DARPA Triage Challenge | Hacker News Reader

DARPA Triage Challenge

triagechallenge.darpa.mil

73 pointsgeox28 comments

FYI it should perhaps be mentioned that the deadline for participation has passed [0]:

   Team Qualification open for self-funded teams
   Sept 1 – Nov 27, 2023

   Challenge Kick-Off
   November 6-7, 2023

It seems that the challenge ends "Fall 2026"

[0] https://triagechallenge.darpa.mil/docs/DARPA_Triage_Challeng...

I’m on the data engineering team prepping the training datasets for delivery to the competitors’ workspaces. It’s been a really fun challenge sorting through all the different modalities of data and wildly different formats coming from the various ER teams to get it all in a useable state.

>Vision >A primary stage of MCI triage supported by sensors on stand-off platforms, such as uncrewed aircraft vehicles (UAVs) or robots, and algorithms that analyze sensor data in real-time to identify casualties for urgent hands-on evaluation by medical personnel. >A secondary stage, after the most urgent casualties have been treated, supported by non-invasive contact sensors placed on casualties and algorithms that analyze sensor data in real-time to predict need for life-saving.

The non-invasive contact sensor sounds interesting. Does anything like that exist today? I'd imagine some of the features would/could be: heart rate monitoring (including ekg and other diagnostics), blood pressure, body temp, oxygen level. Maybe this thing is band wrapped around their arm). Also maybe a blood sample via a minimal/automated prick from the band (doesn't sound as non-invasive anymore but could be helpful).

Sounds like they want a more portable Lifepak and some algorithmic alerts. Probably the major diagnostic you missed is ECG.

No blood samples, all the vitals are from strap-on sensors. They do want to incorporate video and audio from the helicopter/ambulance for the pre-hospital portion and from the ER for in-hospital data in future phases of the competition.

There's been several attempts at noninvasive biosensing using techniques like Raman spectroscopy, but it's a lot harder to make it reliable than people estimate. Apple recently shelved its program for doing glucose sensing from the watch, which I remember Tim Cook had remarked would be straightforward. On top of that, the places where DARPA wants these sensors to be deployed (MCIs, trauma victims) make for very difficult sensing environments (due to blood loss, shock, hyperglycemia etc)

How do people hear about things like this? I would have loved to be involved but it’s too late now.

FedBizOpps. Here is a link to a search for more DARPA opportunities:

https://sam.gov/search/?index=opp&page=1&pageSize=25&sort=-m...

I live in this world and had no idea how to construct that link. This is the sort of stuff that keeps me coming back to HN.

Similar link for Defense Innovation Unit: https://www.diu.mil/work-with-us

That link worked yesterday but not today.

For DARPA (and IARPA) projects it's usually large teams that get selected, generally a university or a large federal contractor--or sometimes consortia from several such organizations. Unless you know how to make a tricorder, your chances of getting in are pretty slim unless you're a member of an organization like that. And that's fairer than it sounds, since usually DARPA-hard means lots of moving parts and therefore lots of experts on different topic areas on any given team, else the problem is hopeless.

Defense contracting can be intimidating to get into, for sure. Like most professions, it's a universe unto itself, and there are meetups in larger cities around the topics (aerospace, for example) to make friends in the industry.

As for the work itself, if you're interested in going into defense contracting, some online courses may be freely available to you at Defense Acquisition University, which educates on a lot of the policy and bureaucracy (hope you like acronyms).

In what situation do you want a triage AI drone rather than like, I dunno, an extra box of blankets and antibiotics?

The whole concept seems silly.

The idea is to identify the casualties that could benefit the most, and fast.

Imagine if there is a bomb that goes off in a stadium or music venue or whatever (e.g. Manchester bomb in UK that was during a music concert where hundreds of people were injured simultaneously) or a major train crash etc where there are loads of major injuries simultaneously. There are say 10 first responders on the scene, but 200 casualties - who gets treated first? Blankets won't help.

Chuck a drone up and it might identify the "priroity list" of who gets first aid first. E.g. someone could be in cardiac arrest who would benefit from immediate CPR, but you might not even get to them if you are manually triaging hundreds of casualties across a wide area, and they'd die "unnecessarily" just because the people on the scene didn't know/were overwhelmed/stressed out/panicking etc. I've only ever been in much much much less severe circumstances and can only say that it is hard to think straight and clearly even in such a "minor" situation - I shudder to think about what it would be like if you are amongst the first on the scene to something big.

Anything that can help save the most lives is laudable.

Minor nitpick but interesting for laypeople:

During a multi-casualty incident (MCI) the commonly used START triage system would invariably categorize a patient in cardiac arrest as 'Black' or 'Expectant'--these patients are typically passed over so that medics can treat more viable patients.

Additionally, pulses aren't used for determining whether a patient is 'Black'. Instead, medics will check to see if the patient has spontaneous breathing: if respirations are absent, the airway is positioned (typically jaw thrust or head-tilt chin-lift). If no spontaneous breathing is present despite positioning, the patient is assessed as 'Black' and the triaging medic moves on.

It is grim, but at a true trauma MCI (# patients >>> # rescuers) there should be no healthcare practitioners or rescue personnel performing CPR. This seems harsh but given the low odds of survival to discharge of traumatic cardiac arrests, it makes perfect sense in an MCI scenario.[0]

[0] Work as a Critical Care Paramedic, to lazy for real references :P

Definitely interesting to me!

I suppose everything is situation dependent, but what is the most extreme injury which would not receive a Black? If someone lost a limb and likely to bleed out - would that be deemed worth treatment?

A single lost limb with major bleed would probably still be breathing, and so would not be Black. Their vitals would place them in Red. They are the people triage is designed to save, by minimizing time spent with Black, Yellow, and Green.

Would depend on the following factors: respiratory rate, perfusion to skin, mental status.

Typically, an amputation with significant bleed would hit the respiratory rate (>=30/min) and perfusion (cap refill >=2 secs) thresholds for 'Red'.

That said, I've been present at a plane crash with four patients where one had a clean amputation that had had a tourniquet applied immediately following the crash. He was speaking clearly, had adequate perfusion, and was breathing appropriately. Technically, this patient could be categorized as a 'Yellow' and was, in this case, considered a lower priority than the head injured patient trapped within the aircraft.

It’s the military, it sounds silly because it is to normal people.

But when you have thousands of troops in combat and hundreds of dead and wounded, how would you triage under fire?

This is a current problem in active wars.

There is also a triage nurse in every ER, whose duty it is to decide who gets treated first. However, of the three triage categories, one hopes in an ER that few if any patients will fall into the category of "dies even with help". But I suppose gunshot victims could fall into that category.

This is an issue before the ER or even stabilization point.

You have a battlefield spread out over miles, with dead and wounded all over, and limited resources to pickup the wounded.

This is about triage prior to sending people out of a bunker into an active combat zone to rescue wounded, not waste time on bodies or people that are going to die before they get to the ER.

That's right. It's been a problem in war for decades.

I'm happy to see a return to the defensive survival evacuation and triage role for AI, CV, and robotics as opposed to the regression from that role into autonomous killing machines.

Building an Alan Alda MASH persona along with supporting performers and portrayers is not a bad idea, even if the dragging on the sled requires much less intelligence save for the head-banging of the injured that's tough to avoid and manage.

I think you misunderstand triage.

Triage happens early on but it isn't about antibiotics or blankets or any other specific help. A triage drone shouldn't be treating anyone.

Triage means identifying who needs what help and allocating people to do it. Often this is about identifying who is at most urgent need of help.

So the job of a triage drone could be to identify the people who need help the quickest. Perhaps identifying someone who is unconscious but has a pulse in a larger group of dead people.

So you give the people who need resources to live instead of wasting them on those who will live without them.

...or die even with them. That's the less pleasant half of triage: Don't waste resources on people who are going to die regardless of treatment.

Yeap. And that's of course why it's called triage, instead of diage (or would it be biage?). People are divided into three groups: those who will get well without help, those who will die even with help, and those who will die unless they get help.

I'm not sure where you heard this, but triage does not use the prefix tri- meaning three. Triage is a concept that is instantiated differently per locale and is not a universal 3 bucket system.

Reply on news.ycombinator.com