Would You Let a Robot Stick You With a Needle?

An anonymous reader writes "IEEE Spectrum has a story about a robot that uses infra red and ultrasound to image veins, picks the one with best bloodflow, and then sticks a needle in. (video included). Veebot started as an undergrad project and the creators are aiming for better performance than a human phlebotimist before going for clinical trials. Robodracula anyone?"

Comments

Why yes, I would.

[Anonymous Coward]

I'll always take a robot over a human when my safety is in question. I want a human involved, but predictable error that can be controlled is far preferable to unknowable error modes of humans.

Re:Why yes, I would.

[phantomfive]

predictable error that can be controlled is far preferable to unknowable error modes of humans.

This is exactly backwards....a human will be aware enough to never jab the needle all the way through your arm. If there's a bug, the computer will do that happily and quickly.

Therac-25 is an example of the dangers of improperly tested computers with lethal equipment.

Re:Why yes, I would.

[ShanghaiBill]

Therac-25 is an example of the dangers of improperly tested computers with lethal equipment.

The Therac-25 was the result of layer after layer of utter incompetence. They assigned a programmer who wasn't qualified to write a javascript button-click handler, to write life-critical sofware. Then no one else even looked at his code. There was no design review, no QA or bug tracking, and very little testing. Even after the defect was reported, there was no review or followup, or realization that it could even be a software problem. But the problem went much deeper. The hardware design was just as defective. There were no interlocks, in either hardware or firmware, to prevent defective software from killing patients. Many books on mission critical embedded system design devote an entire chapter to all the stupid mistakes that made up the Therac-25. If you make a list of the rules of sane system design, the Therac-25 design will have violated nearly every one of them.

Re:Why yes, I would.

[ozmanjusri]

a human will be aware enough to never jab the needle all the way through your arm. If there's a bug, the computer will do that happily and quickly.

And a bit of thought to the mechanical design of the robot will prevent it ever having the physical capability to do that.

Which oddly enough, is how they've designed the robot in TFA....

Re:Why yes, I would.

[DarkTempes]

I'm fairly certain that you can make it mechanically impossible for this to happen in a needle jabbing machine.

And a human could have a seizure or something and jab the needle all the way through your arm.

Re:Why yes, I would.

[Dunbal]

You'd have to pick a very special case - like a mentally impaired, combative patient for example. But then that's why staff with brains exist. You have to know who can get the machine, and who can't.

Re:Why yes, I would.

[BitZtream]

And the exact same thing could happen to any other completely mechanical device. Unless you wrote the software or were closely involved with the development process, you have no fucking clue as an end user of medical devices if its actually safe to use.

Yes, its an example of how to do it wrong, but you CAN NOT ignore the fact that IT HAPPENED.

The example is mentioned not to show how it can be done wrong, but that even in the highly regulated medical industry, where lives are ALWAYS on the line, it slipped right through with a completely improper design and inadequate testing, where as even a 15 year old would be more reliable at noticing the missing filter when reconfiguring the machine more often than the T-25 failed.

Re:Why yes, I would.

[Dcnjoe60]

You sir, are an idiot. Plain and simple.

Humans make quite a lot of errors when extracting blood, they put puncture the backside of the vein, they miss the vein - and this is on healthy adults with fairly visible veins.
When it comes to getting a vein on a child or elderly, it can take them many tries, if a robot can find the vein faster and with less complications involved, then it could ultimately save lives.

Personal experience: when I was a child they had 5 goes at my right hand and another 6 at my left, before hitting a vein. As an adult donating blood, I've had them miss my vein because they where too busy thinking about something else, I've had them puncture the backside of the vein, resulting in quite a huge "bruise" (the blood will flow out and collect in the elbow, which looks absolutely grim, but is not dangerous).

A robot can do the same task over and over, never make mistakes due to personal problems etc. Sure, a malfunction *could* happen, but we have used robots along side humans for ages and we use them for surgeries now, with very few problems.

Most of the problems you describe aren't the problem of the person sticking you but a problem with the veins in the person. Veins roll, they collapse, people involuntarily twitch and all sorts of other factors come into play. A robot sticking a patient will not be immune to that, nor would this be something as simple as a robot doing a repetetive task over and over as each and every stick will be different and unique, even on the same patient from one day to the next as you can't use the same site twice in a row.

Robots and computers are great for doing the exact same thing over and over again. However, differences in peoples anatomy, although similar, adds a layer of complexity that either will lower the efficiency of the automated process to account for it, or will still result in misses.

Re:Why yes, I would.

[Anonymous Coward]

How many doctors have killed people? You can't ignore those either.

Re:Why yes, I would.

[wisnoskij]

Honestly, I think the main problem was far simpler than high school level coding practices.

When this happens: "The feeling was described by patient Ray Cox as "an intense electric shock", causing him to scream and run out of the treatment room."
And then the patient dies of radiation poisoning.
And you continue using the machine for years afterwards without understanding what happened, and patients continue to die over a 2 year period.

That is your main problem, right there.

Re:Why yes, I would.

[ShanghaiBill]

And the exact same thing could happen to any other completely mechanical device.

You design the mechanism so that it is physically impossible for the software to do something dangerous. In the case of Therac-25, there should have been a mechanical interlock that cut power to the radiator when the shield was not in place. In the case of the needle sticking robot, you use an actuator with a stroke of, say, 5mm. Then there there is no way it could "jab through your arm". You also use a weak actuator, that doesn't have enough physical strength to push into bone, even when given full power. You put a spring-loaded (not software controlled) sheath over the needle, so the needle is never exposed unless it is pressed against skin. You design the hardware assuming the software is malicious. You design away any way you can think of for the software to do harm.

Then you design the software assuming the all the mechanical interlocks have failed, and use sensors to double check everything.

Re:Why yes, I would.

[AmiMoJo]

This is how much safety critical software is written. I used to do code for fire alarm and smoke extract systems. The company very clearly defined the circumstances under which it would work: one fire in one location. Two fires and you were screwed. No-body reviewed my code, and the only testing was when the guy installing it got on site and tried it. Finally a fire martial would ask for a demonstration of the agreed functionality, not bothering to check any more than the simplest of fault conditions, and then it was signed off.

Companies see safety as something they are required to have by law and which gets in the way of their profits. Every building needs a fire alarm system but why waste money on a good one when the lowest bidder can supply something that meets the legal requirement. Fires are pretty rare and there is always someone else to blame - it's never the fire alarm's fault that people died.

Re:Why yes, I would.

[icebike]

You design the mechanism so that it is physically impossible for the software to do something dangerous. In the case of Therac-25, there should have been a mechanical interlock that cut power to the radiator when the shield was not in place. In the case of the needle sticking robot, you use an actuator with a stroke of, say, 5mm. Then there there is no way it could "jab through your arm". You also use a weak actuator, that doesn't have enough physical strength to push into bone, even when given full power. You put a spring-loaded (not software controlled) sheath over the needle, so the needle is never exposed unless it is pressed against skin. You design the hardware assuming the software is malicious. You design away any way you can think of for the software to do harm.

Then you design the software assuming the all the mechanical interlocks have failed, and use sensors to double check everything.

Exactly.
But you also make the output of the sensor system visible to the human attendant (who will be there anyway), and provide them a safety override switch, for when the needle breaks off in the patients arm due to a flinch or for what ever reason.

Come to think of it, maybe just providing the human with the sensor system may yield better results, and probably provide faster service in a wider variety of situations.

Unless you envision this as a self service device, you are going to have a nurse there anyway. Drawing blood is not all that dissimilar from any other use of needles for intravenous injections, transfusions or drips, other than that its in and out relatively quickly. So maybe a wearable sensor system would be more appropriate.

I'll give it a try

[rgmoore]

I'd be willing to give it a try. I've been stuck by enough nurses in my life that I'd be willing to give a robot a try. I wonder if it would identify the same spot that the human phlebotomists always use; I've given blood enough times that I have a nice scar to show where the needle ought to go.

sadistic blood workers

[bzipitidoo]

They used to jab the tip of your finger. That's just about the most sensitive, painful place they could choose to get a blood sample. Fingertips have the greatest concentration of nerves. Being medical professionals, they of all people should know that. So why couldn't they prick some other spot, like the forearm? It really seemed like they were at best indifferent to causing their patients unnecessary pain. At worst, I wondered if some of them were sadists.

Some years ago, a change in this procedure came along. Now, they prick the side of the finger, not the tip. Much, much less painful.

Would You Let a Human Perform Surgery?

[Snufu]

Machines and computers are ideally suited to many of the tasks that make up medical care. Autonomous systems are already omnipresent throughout medicine and its only a matter of time before we trust them ahead of error-prone humans, especially for tasks that require fast reaction, repetition, or precision.

Future generations may gasp at the thought that at one point in our history we went under the knife to mortal hands.

There are good reasons...

[sirwired]

The skin on the tips of your fingers is both thick and generally well-vascularized, (but not so much that there is any chance of hitting a larger vessel).
They don't have to pinch your skin to force sufficient blood to the surface to collect. (This causes bruising in people with fragile skin.)
There is a very high concentration of nerve endings, the pain receptors are not nearly as dense.
There's no muscle, which is sore for some time when injured.
It's consistent from person to person; a forearm stick will vary widely depending on the thickness of the skin, fat, and muscle layers. That's not a worry on the fingertip, where everybody will have enough skin that that's the layer they'll always be drawing from.