed that tofu is desirable as a phantom material, both because it is fast to get and because it has similar physical properties (density, speed of sound) as soft tissue.' Haha wonderful.
Progress in making measurements through the skull useful might be how we finally get to precisely measure side-effects elsewhere: comparing healthy adult skulls to proper control groups. Always seemed odd to me how unspecific the thermal safety limits are, though the peak is expected depend on localized unknowns.
love the writing in this
"In physics, there's a word for 14 orders of magnitude of attenuation. It's called zero, i.e., you will measure nothing."
Lots of great sentences in here as noted in the other comments.
Nice timing, tomorrow I'll be participating in a study doing transcranial ultrasonic neuro-modulation, meaning using ultrasound not just to map brain activity but to influence it (the point of the study is inhibiting the Default Mode Network).
If anyone's interested I found those two paper really interesting:
- Aubry et al 2023[1], on potential risks and limitions of using focused ultrasound in the brain (tldr we don't know but have conservative estimates. Really interesting for me to see that HN article adding to that)
- Lord et al 2024[2], a first study on using Transcranial Focused Ultrasound to modulate the DMN and subjective experience
[1] https://arxiv.org/pdf/2311.05359
[2] https://www.researchgate.net/publication/381488518_Transcran...
Awesome! I know of efforts to leverage focused ultrasound to shorten sleep cycles and improve mental health. There’s so much more possible in neuroscience, great to see this work is gaining steam.
By the way Bryan Johnson recently mentioned a novel technique of using ultrasound to de-calcify pineal gland for improving sleep as it gets worse with aging - this sounds really intriguing to me.
It mentions using this as a computer interface but wouldnt prolonged use eventually open up the blood brain barrier?
In case you didn't scroll to the end, they open sourced their code:
How do you degas a skull in a living mammal and have an unharmed animal after?
Very neat.
The big question for me is - how will it feel on a live person. Is it going to be painful? Could it alter/damage the brain tissue?
So basically - how safe is this tech?
22 dB/cm/MHz, 1.4cm, 10 MHz gives ~40dB, which happens to be 4-orders of magnitude difference in power. Not sure how they got 14 orders of magnitude for the attenuation.
I don't understand why they believed in such high attenuation numbers.
They quote a book that the public at large (including me) can not check for the 22dB/cm/Mhz number.
The next best quote is the 8.3 dB/cm/Mhz quote. That article is available to the public:
https://pmc.ncbi.nlm.nih.gov/articles/PMC1560344/pdf/nihms94...
However I don't see any expression claiming a linear frequency dependence of attenuation up to 10 MHz.
> the number people will tell you in conversation.
Is very vague. Do your own research, find actual measurement data, don't extrapolate a few sub MHz measurements out to 10 Mhz, especially not if the error bars become ludicrously big.
Since I can't find the quoted frequency coefficient for attenuation I look at the possible candidates in that article: there's Figure 11, Table 1 and Table 3.
My gut feeling tells me they used table 3:
frequency in MHz | Longitudinal attenuation in Nepers per meter
0.272 | 14 +/- 17
0.548 | 53 +/- 43
0.840 | 70 +/- 28
I suspect they discarded the middle frequency because of the large error bar, so they are left with
Mhz | Np per m
0.272 | 14 +/- 17
0.840 | 70 +/- 28
the difference in frequency is 0.568 Mhz
so the difference in attenuation is then 56 +/- 45 Np per m. Yes the standard deviation is almost as large as the value. Let's see if we arrive close to their supposedly "quoted assumed linear frequency dependence of 8.3 dB / cm / Mhz "
2 x (56 Np per m) / ( 0.568 MHz x 100 cm per m x log(10) Np per 10 dB)
= 8.56 dB / cm / Mhz
close to their 8.3 "quote" which is really their own deduction, or whomever "derived" it in "conversation".
If you calculate the error bar: 8.56 +/- 6.88 dB / cm / MHz.
What they independently measured (props! actually good science):
11.18 dB / cm / MHz
Thats 2.62 / 6.88 = 0.38 standard deviations away. Thats not new science in the sense of hypothesis rejection, but a valuable extra datapoint refining the literature of values.
The likelihood of measuring a value 0.38 or more standard deviations away from the expected value would be: 70.4 % so not very surprising at all. Basically in conformance with the 8.56 +/- 6.88 dB / cm / MHz value.
https://www.mathportal.org/calculators/statistics-calculator...
Yeah I did a (mostly failed) PhD on ultrasonic imaging and found many things that worked in simulations but not in practice. The fancier your imaging algorithm gets the most ill posed it becomes and more sensitive to noise and errors.
Even if you add noise to your simulation , when you go to the real world it will have lots of sources of noise and errors that you didn't model. In this case I suspect aligning the CT scan with the ultrasound probe will be extremely difficult.
Also there's a reason ultrasonographers are so highly paid, and it's mostly used for pregnancies. In normal tissue it kind of sucks as an imaging method. (On an absolute scale; obviously it's amazing technology.)
Eh maybe it will work though. You never know.
I recently had the idea to start a company that measures specifically properties of the pineal gland, I think people would pay for that. I have no domain expertise whatsoever. If anyone wants to investigate this deeper with me let me know
Surely this is easily solved with time-reversed acoustics. Just stab a transmitter into the brain with an ice pick to the point you want to measure, and pick up the signal at lots of locations around the skull. Now you have both a mapping from an input signal (the reverse of the signal you picked up) that you can send to precisely target that point, and you know it looks like after it comes out from that point (the original signal you picked up).
Now you can tell exactly what is going on and the person is thinking! Specifically it'll be either: (1) "oh my god, I have an ice pick in my brain" or (2) nothing, because they have an ice pick in their brain.