CNC Machining for Medical Imaging: How Tolerances Affect Image Quality

• 1
  What CNC machining for medical imaging actually controls
• 2
  Material choices for MRI, CT, PET and ultrasound hardware
• 3
  Where five-axis machining earns its cost
• 4
  Where machining stops being the right answer
• 5
  Matching process to imaging component

An imaging system is a mechanical assembly that holds a geometric relationship while it spins, slides or vibrates. A CT gantry rotates a source and detector array around the patient at speed. An MRI gradient coil set fires thousands of times per second. Ultrasonic probes sweep a crystal array across a few centimeters.

Each of those motions turns small machining errors into visible artifacts. A bore that is 0.02 mm out of round on a rotating anode assembly shows up as ring artifact. A detector rail that is not flat by 0.01 mm shifts the reconstructed slice position. The scanner software cannot correct what the metal does wrong.

That is why the useful question is not whether a shop can hit a number on a drawing. It is whether the shop can hold that number across a production run, after heat treat, after plating and after cleaning. Those steps move the part.

CNC machining for medical imaging parts usually means milling, turning and mill-turn work on metals that are non-magnetic, dimensionally stable and compatible with the sterilization cycle the device will see. The process is subtractive, so the geometry comes from a solid block or bar and the surface is generated by a controlled cutting edge.

• 1
  Geometric error becomes image errorRoundness, flatness and position drive reconstruction accuracy.
• 2
  Stability matters as much as accuracyA part that moves after plating is no longer accurate.
• 3
  Cleaning is part of the processBurrs and trapped chips are functional defects in imaging hardware.

The scanner type narrows the material list before machining strategy enters the conversation. MRI puts parts in a strong static field and a pulsing gradient field, so ferromagnetic behavior is disqualifying. 316L stainless, titanium and some aluminium grades are common where structure is needed. Beryllium copper appears in RF coil contacts and spring fingers, and it machines well but needs its own chip handling.

CT and PET hardware spends more time on stiffness, thermal stability and X-ray transparency. Aluminium 6061-T6 and 7075 are typical for housings, brackets and carriage plates because they are light and hold tolerance. Tungsten and lead-loaded polymers handle shielding, though those are usually cast or molded rather than machined.

Ultrasound probes sit at the other end of the size scale. Acoustic stack components, lens housings and backing blocks are small and often have wall sections under 1 mm. PEEK and POM are used where acoustic impedance and wear matter. 303 and 316 stainless show up in needle guides and biopsy hardware.

The material also sets the finishing path. Anodizing adds a few microns and is not uniform on threads. Electroless nickel is more even but still changes a critical bore. If a drawing calls ±0.005 mm on a surface that will be plated, the machinist has to cut to a pre-plate size, and that number belongs on the drawing.

• 1
  MRINon-magnetic only: 316L, titanium, aluminium, beryllium copper.
• 2
  CT / PETStiff and light: 6061-T6, 7075, shielding alloys.
• 3
  UltrasoundSmall features, thin walls: PEEK, POM, 303 stainless.
• 4
  Plan for platingPre-plate dimensions must be on the drawing.

A detector housing with compound angled faces, deep pockets and a sealing groove on four sides cannot be made in three setups without stacking tolerance. Each refixture adds error. Five-axis work holds those features in one setup, so the positional relationship between the groove, the mounting face and the bore stays under a single datum.

The gain is not just accuracy. Fewer setups means fewer chances to mark a cosmetic surface and less time in the machine. On a part that needs Ra 0.8–1.6 μm on sealing faces and Ra 1.6–3.2 μm everywhere else, the tool path and the setup count are the same decision.

Five-axis is not automatically better. A simple flange with a bolt circle and a through bore belongs on a lathe or a three-axis mill. Putting it on a five-axis center adds hourly cost with no accuracy benefit. The judgment call is whether the part has features that must be related to each other across multiple directions.

Mill-turn centers cover the middle ground. A gradient coil former or a rotating anode hub often starts as bar stock and needs turning plus cross-drilling plus a milled flat. One machine, one setup, one datum.

• 1
  Use five-axis whenFeatures on different faces must share a datum.
• 2
  Use three-axis whenThe part is prismatic and reachable from one direction.
• 3
  Use mill-turn whenTurned body plus cross features, single setup.

Machining is a one-part-at-a-time process with a fixed setup cost. That is fine for prototypes, for low-volume scanner hardware and for parts where the geometry is the function. It stops making sense when the part is a simple shell in a volume of tens of thousands.

Die casting and vacuum casting take over when wall thickness is uniform and the surface does not need to be generated by a cutter. A CT housing cover with no critical bore is a casting. A detector mounting plate with a ±0.005 mm bore pattern is not.

There are also geometries a cutter simply cannot reach. Internal channels that curve in three dimensions, lattice structures and hollow shells with no split line are additive territory. A common route is hybrid: print the complex core, then machine the sealing faces and bore to tolerance.

The practical rule is to machine the surfaces that set alignment, sealing, heat transfer or wear, and to use a different process for the surfaces that only need to be there. Mixing processes on one assembly is normal in imaging hardware.

• 1
  Machine it whenBores, seal faces, datum surfaces, low to mid volume.
• 2
  Cast it whenUniform walls, no critical bore, high volume.
• 3
  Print it whenInternal channels or lattice, no split line possible.