Surgical Planning and 3D Printing: What Step Are We Now?

• 1
  Key takeaways
• 2
  What surgical planning and 3D printing actually changes
• 3
  Reading the scan: slice thickness, thresholds and mesh quality
• 4
  Matching the printed part to the job it has to do
• 5
  Where 3D printing stops being the right answer
• 6
  Verification, documentation and what a supplier should hand back
• 7
  Step by step: from DICOM to a part you can hand to the OR
• 8
  3D printing vs CNC machining for patient-specific parts

The shift is simple to state. A CT or MRI scan used to end as a stack of 2D slices on a monitor. Now the same scan can become a physical object on the surgeon's desk a day later. That object lets the team rehearse an approach, check screw trajectories and show the patient what the operation involves.

For the engineer, the interesting part is not the printer. It is the conversion from imaging data to a manufacturable solid. DICOM files carry voxel intensity, not geometry. Thresholding picks bone, soft tissue or contrast-filled vessels, and every threshold you choose changes the resulting surface by fractions of a millimeter.

That is why two hospitals can print from the same scan and get models that differ at the margins. The segmentation settings, the smoothing pass and the mesh repair all leave a fingerprint. If the printed part is used only for discussion, this is fine. If it is used to decide where to drill, it is not.

So the honest answer to "what step are we now?" is: additive is routine for visualization and increasingly common for guides, while load-bearing implants and precision drill tooling are still machined. The two processes sit next to each other in the same workflow.

Start with the source data. A slice thickness of 0.5–1.0 mm on a modern CT gives enough axial resolution for most craniomaxillofacial and orthopedic work. Thicker slices force the software to interpolate, and interpolation rounds off the very edges a guide has to sit on. Spiral scans with 0.625 mm slices are a practical baseline.

Next, threshold. Bone sits roughly between 200 and 3,000 Hounsfield units depending on density and the scanner's calibration. A single global threshold will merge the mandible with the teeth in one patient and drop thin cortical walls in another. Region-growing or per-region thresholds take longer but produce cleaner anatomy.

After segmentation comes mesh work. Export an STL and check it before it ever reaches a slicer. Look for non-manifold edges, flipped normals and shells thinner than 0.8 mm. Slicers will silently repair some of this, which hides a geometry error behind a printable-looking file.

Keep a record of every parameter. If a guide does not seat in the OR, the first question is whether the mesh changed between the planning model and the printed guide. Without a parameter log you cannot answer that.

• 1
  Slice thickness0.5–1.0 mm axial; avoid anything above 1.5 mm for guide work.
• 2
  Bone thresholdRoughly 200–3,000 HU, split per region rather than one global value.
• 3
  Minimum wallKeep anatomy and guide walls at 0.8 mm or thicker for handling.
• 4
  Export formatWatertight STL or 3MF; check the mesh, do not trust auto-repair.

Not every part in a surgical plan needs the same accuracy. Sort them by consequence. A model that sits on a table for a case conference can be printed at 0.2 mm layer height on a filament machine and nobody will care about 0.3 mm of surface error.

A cutting guide is different. It has to seat on bone with no rock, and the saw slot or drill bushing position carries directly into the patient. Here layer height, resin shrinkage and post-cure distortion matter. Print the guide in the orientation it will be used, and verify the seating surface on the actual model before the case.

Trial implants and instrumentation sit at the top of the list. If a trial has to check whether a press-fit stem enters a reamed canal, ±0.3 mm is not enough. We machine these from 6061-T6, 316L or Ti-6Al-4V on 5-axis centers and hold ±0.005 mm with Ra 0.8–1.6 μm on mating surfaces.

A practical split: print for anatomy, print for guides where the geometry is bulky, and machine anything that has to fit, slide, thread or bear load.

The cost argument is weaker than it used to be. A desktop resin printer is cheap, and a hospital can run it in a basement room. The real limits are geometric and mechanical, not financial.

Threads are a good example. A printed M3 thread in a photopolymer strips after a few assembly cycles. If a guide has to be clamped, screwed or adjusted in the OR, machine the threaded component or press in a metal insert.

Thin features are another. A drill bushing wall below 1.5 mm in resin will ovalize under load and the drill will wander. The same bushing in 316L or hardened steel holds its bore. This is why hybrid guides exist: a printed body that follows the patient's bone, plus machined metal inserts at the working points.

Surfaces matter too. If a part slides against another part, printed layer lines create friction and wear. Machined Ra 0.8–1.6 μm surfaces behave predictably. In our shop, hybrid guide work usually means printed anatomy plus machined metal inserts, both checked against the same mesh.

Patient-specific parts sit in a regulated chain, so paperwork is part of the deliverable. At minimum, keep the source scan hash, the segmentation parameters, the final mesh file, the print or machining parameters, and the inspection record.

Inspection should match the risk. A planning model needs a dimensional spot check on overall length and one or two landmark positions. A cutting guide needs the seating surface checked against the model and the bushing bore checked with a pin gauge. A trial implant needs full dimensional inspection on the features that control fit.

We run raw material checks, in-process monitoring and a final inspection before shipment, with reports on request. Our facilities hold ISO 9001:2015, IATF 16949:2016, ISO 13485:2016 and ISO 27001:2022. The 13485 scope covers medical device work; ISO 27001 covers how patient scan data is handled.

Uploads are treated as confidential and an NDA is available on request. For hospitals, that last point often decides the supplier more than the tolerance does.