CONTACT US

X-ray News

Fireside Chat with the Xperts: Medical Device Inspection

When it comes to healthcare, X-ray isn’t just for broken bones and the like.  X-ray inspection is playing an increasingly valuable role in developing and manufacturing medical devices.  In this Fireside Chat, our Xperts discuss developments in X-ray system technology that is aiding in the production of lifesaving medical devices.

The conversation ranges from research and development, manufacturing, and quality assurance to the strength of artificial intelligence and deep learning.  Listen in, and learn how automated and autonomous X-ray inspection solutions enable fast and accurate manufacturing, support human operators, and collect valuable data in the process.

 

Transcript:

David Kruidhof:
All right. Well, thank you for joining us for another Fireside Chat with the Xperts. I’m David Kruidhof, here at Creative Electron. Today, we’re going to be talking about medical device inspection. I have Dr. Glen Thomas with us and Carlos Valenzuela. So thank you two for joining me this week, as I said, we’re going to talk about medical device inspections. So Carlos, maybe you can give us a little bit of an overview on that. What that entails, what it doesn’t entail.

Carlos Valenzuela:
Yeah. So I think that’s the medical device industry for us has been expanding over the last few years. I think that’s… We started attending more trade shows and things like that, to understand what the market needed from an X-ray machine. And I found out that there’s… it’s so broad, but there’s a need for it. And because there’s so many liabilities, and kind of Glen, we’ll go over that a little bit later, more in the R&D side, what these devices need to be inspected for. Our new capabilities, our speed, our software will allow you to X-ray your medical device, whether it’s an infusion pump, an insulin delivery system or device, whatever device it is, we can help analyze it and figure out if it’s been manufactured correctly. But, like I said earlier, because of the liabilities associated with medical devices, this is a must and is becoming a little more common to use our… Not just our technology, but X-ray in general. I think that’s been our experience so far. I don’t know if you want to add anything, Glen?

Dr. Glen Thomas:
Yeah. Medical devices are highly regulated and that regulation sometimes requires X-ray inspection, especially if it’s written into the SOP of how they’re building the product. So it’s good to get in early. And if you think that X-ray will be a benefit, it’s good to buy in prior to the actually starting up the line and getting everything together. One of the worst things that can happen to a medical device is they find out they need X-ray later, rather than sooner. That’s usually after they’ve had a major issue and they found that only X-ray can help them solve that issue and conform to the quality standards that are required for the product. Right?

Dr. Glen Thomas:
Once you get a cease and desist letter, it’s a little late, right? And then adding X-ray into the production line after the fact is oftentimes very difficult, based on the FDA regulations and the initial report and your quality regulations that the FDA and all parties have agreed to. So it’s always good to look at X-ray early on in the project, whether you need it or not, it’s another story, right? Not everyone needs X-ray in medical devices, but if you… it should be standard procedure for any R&D engineer to look at how, or even the quality engineers and how to preserve the quality of the product as it goes out the door. And sometimes X-rays…well, other times it’s a nice have, but putting the X-ray into the product line after the fact, sometimes it’s very difficult and it’s costly.

Carlos Valenzuela:
Yeah. I think our experience so far, at least when we started getting into the market was that people… these device manufacturers were kind of becoming a reactive, they were getting a lot of returns or they were getting issues and then they would ask, “Hey, can we see this with X-ray?” And it turns out we could. Right? So if they kind of implemented this technology from the development of the device, it would have become a lot easier. So I think as we were exploring more markets and more devices, I think it’s easier for them to get this technology implemented beforehand, not just being a -we have a problem. Can you help us fix it?- Even though we can still do that, but it’s more efficient that becomes part of your testing procedure.

Dr. Glen Thomas:
Absolutely. Nothing worse than a phone call from a frantic customer that says, I have $2.6 million worth of product sitting in a warehouse I can’t ship. What do we do, right?

Carlos Valenzuela:
Yeah. And sometimes is just a missing screw or a bent piece of plastic inside. It’s something very, very simple. And something with our equipment and in any sort of test equipment, it that becomes almost like a…it closes your loop and your production line. So it feeds back information into how other processes are working. So let’s say you have this device and it has screws, springs, plastic, wires. And if our X-ray machine starts seeing a defect on the wire, you can go back to the machine or the person that adds that wire and kind of figure out what the issue is. So you kind of close this system and then you become…you create that feedback into a more efficient production line.

Dr. Glen Thomas:
With our Business Intelligence type software that we have now, you have the ability to track that component really well. You can track that component from an inception essentially, when you barcode it into the X-ray system, right? Or in the barcode as it travels through the process. And then you have a line in the sand per se, right? So if you do get a recall, you would have serial number, 1,000,002 to 2,000,002, right? You have data on those products. So when you look at your X-rays because, every X-ray would be stored in the database based on serial number, date of manufacture, customer PO, whatever you would use as a image tag, you could actually do a search of the database of those images and pull up the… Let’s say you had product 1,000,001 that had an issue that came back from the field, and you suspected that would be a problem. You could pull up 100 parts or 100 samples prior, and 100 samples after that image.

Dr. Glen Thomas:
And you could actually look at the X-ray images and start to see, to do degration of the process, right? If it’s a wire, like your tooling was starting to wear out, you could start to see some variances as you go along to that failure point. And then you would see failures after that point, right? And then maybe in the next shift, they changed the tooling out, and then you would see a stop. So you could actually go back and you could say, from point A to point B, we had a problem. To be safe we’ll go 100 on each side of that serial number and recall that number of devices rather than recall the whole product line for the last 10 years.

Dr. Glen Thomas:
So it actually helps you verify the product, save the images in the database and it actually helps R&D in the long run, because you can look at that image of a product if it comes back from the field and you can understand what may have changed, why did this internal component shift, right? Was it a natural progression or did somebody drop it? And what happens if you drop it? We never tested for what happens if you drop it. Right? So it actually helps fill in the blanks.

Dr. Glen Thomas:
And our database is very detailed in its ability to store a lot of data upfront with serial numbers, parts, dates, times, operators, and all that information, so that you can actually do quite a search on a lot of images. And the more images you put into the X-ray system, the smarter it gets, right? With our AI technology, you have a large database of images. You really have a lot of information there.

Carlos Valenzuela:
Yeah. Nobody’s ever complained about too much data. Right?

Dr. Glen Thomas:
Absolutely.

Carlos Valenzuela:
Just ignore it if you don’t want it. But I think if you’re going to need it, I think that you brought a good point. You can see things not just from, so you can see if the late shift works…what if the late shift only does 80% of production that the early shift does? So all these things are kind of pluses aside from our X-ray technology.

Dr. Glen Thomas:
Absolutely. Yeah. That’s the main reason we developed the business information systems. And that was to expand on the data. Because in the past, what you would get…it’s you get a standalone X-ray system that was barely even able to save an image, and if it’s saved it, it was in some weird format that was proprietary. And then now you’d had a thermal printer. You would print out the image, right? Computers have eliminated all of those issues. We can store images to the cloud. You can store them to the image buckets on the cloud so that the images are available to the whole organization. Right? If you have some R&D type applications, or you saved some process engineers that want to take a look at how the product lines going in, say China or Italy or France, and you’re sitting in your office in the United States, you could literally log onto the system, you could pull the data from the database, or you could watch the system and watch the X-ray operation as it’s going along.

Dr. Glen Thomas:
So for managers, it gives you the ability to remotely monitor in lines as well, right? So you have the ability to log onto the system and watch not only trends in the database, but you can actually see the live images on your screen from your desk. So there’s a lot of advantages for that, especially with the reverse manufacturing facility, right? You’ve got 16 facilities worldwide. You may…and all of them are manufacturing the same product that would give the ability of engineers to get together and look at the product, right? You could X-ray it, everyone’s on a team viewer meeting or a zoom meeting, you can see the data on the screen real time. So there’s a lot of advantages to that data port.

David Kruidhof:
Now, so Glen mentioned the attribute BI with all of its tracking and analysis capabilities, but what about analyzing the image itself? What, on a medical device, is this industry inspecting in terms of… are they measuring things? Is it AI analysis? What kind of stuff are they looking at and making sure it need to be there?

Carlos Valenzuela:
I think it varies from product to product, but we are… since we manufacture all of our systems in Southern California, every product kind of requires a little bit of different kind of recipe, let’s say we have detectors for large field of view. Let’s say you have a medical case and you want to verify that every single component is in place, that’s going to require slightly different hardware than verifying a needle it’s not bent. You know, micro level resolution or micro level magnification. So, aside from that we… through some of our partners like FANUC and Cognex, we’ve been able to acquire more capabilities as far as automation, and machine vision, and machine learning capability. So we can do measurements. We can do placement. Some very simple applications can be done with just plain, not just comparison, but like pattern recognition.

Carlos Valenzuela:
That’s a little bit more precise than just comparing images. If you compare images sometimes they can change a little bit. So you have a lot of false positives. If we do more of a pattern recognition, then the system kind of tends to understand where components are and where they’re supposed to be, even if the image varies a little bit in grade level. So if it got a little darker, it’ll still understand what that pattern is and in anything in between, if you’re looking for something that’s very complex that you can’t put into… sometimes you can’t put it into words, like you have to show an image and explain it, that’s where AI and Deep Learning comes in. AI and Deep Learning can essentially replace operators for making these very hard decisions, not in a sense of physical replacing them, but aiding them, helping them make the right choices.

Carlos Valenzuela:
You might be in a facility that you have your lead operator or your tester, right? And he’s been in the company 35 years and he knows when things are good and he knows when things are bad. He can be the source of training, he can be… he can gauge when a sample is good and when it’s bad. And basically he becomes the AI. So from his knowledge, you can train an algorithm to make decisions on the product. So that’s where this comes in handy. And then the next person comes in and then, the AI will tell them, okay, because my knowledge is that this product is bad, I’m going to display where I think is bad, or some of the defects are, you can decide to agree, you can decide to bypass it. And that’s for the more complex. But aside from that, machine vision can do measurements. Like I said, pattern recognition, placement, location, and anything in between? I don’t know. Do you want to add something Glen?

Dr. Glen Thomas:
That’s major advancement in the X-ray field for medical devices, right? So you put a lot of stress on your operators, when you say, if you mess up interpreting this image, it could cost us $3 million or $4 million, or you could kill someone, right? So there’s a lot of stress on the operator to get it right. And the more stress you have and the harder the operators staring at the screen, the less their efficiency is, right? Everything goes down. So you can’t have an operator on the machine more than an hour, two hours. If you’re making those manual algorithms decisions, right? If you are playing AI, making mistakes, you just get a headache. It’s a very tedious job because of the severity of your screw up. Right? You could literally kill someone. So AI taking that place, actually, it makes a lot of sense because the algorithms are more consistent in their results, right?

Dr. Glen Thomas:
It’s either good or bad. It meets the criteria or it doesn’t. Whereas, a human, ‘yeah that looks like it might be okay.’ And you’ll let it go. Right? AI is not going to pass that. It’s going to say, “No, it’s off because of 0.0004%.” Right? So you get a lot more consistent results with the AI. So the advantage of including AI into the X-ray analysis is huge. And it reduces a lot of that operator fatigue. You could actually have one operator work in three or four different lines rather than one specific product for two hours in a day, right? So it’s key that you switch out your operators when you’re doing the manual algorithm type inspection, right? The AI just adds a huge amount of consistency, reliability, and reduces that operator burden, it would be a terrible job, looking at something like that every day for three or four hours. And AI gives you the ability to speed up as well. You can do more products in a shorter period of time with more reliability and consistency, which is what you’re looking for.

Carlos Valenzuela:
Yeah. And going back to the data that we talked about earlier, if a product failed, you know why. It doesn’t just say, “Oh, this operator number five failed it, ‘why?’ it just thought it was bad.”

Dr. Glen Thomas:
Right.

Carlos Valenzuela:
If you using algorithms, AI and all this machine vision stuff, you have a documented reason why that product failed. And if it became… if it’s something out of the norm, out of the ordinary, you have the image and you can figure out why it made that decision. So that is kind of the main benefit and yeah, things are fast, with a new digital detectors, conveyors and all this technology a system can be less, I mean, a decision could be less than a second. You know, it’s really, really quick. I mean, if a person can do that, I’ll be very impressed.

Dr. Glen Thomas:
Exactly.

David Kruidhof:
What about the people who aren’t quite ready to put full trust in the machines? There are some sort of a middle ground here?

Carlos Valenzuela:
Yeah. I think that’s kind of touched on it a little bit, where we…it always tends to go that way, where we develop a system and even if the system works perfect, they all have this like a manual bypass where, just to build trust on the machine, an operator and the machine have to agree and then slightly… Later, then you become more comfortable with the machine is doing. But yeah, I mean, it could also be part of training operators, right? Your machine is your number one operator. When you train new ones, have him analyze image. And then if they both agree, then that’s kind of part of the operators training.

Carlos Valenzuela:
So yeah, there’s everything in between. A lot of our systems that we do, do have also the automated inspection, but they do also have a manual inspection where what we found is that some procedures don’t allow for automated decisions, they have to be made by a person so they can use an automated inspection. So they have to use a manual inspection or we can use… or you can use automated, but the decision that you actually use it’s a person, but it was aided by the system.

Dr. Glen Thomas:
Right. It’s much simpler. We have the operator alerted that there’s a problem. And look at one image or a series of images rather than staring at that screen nonstop. Right?

Carlos Valenzuela:
Yeah.

David Kruidhof:
There’s a way to present, basically do the automated analysis and present those results for the operator to then make an actual decision, or the decision. Right?

Dr. Glen Thomas:
Right.

Carlos Valenzuela:
Well, I mean, if you look at an image database is a lot easier than do that, than have a machine that’s waiting for you. Right? And maybe the next step is also waiting for you. So there’s that sense that you have to make a decision quick or you’re the bottleneck. But if you look at an image on a computer, on a database, it’s a little bit easier. So there’s not that…you’re not feeling rushed. Maybe you want an image is going to take you 15 seconds and the next five are going to take you five, but you’re not holding up the line. And, we’ve been to some facilities that have two, three people per system, where one’s loading, the other one is in tagging it or something. So like, ‘Hey, hurry up. That was good. Just click, click the green one, click the green one’ right.? Just to, you know, cause you have three people waiting on you, but you have a system that makes those decisions. It just becomes, you know, a very, very efficient production line.

Dr. Glen Thomas:
And if it’s a very complicated component or a sample, you have the algorithms will actually point out where it thinks there is an issue in that. Right? So if you’ve got 10 mechanisms that you need to check in the image or in the sample, the algorithm will pick out the flawed or the suspect area. And that will enable that operator to look at that suspect area and disregard the rest of the noise, because essentially at that point, everything else in the image is noise, except for that area of concern. Right?

Dr. Glen Thomas:
So that makes it quicker as well, for the manual verification, the operator can look at a specific area that the algorithm, the AI says, “Hey, we got a problem in this area, in this segment or this sector,” right? They don’t have to look through the whole image and say, I don’t understand why it failed. Right? So the algorithms, even when you’re doing that manual verification can help lead the operator into the area of interest much faster so that it can make a decision, yes or no. Right? They don’t have to look at a lot of data points, just the data point that matters.

David Kruidhof:
So you spent most of our time talking about manufacturing line inspection, right? What about other uses in the medical device industry, past manufacturing. Now you mentioned that having a stack of product sitting around that needed to be inspected, but-

Dr. Glen Thomas:
Exactly. Since our X-ray systems are real-time, which means they’ve run at least 30 frames per second, you can watch movement in the X-ray, so a lot of R&D applications. For medical, what they’ll do is we’ll put in some ports in the side of the system and I’ve had one company in the past that fed some imitation of veins and arteries into the system. Right? And they were looking with a pump and they would actually pump liquid through the arteries in the veins and use the X-ray system to train doctors in placing some stints like products, right? If they weren’t placed properly with exact precision, it would cause some Eddy currents in the blood flow. If they weren’t placed… They were totally placed wrong, they would block the blood flow completely. Right? So essentially you don’t want to learn that process on the operating table.

Dr. Glen Thomas:
You want to get that process right, so they would use the X-ray system to train the doctors. Because they were essentially doing the, installing this components or a catheter in under fluoroscopy. So it was more or less the same concept as when they were in the operating suite minus the human and the major issues with flaws. So the real time operation of the X-ray system, the 30 frames per second, really helps them in training a lot of medical devices. So it’s R&D training kind of a strange place. But I have sold systems in that application.

Dr. Glen Thomas:
We sold systems for things like bottle fill, right? Medical includes pharmaceuticals, in some cases, do you have the right amount of pills in the package, right? If you’re dealing with a controlled substance and your machines leaving out two then the doctor’s prescribing your 20 doses, and you’re only have you only have 18 doses in the package, there’s an issue. If there’s 22 dose package, now we have another issue, right? Because most pharmaceuticals have to be within 10% of their stated count, whether it’s milliliters or number, right? Number of pills. So you would use it for bottle fill inspection. So it’s not necessarily a medical device, but it is medical related. It’s got the same complications with dealing with external regulations, government regulations.

Dr. Glen Thomas:
So you have returns from the field, right? You get a catheter, you get a device. I had a company that was building some catheters. They had 100 wires inside. And essentially if they got a return from the field, it was required under their FDA submission that they would dissect that and find out what that return problem was because this was a life critical component that they were actually supplying. They would get the product back from the field that had a problem out of the operating suite. And it would take one operator, 40 hours to tear that cable down because, they had to go level by level. It was encapsulated until they found the area that had a problem. After a few hundred, they kind of knew where the problems were going to be in the breaks were going to be, but they still had to manually tear that down.

Dr. Glen Thomas:
And in some cases they would tear it down and find that the brake wasn’t there, it was in a different location. So they go through that same process. We were at a 40 hour return evaluation down to three minutes with X-ray. It was a huge savings across the-

David Kruidhof:
Yeah, quite a change.

Dr. Glen Thomas:
Another one that I had was fun was an ultrasound company building ultrasound probes and repairing ultrasound probes doctors would, or the hospitals would send these probes in. That was with some damage. They were fairly expensive and they would actually need to tear them down. Some of them could be repaired. Some of them weren’t, they would spend $600, $700 per probe to tear it down, to get, to be able to give a quote to the customer, whether you need to replace it, or we can repair it. Right? And in some cases, the cost of tearing it down to evaluate it exceeded the cost of a new product, right?

Dr. Glen Thomas:
So essentially everyone would be thrown away, would they would save money, in a rare instance, or in 50% of the time, they would find that they could repair it and save $100 or $200 or $300. Right. We were able to get that analysis down in their quote process streamlined within an hour. So the product came in, they were able to evaluate the problem, write up a quote and get it back to the customer within an hour versus three or four hours of labor, just to tear it down, to get to the point they would say, “I can’t fix it.” Right? So they saved a ton of labor as well. So there are a lot of different applications that aren’t necessarily medical producing product, but validating and repairing product as well.

Carlos Valenzuela:
You know, one thing we kind of…we haven’t really talked about is working in the medical device industry, X-rays are actually used for other parts of the process. If you work with a medical company, they heard of X-rays because X-rays are very common in the medical field, right? For people, for bones, for CTS and things like that. They’re also used for to sterilize their product. So I think that, just to end on this, that’s one of the barriers that sometimes people don’t know. We talk about X-ray and they’re like, ‘Oh no, no, that’s… we’re not, we don’t, we don’t need it,” but this X-ray in the entity field for medical devices it’s a huge, huge benefit to them. But because X-rays are used for people sterilization in different parts of the process, there’s that the different meaning in the medical field of X-rays.

Dr. Glen Thomas:
Right, right. It’s key. The sterilization process is typically really high power X-ray stuff.

Carlos Valenzuela:
Yeah.

Dr. Glen Thomas:
It’s quite a process. Right? So they go, no, we sterilize our stuff some at another facility. Right. I’ve heard that. No, no, wait a minute, slow down back up. We’re not talking about sterilization, but on that same token, you can use X-ray to look at a product that has been sterilized and it’s still in its sterilized packaging, right? So you don’t need to break your sterilization packaging to actually look at a product in a lot of cases, some of these medical products are in opaque packaging that has been sterilized. So you could still look at the product before it goes out the door, just for a confirmation, if you thought there may be an issue without breaking the sterilization. Right?

Carlos Valenzuela:
Yeah. The amounts of radiation that go through sterilization are way, way, way more than it takes for us to X-ray and get an image. It’d be more kind of scared sometimes like, Oh, you can’t, radiation. And we did some studies and it was a crazy amount of time that they had to be in our machine to get to the levels that… The limits that they had for their product. We were talking about few seconds and, there was months or something like that, but don’t quote me on that one. Right.

Dr. Glen Thomas:
Product dependent.

Carlos Valenzuela:
Product Dependent. We need a little disclosure here. Nothing that Carlos says it’s true.

David Kruidhof:
We’ll add that to the bottom of that.

Carlos Valenzuela:
Take everything with a grain of salt.

David Kruidhof:
All right, guys. I appreciate your time a little bit over again today, but again thanks Glen, thanks, Carlos. And if you all have any more questions, definitely let us know. I’m happy to get to them and we’ll see you all again in two weeks. Thanks.

Related Posts

Fireside Chat with the Xperts: Armor Plate Inspection

As armor plate technology for body armor continues to advance, quality assurance tools for this life-saving equipment needs to improve as well.  At Creative Electron...

Read More
Fireside Chat with the Xperts: Armor Plate Inspection

Fireside Chat with the Xperts: Refurbished vs. Used X-ray Systems

When your budget demands a bit of belt tightening, new equipment may be out of reach.  And when researching the potential purchase of a preowned...

Read More
Fireside Chat with the Xperts: Refurbished vs. Used X-ray Systems

Fireside Chat with the Xperts: “Barbie Dolls to Bombs”

You know a guy has been in the business for a while when you ask what kind of interesting things he's X-rayed over the years...

Read More
Fireside Chat with the Xperts: “Barbie Dolls to Bombs”