Show HN: A software solution to ultrasound blood flow monitoring | Hacker News Reader

Show HN: A software solution to ultrasound blood flow monitoring

bigwww.epfl.ch

85 pointsrscho21 comments

For those interested, this is destined to any situation where monitoring blood flow is useful, in any vessel available to ultrasound imaging. HN offered me a repost of this story due to my comments at https://news.ycombinator.com/item?id=13423421 , where automated anesthesia has been mentioned. The problem is that making people sleep is only part of the job, and the feedback required to create an automated feedback loop is lacking in many aspects of the process. This was created as a potential piece of the puzzle. Unfortunately, it was deemed too technical by my boss to be worthy of support (the concept is high-school physics level, notice...). I am looking for companies/research groups/individuals active in ultrasound research who would be interested in attempting to refine the technique, so if you know someone, don't hesitate to contact me: raoul.schorer@gmail.com

I am missing something: doesn't a typical duplex ultrasonography machine already output blood flow rate?

I recently had such a test and I vaguely recall seeing flow rate in the UI of machine. Oh, maybe the machine measures blood velocity, and you calculate flow rate in units of volume/time?

The scale present on color flow ultrasonography exams does indicate the instantaneous speed at all locations on the plane defined by a cross-cut of the vessel. Duplex ultrasound outputs a time->velocity signal, but only focused on a single point. The traditional technique to estimate flow uses duplex, but is imprecise (~30% flow volume error, at best) and hindered by a lot of manual manipulations making it not reproducible along time and machine users.

Both techniques have disadvantages, making it difficult to obtain a good flow estimate. In the clinics, color flow Doppler is used as a mere location and speed rough indicator. I tried to solve this by including the vessel shape in a mathematical model, which makes the automated method surprisingly practical and efficient on those preliminary tests.

Can something like this be used to detect (and prevent) DVT?

DVT detection by ultrasound (when there is already a DVT installed) is already pretty efficient, because you don't need to know flow volume for that. We test it by the compression test, where a blood clot will prevent squashing of the vein when pressure is exerted on it with the ultrasound probe. Furthermore, the shape of veins is usually far from circular, making our method less than optimal.

Our method is geared towards vessels which have enough pressure in them to constrain them to a circular shape, namely arteries, but also (most likely) great veins. So it could actually prove useful not in DVT, but to detect one of its deadly complications: pulmonary embolus. Apart from this, knowing the flow passing through the pulmonary arteries would prove extremely useful in many cases. It would have to be tested, though. I cannot be sure that it works in this particular case.

Speaking as someone who had DVT/PE there was this stage before experiencing the PE (to the lungs) in which the calf muscle was red/swollen/cramp - it's this period of time in which case I felt like (and still feel like) a device of some type (either low cost ultrasound w/ software) could determine if a blood thinner or something else is needed - and probably save your life. It's evident that between sitting long periods (either at your desk programming or after long flights) that checking the flow is what needs to be done to prevent DVT/PE in the first place (minus other factors of course). I applaud this work because it does sound like a great effort towards something like that.

Did they do a D-Dimer test (https://en.wikipedia.org/wiki/D-dimer)? My understanding is that this can be more reliable than ultrasound (subject to misinterpretation). I'm sure a doctor would tell you to do both if a PE/DVT is suspected.

I hope you have recovered well. My wife nearly died from medical negligence related to a DVT/PE - it can be life changing if it becomes a PE.

Actually, D-dimers alone are used in a low-probability clinical setting, because they are extremely sensitive. When you already have DVT symptoms, the standard practice is to directly go to ultrasound, which is a much more specific exam. And in case you already have PE symptoms, then most MDs would directly go for the spiral CT. http://www.aafp.org/afp/2012/1115/p913.html

Recovery is a weird word. The anxiety you face after you have one can be nearly as crippling as what you went through in the first place (this can probably be said for anything high trauma though). What I can say is that you learn a lot about blood in general after you have had one and things like Vitamin K, green food, etc. It did take 2 ultrasounds on the leg to get an OK to come off thinners, but I still feel there is a window of unknown which healthy and unhealthy people go through in the clotting of blood that is severely under treated and was a great case of where knowledge would have saved a trip to the hospital. I don't in anyway say do-it-yourself-at-home kits with ultrasound and machine learning algorithms should replace medical advice but more and more people are dying from blood clots (whether it be food related, environment, clothes(?)) that it warrants a better look - and I am up for anything non medicinal - especially ones which can cause excessive bleeding like current thinners. If I had a friend who had swollen calf muscles and pain and told them to try something at an urgent care facility to determine a clot was present and it was low cost for all parties involved ... win/win.

US > D dimer to rule in VTE, as many things can cause the latter to be elevated [1]. We almost always go straight to imaging when there's even a remote possibility of DVT.

[1] https://www.ncbi.nlm.nih.gov/m/pubmed/23948628/

Could this technology be used for measuring water flow in pipes?

Absolutely, given the pipe is transparent to the ultrasound signal. Actually, it is a well-known technique (see for example http://www.sensorsmag.com/sensors/acoustic-ultrasound/ultras...), because if the angle of incidence of the ultrasound beam to the pipe flow velocity vector is known (which is not the case with blood vessels), the problem becomes trivial.

Do you have hardware or can you let me know hardware you are currently using?

The soft is coded in Java (source code available on the linked website). As for hardware, I had no choice other than using the S-video output of a clinical ultrasound machine (Sonosite-S) through a frame grabber card (http://www.hauppauge.fr/site/products/data_hvr1975.html). The signal processing was done on a standard laptop from 2010.

The problem in using a clinical machine is that we had no access to the internals, and the video output was an interlaced PAL signal, which translated to a raw 100Hz 180*144 px numerical matrix, so the spatial definition could be much improved by using a better machine. And of course, getting access to the machine internals would allow the use of an optimal software solution, which we couldn't implement in our case.

I would be interested to see how well this algorithm would work w/ mobile setups (ie: https://www.clarius.me/aium-debut-pr/)

This would definitely work, since the device has color flow Doppler, and the performance of the whole thing would mainly depend on the performance of the underlying machine (spatial & temporal definitions). Our solution was not tested in the clinical setting yet, though. Thanks for the link!

How common are Ultrasound machines with digital output on the field ? I guess there are models with DVI/HDMI/DP in catalogs already.

They are still quite rare in most places, actually. Only the brand new models have digital output. I was unlucky, because my hospital had just bought new machines not long before digital output became mainstream on the market. And so for now, we are stuck with analog.

No regrets I guess, you're showing promising results even at that crude resolution and analog contraption.

Can't wait for your setup to be done with less absurd hardware and data.

Do you think the medical hardware industry is ... "milking" their customers ? Do we need an ElonMusk kind of management to improve monitoring, nurse time at lesser cost ?

(sorry if that's too far out of your domain)

Thank you! In my opinion yes, the medical hardware industry absolutely is milking their customer base. In ultrasound systems for example, you buy the machine equipped with the basic software at an already prohibitive price. After that, if you want to go beyond the basics, for example have the UI for cardiac measurements, you buy supplementary modules which are also very expensive although they are by now standard and well-known. I also think that the healthcare system does not lend itself too well to aggressive management models. It does not improve things to try to get people as efficient as possible in the management sense. For example, I think patients are far better off if nurses are let free to do their jobs and tend to the patients instead of imposing too much control on them. Now of course, there are financial imperatives, and technology could provide solutions to maintain/improve care while diminishing the staff numbers/costs. That is my opinion as a practicing clinician. Others may think differently.

Thanks a lot for the details. I want to respect everybody in every domains. But I can't bear the absurd blindness of the market hand when it comes to health and safety. Especially since today's tech makes it obvious that many important tools [1]can be made for very accessible prices. Haa politics. Anyway, I'll keep an eye on your page and the BIG too. I wish I could contribute there.

[1] firefighter came to help my mother suffering from cardiac pain, the EKG monitor couldn't grab a cellphone signal in my whole area. They can't interpret the plot without a doctor so it's basically useless. They told us this things weren't cheap. Yet it's less useful than a 100$ smartphone.

Reply on news.ycombinator.com