Dual-Filament 3D Printing of Patient-Specific CT Phantoms with Embedded Implants and Tunable Metal-Artifact Intensity

Imagine you are a detective trying to solve a mystery, but every time you look through your magnifying glass at the most important clue, the glass itself smears the image with a giant, confusing smear of black ink. In the world of medical imaging, this "magnifying glass" is a CT scanner, and the "ink" is a metallic implant like a screw or a hip replacement. When X-rays hit metal, they get confused, creating streaks and dark shadows that hide the delicate tissues around them. This makes it incredibly hard for doctors to see if a surgery went well or if a patient is healing. To fix this, engineers have invented clever computer programs called "Metal Artifact Reduction" (MAR) algorithms. These programs try to digitally wipe away the ink smears. But here's the problem: how do you test if these programs actually work? You can't just ask a patient to swap their titanium screw for a plastic one to see the difference, and you can't use a simple plastic block because it doesn't look like a real human body. Scientists need a perfect "test dummy" that looks exactly like a patient, has a metal screw inside, and can magically change how much "ink" it creates so they can test their cleaning software fairly.

This is where a team of researchers from the University of Pennsylvania stepped in with a clever new trick. They developed a special way to 3D print these perfect test dummies using a method they call "PixelPrint." Think of it like a high-tech baker who doesn't just bake a cake; they can bake a cake where every single crumb is a different flavor, and they can even hide a chocolate chip inside a strawberry crumb without the flavors mixing. Usually, 3D printers for medical scans are like single-flavor bakers: they can print a bone or a muscle, but they can't print a bone and a metal screw in the same object while keeping the texture of the real body. This team, however, upgraded their printer to use two different "filaments" (the plastic strings the printer uses) at the same time. One filament is a special plastic mixed with calcium to look like soft tissue and bone, and the other is a plastic mixed with metal dust to look like a screw.

The researchers' main discovery is that they can control exactly how "metal-y" the screw looks by changing how much of the metal-plastic they pack into the screw area. They printed three identical copies of a patient's neck (specifically the cervical spine from vertebrae C4 to C6) with six spinal screws embedded in them. The first copy had no metal plastic at all (0% infill), acting as a perfect, clear "ground truth" reference. The second had a medium amount of metal plastic (50% infill), and the third was packed with almost all metal plastic (85% infill). When they scanned these with a CT machine, the results were exactly what they hoped for: the 0% version looked like a clean neck, the 50% version had mild streaks, and the 85% version had heavy, dramatic streaks, just like real metal implants do. Crucially, the rest of the neck—the muscles, the airway, and the bone texture—looked exactly the same in all three, proving that the "ink" was only coming from the adjustable metal parts.

The team also tested how different types of X-ray energy affected these streaks. They scanned the phantoms at various energy levels, ranging from 50 keV to 190 keV. They found that at lower energies (like 50 keV), the streaks were huge and the screws looked like they were blooming into giant, blurry blobs. But as they turned up the energy to 130 or 190 keV, the streaks faded away, and the screws became sharp and clear again. This confirmed that their new printing method could reliably mimic the real-world problem of metal artifacts and how they respond to different scanning settings.

However, the paper is careful to note a few limits. While the plastic mixed with calcium did a great job of mimicking soft tissue and spongy bone, it couldn't quite reach the super-high density of the hardest, outer layer of real bone (cortical bone). In the real patient scan, that hard bone was over 1000 Hounsfield Units (HU), but the printed version topped out around 620 HU. So, while the print was excellent for most things, it wasn't quite "hard enough" to perfectly copy the densest parts of a real skeleton. Additionally, the metal they used in the plastic (stainless steel) isn't exactly the same as the titanium or cobalt-chromium used in real surgeries, though it creates very similar-looking streaks.

The researchers measured the "messiness" of the streaks using a specific mathematical tool called the Gumbel p-index. They found that as they increased the metal content from 0% to 50% to 85%, the "messiness" score went up in a perfectly predictable, straight-line fashion. The score jumped from 46.7 HU for the clean version to 57.1 HU for the medium version, and finally to 90.5 HU for the heavy version. This proves that their method allows scientists to dial the artifact intensity up or down like a volume knob, without changing the anatomy underneath.

In short, this paper suggests that this new dual-filament 3D printing method is a powerful, reproducible way to build test dummies for checking how well computer programs can clean up metal artifacts in CT scans. It's not a magic cure that fixes the problem in patients yet, but it provides a reliable, adjustable "training ground" where engineers can test their software against a known, perfect reference. The authors believe this approach could eventually be used to test scanners for knees, hips, and dental implants, helping to make medical imaging clearer and safer for everyone.