Neural engineering without formal engineering backgorund?

[SpanishMusical]

Hi all,

First off, I apologize if this is in the wrong place; please move it as necessary. I'm finishing up my last year of undergrad and as I've been applying to schools, I've been finding myself getting more interested in PM&R and especially brain-computer interfaces. I'm not particularly interested in device development or design, but moreso the decoding of neural signals from implanted electrodes, or using feedback from these devices to stimulate sensation and help people better use these devices. The problem is, I have a limited quantitative background, andno formal engineering background. My math background is Calc I and II as well as linear algebra; CS is intro to programming, AI, machine learning (a more mathy look at the algorithms), and a neural networks course. I'm doing research in time-frequency analysis of EEG signals and applying machine learning (clustering) to my results. I've coded a lot of this in MATLAB, and written basic programs to automate EEG flatline detection (when electrodes would drop out in recordings) as well as used dimensionality-reduction methods (PCA) to handle all the data. That said, my major is in neuroscience, and it's a bit too late to take any more undergrad classes. Have I locked myself out of working in neuroprosthetics/neural engineering, or is there still hope to get involved in this research in med school/after?

Not sure who would know the answer to this, so @Neuronix ?

---

Members don't see this ad.

 

[Neuronix]

That's a pretty niche area and it's gonna be program dependent. I'd keep a fairly open mind in neuroscience, write your essays for neuroscience and give a flavor of human/machine interface work on interviews and see what opportunities you get. The details of how you get there (i.e. through a neuroscience PhD, bioengineering PhD), and whether those will be open to you as a graduate student depends on the program.

 

[deleted728364]

I am just a postbac planning to apply but yeah I'd say it sounds like you're getting the experience/skills necessary to do neural engineering. Definitely haven't "locked" yourself out. I'm gonna push back on the notion that this field is niche or esoteric. Systems neuroscience is booming. I think you should aim to (1) keep up w literature to figure out which PIs you want to work with (and which unis have the best communities for these people) (2) make sure youre doing a great job in the lab right now so you can get great LORs to help your chances at admission to those unis (and gpa, mcat etcetc) and then just hope you get accepted to a place that provides a good community for this research.

BTW imo Pitt is really the best place for this stuff. Byron Yu, the RNEL lab, and Aaron Bautista do super cool stuff. Lots of great PIs to find in the references of this review: http://stat.columbia.edu/~cunningham/pdf/SaxenaCONB2019.pdf I can send you more review articles if you want.

Also check out LFADS by Chethan Pandarinath. As well as Krishna Shenoy's lab. If you can understand the math there and across this literature then you should have the skills necessary. FYI everyone uses python these days, a lot more flexible than matlab. (also the people/article i mentioned focus on latent factor analysis and dimensionality reduction/manifolds. thats my exposure to neural engineering/BCI stuff so it may be a bit biased from where the field is).

Also check out the COSYNE conference. it's more theoretical and less applied engineering focused but you may find it interesting. again if you can keep up with this stuff then you have the skills. It definitely takes time to learn the methods/jargon and understand everything, but just keep at it and youll be fine. you already have exposure to ML/DL, so you should be fine. If you really want to "prove" this to yourself, perhaps try emailing around for a gap year position at these labs. Though thats kinda hard given COVID hiring freezes :/

Also twitter is an excellent tool to find more researchers in this field. You'll find that there are a lot more than you thought.

@AZDev, paging you once again

 

[deleted728364]

but moreso the decoding of neural signals from implanted electrodes, or using feedback from these devices to stimulate sensation and help people better use these devices

yeah, definitely check out COSYNE and the PIs whose work is related to (cited by, who cite) LFADS. Theres a huge gap between being able to throw an ML alg for EEG data together on matlab and really understanding this literature and being productive in one of these labs. So i'd start bulking up on the literature and picking up skills along the way.

The problem is, I have a limited quantitative background, andno formal engineering background.

everyone just picks these skills up by self studying. your major really doesnt matter. youre gonna be self studying stuff for the rest of your life if you do a research career i feel like. highly recommend strogatz's nonlinear dynamics book and kailath's linear systems book.

 

[AZDev]

"I'm finishing up my last year of undergrad and as I've been applying to schools, I've been finding myself getting more interested in PM&R and especially brain-computer interfaces."
What major / minor / concentration are you? (I saw neuroscience, but any other info?)

"I'm not particularly interested in device development or design, but moreso the decoding of neural signals from implanted electrodes, or using feedback from these devices to stimulate sensation and help people better use these devices. The problem is, I have a limited quantitative background, and no formal engineering background."
I don't think this is an issue if you're not interested in the device itself. There are plenty of COTS devices out there for researchers and definitely a place for a computational researcher in the space without engineering experience (especially with a strong computing background).

"My math background is Calc I and II as well as linear algebra; CS is intro to programming, AI, machine learning (a more mathy look at the algorithms), and a neural networks course."
All great things from my POV for this kind of work. Especially linear algebra.

"I'm doing research in time-frequency analysis of EEG signals and applying machine learning (clustering) to my results. I've coded a lot of this in MATLAB, and written basic programs to automate EEG flatline detection (when electrodes would drop out in recordings) as well as used dimensionality-reduction methods (PCA) to handle all the data."
Sounds like signal processing software engineering / computer science is your game. Given the complexity of the kind of work you're doing, you're already well on your way to proving your case for competence in that realm IMO.

"Have I locked myself out of working in neuroprosthetics/neural engineering, or is there still hope to get involved in this research in med school/after?"
I think you have plenty of options as long as you target things you're competent in doing.

Re: Schools
Pitt/CMU is great for this in general. I'm biased b/c my research interest is in sensory perception, but Case Western has a great set of PNS researchers in the space of touch sensation, as does the University of Rochester in CNS (to a more limited extent). MIT is working on decoding / stimulating proprioception too (accessible through the Harvard HST). Plus there's plenty of work in pain, vision, and hearing in general, all of this work requiring computational efforts. I would be surprised if you could not find a place for yourself in one of these groups if you really showcase your passion.

I do not agree that everyone is using python. That being said, it is extraordinarily common due to its relative ease of use, creation, and modification. Languages at lower levels are often used for tech at the far edge of neurotech's capabilities (since it gets really hard to use hardware to improve speed of processing at those scales).

 

[SpanishMusical]

Thank you all for the encouragement and responses! I will absolutely check out the LFADS papers as well as the other ones attached. Any other good review articles, @pretysmitty ?

@AZDev : about your first question, technically I'm majoring in neuroscience and in Spanish (separate majors) with a minor in chemistry. I found this interest a bit late in the game, so I've only recently started beefing up my quantitative skills (neural nets, ML, linear algebra). My research experience is basically as I described it in my first post.

 

[AZDev]

Thank you all for the encouragement and responses! I will absolutely check out the LFADS papers as well as the other ones attached. Any other good review articles, @pretysmitty ?

@AZDev : about your first question, technically I'm majoring in neuroscience and in Spanish (separate majors) with a minor in chemistry. I found this interest a bit late in the game, so I've only recently started beefing up my quantitative skills (neural nets, ML, linear algebra). My research experience is basically as I described it in my first post.

I'm not remembering the names right this minute but you might find it interesting to read about the neural interfaces researchers working in Spain. There are quite a few that regularly show up at the relevant conferences in the USA. Could be a good application of your Spanish skills to look into the possibility of collaborations in your future work.

 

[deleted728364]

FWIW (you sound like you already have a good foundation) but if you want a good reading list, I'd check out this rather hilariously written "Super Harsh Guide to ML". Only change i would add is doing the coursera first and then try out "Introduction to statistical learning" if ESL is a bit too much (pretty dense book). Also, i'm kinda crap when it comes to Bayesian stats so im open to suggestions for reading if anyone has anything.

Any other good review articles, @pretysmitty ?

Hmmmm i think you search on google scholar for "neuroscience [BCI]" or dimensionality reduction, manifolds, decoding etc and look for highly cited reviews you'll be good. The articles i initially sent will cite some good reviews as well.

but heres a big ol list someone sent me a while back. still relevant imo:

• Neural data science: accelerating the experiment-analysis-theory cycle in large-scale neuroscience
• Dimensionality reduction in neuroscience
• Dimensionality reduction for large-scale neural recordings.
• Computational principles and models of multisensory integration
• Computational Neuroscience: Mathematical and Statistical Perspectives
• On simplicity and complexity in the brave new world of large-scale neuroscience
• Computational training for the next generation of neuroscientists
• Using computational theory to constrain statistical models of neural data
• Big data and the industrialization of neuroscience: A safe roadmap for understanding the brain?