Stent-testing smart robot makes the medical grade

The Raspberry Pi often makes the world a better place. This time, it’s helping to test 3D-printed stents using a smart stent-testing robot.

Stents are small tubes used to prop open a patient’s airway. They keep people alive, so it’s incredibly important they don’t fail.

In fact, the FDA (Food and Drug Administration) requires testing of each design by compressing it over 300,000 times. That’s a sturdy challenge for any human, which is why machines are normally used to mash up the stents.

The usual stent-destroying machines are dumb clamps, with no idea whether the stent is breaking or not.

Stent Testing Robot Camera

A smarter stent-testing robot

Enter the Stent-Testing Robot, an intelligent arm that mashes stents while a Raspberry Pi Camera Module keeps a sharp eye on how it performs.

It’s designed by Henry J. Feldman, Chief Information Architect at Harvard Medical Faculty Physicians.

“We start with a CT scan of the lungs, and via a 3D reconstruction get the size and shape of the bronchus that we wish to stent open,” explains Henry. “The trick is to make it the exact shape of the airway.”

The challenge with testing is if stents start to fail before the end of the test. The dumb devices currently used continue to pulverise the stent when this happens.

Stent Testing Robot Camera Squisher

Machine vision to control stent-testing

The Raspberry Pi, meanwhile, uses machine vision to stop the mashing at the moment of failure.

The instant-stop approach enables Henry’s team to check which part failed, and view a time-lapse leading up to the failure. The video helps them design more reliable stents in the future.

Henry explains:

Naturally, we turned to the Raspberry Pi, since, along with a servo control HAT, it gave us easy OpenCV integration along with the ability to control a Hitec HS-5665MH servo. We also added an Adafruit 16-channel Servo/PWM HAT. The servo controls a ServoCity Parallel Gripper A.

Python was used to write the servo controller application. The program fires off a separate OpenCV thread to process each image.

Henry and his medical team trained the machine learning system to spot failing stents, and outlined the likely points of failure with a black marker.

Each time the gripper released, the robot took a picture with the Pi Camera Module and performed recognition of the coloured circles via OpenCV. If the black marker had a split or was no longer visible, the robot halted its test.

The test was successful:

While the OpenCV could occasionally get fooled, it was remarkably accurate, and given this was done on an academic budget, the Raspberry Pi gave us high-performance multi-core capabilities for very little money.