Realize Medical

XR Demystified for Surgeons

When 3D helps, and when it doesn’t

Last reviewed October 2026

The short version: a CT or MRI scan is a 3D volume, but most clinicians still read it one flat slice at a time and rebuild the anatomy in their heads. For a straightforward case, that works fine. For complex anatomy, several studies report that a 3D view helps clinicians understand spatial relationships, agree with each other more, and make plans closer to the operation actually performed. It only helps if the model is built from good images, by someone exercising clinical judgement, and checked against the slices.

Slice-by-slice reading is a habit from when scanners and screens couldn’t do better. Early CT scanners built the body one cross-section at a time, and a flat monitor was the only way to view the result. Neither constraint holds anymore.

What the video shows: real CT slices through a kidney with a cyst lift into a stack and become a 3D model. Then the four jobs that slices leave to the reader, set out below. Then a real artery running at 45° to the scan: on the slice it measures 15.3 × 10.6 mm, at right angles 11.3 × 10.5 mm, so the slice reads 35% wider. It ends with when 3D adds value and what a trustworthy model needs. 69 seconds, text on screen, music only. CT data: TotalSegmentator dataset (CC BY 4.0, see References).
Six axial CT slices from one de-identified kidney CT, fanned out as a stack with each structure outlined in colour, beside the 3D model built from the same scan in the same colours: kidney (tan), tumour (amber), smaller lesions (pale yellow), vessels (red) and ureter (green), with the six slice planes drawn through it from the same viewpoint.
From a de-identified Elucis case. Imaging data: TCGA-KIRC collection, The Cancer Imaging Archive, doi:10.7937/K9/TCIA.2016.V6PBVTDR (CC BY 3.0).

What 2D slices make you do in your head

WHAT SLICES LEAVE TO YOU

Four jobs for your head

  1. 1

    Mental reconstruction

    Hundreds of cross-sections, assembled into one object that lives only in your head. You can't hand it to the anaesthetist or the patient.

  2. 2

    Spatial relationships

    “Is the tumour touching that vessel, or just next to it?” On slices, you answer by scrolling back and forth across several images.

  3. 3

    Oblique measurement

    A vessel or valve running at an angle looks larger in cross-section than it really is. Device sizing depends on the right plane.

  4. 4

    Reader-to-reader variation

    The 3D picture is rebuilt privately, so two clinicians can look at the same scan and plan different things. Partly expertise, partly the medium.

Experienced readers are very good at this. It is still four jobs.

Reading slices is a skill, and experienced readers are very good at it. It still leaves four jobs to the reader’s head.

From one de-identified chest CT: a pulmonary vein running at 45 degrees to the scanner's slice plane, shown in red in 3D with both planes. On the slice as scanned its cross-section is an oval 8.8 by 6.3 mm; on a plane at right angles to the vessel it is 7.0 by 5.5 mm. Both outlines are traced from the CT intensities at half maximum.
From a de-identified Elucis case.

Anatomy rarely lines up with the scanner’s axes, so build the right oblique plane first.

When 3D adds value, and when it doesn’t

WHEN 3D ADDS VALUE

Mostly complex cases. Less for experienced readers of simple ones.

CASE COMPLEXITY

SimplerMore complex

Simple liver models: VR and desktop comparable

Complex liver models: VR beat desktop

Zolkin et al. 2026 · randomised crossover · 58 medical students. The authors can't isolate which immersive features drove the difference.

READER EXPERIENCE

Less experienceMore experience

More than ten years: no change in treatment using any method (3D print, VR glasses or a 3D display)

Muff et al. 2022 · 20 physicians

NOT EVERY STUDY FOUND A BENEFIT

Orbital CT before tear-duct (DCR) surgery: the VR system did not help surgeons interpret the CT better. ENT surgeons and consultants read the anatomy more accurately than ophthalmologists and residents.

Priel et al. 2025 · 6 surgeons, 10 patients

benefit reported comparable / no benefit
Sources: 2, 4, 9

The honest answer: mostly in complex cases, and less for experienced readers of simple ones.249 If the anatomy is standard and the plan is obvious from the slices, a model may add time without adding information.

Reach for 3D when the anatomy is unusual, the relationships are the hard part, the team needs one shared picture, or a device has to fit.

What a good 3D model needs

WHAT A TRUSTWORTHY MODEL NEEDS

Four steps from scan to plan. Three places it can go wrong.

  1. Scan

  2. Segment

  3. Check against the source slices

  4. Plan, with clinical judgement

  1. 1

    Scan

    Thin, ideally near-isotropic slices, in the right contrast phase. A model only shows what its series shows.

    • Where a model can fail

      Thick slices give stair-stepped surfaces and can hide small structures.

    • Where a model can fail

      Lyuksemburg et al. could not build one patient's model because the 2D images were poor quality.

  2. 2

    Segment

    Decide which voxels belong to which structure. Tools give a starting point; a person checks it.

    • Where a model can fail

      Grey values are ambiguous (vessel or lymph node?). That's a clinical call, not a rendering setting.

    • What it costs in time

      Colombo et al., 107 cranial cases: mean segmentation time 39.4 ± 20.4 minutes.

  3. 3

    Check against the source slices

    • Where a model can fail

      Croci et al., 8 complex spine cases: VR did not remove the need to review the multiplanar reconstructions.

  4. 4

    Plan, with clinical judgement

Sources: 1, 8, 10

A model is only as good as the images and judgement behind it. Segmentation means deciding which voxels belong to which structure; automated tools give a starting point that a person still has to check.10 More: /research/q/segmentation-time.

Your options: screen, print or headset

YOUR OPTIONS

Screen, print or headset. None is “best”.

3D on a screen

Surface or volume rendering on a flat monitor

Strengths
Already on most workstations. Quick. Fits the reading-room workflow.
Trade-offs
Depth comes from rotation, not stereo.
Watch for
Volume rendering alone can look convincing while hiding ambiguity.

3D printing

A physical object you can hold

Strengths
Tactile. No hardware needed to view. Can go to the OR.
Trade-offs
Hours to days to print. Cost per model. Fixed once printed. Hard to see inside.
Watch for
Stale models if the plan changes.

VR / immersive

A life-size, stereoscopic model you can walk around and look inside

Strengths
True depth. Scale up or down. Slices and model together. Several people can join.
Trade-offs
Needs a headset and setup. Some people get motion-sick. A new habit to learn.
Watch for
A flat workstation copied into a headset adds little.
Sources: 4, 7

The studies don’t crown a winner. In Muff and colleagues’ study, VR glasses were rated best for understanding the pathology in most disciplines, but the 3D display was rated best for ease of use across every level of experience.4

Wellens and colleagues found no difference in anatomical assessment between 3D prints and AR holograms for children with Wilms tumours.7 See /research/q/vr-vs-3d-printing.

What the evidence says, and its limits

KEY NUMBERS FROM THE RESEARCH LIBRARY

Each from a single study

92% vs 54%

How often the plan matched the operation performed: VR model vs 2D imaging

VR model2D imaging

Operating surgeon

92%
54%

Consulting surgeon

69%
23%

Lyuksemburg et al. 2023 · 13 prospective cases

Complex vs simple

VR beat a desktop display on complex liver models. Simple ones were comparable.

Zolkin et al. 2026 · 58 students

10+ years

Physicians with more than ten years' experience reported no change in treatment with any 3D method.

Muff et al. 2022 · 20 physicians

Sources: 1, 2, 4
  • Plans closer to the operation. Lyuksemburg and colleagues compared 2D-based and VR-based plans with the operation performed, in 20 complex oncologic resections at one centre.1
  • Faster, more accurate MRI reads. El Beheiry and colleagues had 18 breast surgeons read MRI as slices and in VR. VR reads were significantly faster and better at identifying the affected breast; tumour-quadrant accuracy improved for practising surgeons but not significantly for residents.3
  • Less variation between surgeons. Dust and colleagues had 12 trauma surgeons plan 22 tibial plateau fractures. Mixed reality gave the highest agreement on approach and patient positioning, most among junior surgeons.5
  • Measurements that hold up. In 60 consecutive TAVI patients, Kanschik and colleagues found no significant differences and strong correlations between valve sizing in VR and in standard CT software.6

Limits. Most of these studies are small, single-centre and often retrospective. They measure what clinicians understood or planned, not what happened to patients. The benefit isn’t universal. Read the evidence as promising for complex cases and still maturing. See /research/q/plan-change and /research/q/randomized-trials.

Questions to ask before you plan in 3D

Yes to the first four: 3D likely helps. Yes to the last four: go ahead.

BEFORE YOU PLAN IN 3D

Eight questions

IS THIS CASE WORTH 3D?

Yes: 3D likely helps

  • Is the anatomy unusual, distorted or congenital?

    If no: Slices may be enough

  • Is the key question a spatial relationship (abutment, clearance, approach)?

    If no: Slices may be enough

  • Does a device or implant need sizing on an oblique structure?

    If no: Standard measurements may do

  • Do several people need one shared picture?

    If no: Your usual review may do

CAN YOU TRUST THE MODEL?

Yes: go ahead

  • Is there a thin-slice series in the right contrast phase?

    If no: Fix the imaging first, or expect a weaker model

  • Is someone with clinical knowledge doing or checking the segmentation?

    If no: Don't plan on it yet

  • Can you check the model against the original slices?

    If no: Treat it as a picture, not a plan

  • Is there time to build it before the decision is made?

    If no: Keep 3D for the cases that matter most

How Elucis does this

Elucis is FDA-cleared (510(k) K220649)11 software for building 3D models from medical images and planning with them. Its cleared indications are as a software interface and image segmentation system that transfers medical imaging information to an output file, and for measuring and treatment planning, used in conjunction with expert clinical judgement.

Slices and model side by sideCheck every surface against its source.
You build the model, in 3DManual and semi-automatic segmentation, in VR or on the desktop.
Measure and plan on the patient’s anatomyMeasure and plan in the same scene.
Plan togetherJoin one scene from VR or desktop, in the room or remotely.
ExportImages, measurements, 3D models (e.g. for 3D printing).

Elucis supports the clinician’s judgement; it doesn’t replace it.

See the evidence for yourself at /research, where every study links to its source, including the ones that found no benefit. Want to see it on one of your own cases?

Request a demoBrowse the research

References

  1. Lyuksemburg V, Abou-Hanna J, Marshall JS, et al. Virtual Reality for Preoperative Planning in Complex Surgical Oncology: A Single-Center Experience. The Journal of surgical research. 2023. PMID 37540972. https://doi.org/10.1016/j.jss.2023.07.001
  2. Zolkin A, Rüger C, Remde C, et al. Effect of virtual reality on spatial-anatomical understanding in preoperative liver surgery: a randomized crossover study. Scientific reports. 2026. PMID 42420374. https://doi.org/10.1038/s41598-026-61007-6
  3. El Beheiry M, Gaillard T, Girard N, et al. Breast Magnetic Resonance Image Analysis for Surgeons Using Virtual Reality: A Comparative Study. JCO clinical cancer informatics. 2021. PMID 34767435. https://doi.org/10.1200/cci.21.00048
  4. Muff JL, Heye T, Thieringer FM, et al. Clinical acceptance of advanced visualization methods: a comparison study of 3D-print, virtual reality glasses, and 3D-display. 3D printing in medicine. 2022. PMID 35094166. https://doi.org/10.1186/s41205-022-00133-z
  5. Dust T, Henneberg JE, Hartel M, et al. Mixed reality improves agreement on surgical approach selection and patient positioning in tibial plateau fracture planning compared to CT, 3DCT and 3D printing. European journal of trauma and emergency surgery : official publication of the European Trauma Society. 2026. PMID 42329453. https://doi.org/10.1007/s00068-026-03230-4
  6. Kanschik D, Haschemi J, Heidari H, et al. Feasibility, Accuracy, and Reproducibility of Aortic Valve Sizing for Transcatheter Aortic Valve Implantation Using Virtual Reality. Journal of the American Heart Association. 2024. PMID 39041603. https://doi.org/10.1161/jaha.123.034086
  7. Wellens LM, Meulstee J, van de Ven CP, et al. Comparison of 3-Dimensional and Augmented Reality Kidney Models With Conventional Imaging Data in the Preoperative Assessment of Children With Wilms Tumors. JAMA network open. 2019. PMID 31002326. https://doi.org/10.1001/jamanetworkopen.2019.2633
  8. Croci DM, Guzman R, Netzer C, et al. Novel patient-specific 3D-virtual reality visualisation software (SpectoVR) for the planning of spine surgery: a case series of eight patients. BMJ Innovations. 2020. https://doi.org/10.1136/bmjinnov-2019-000398
  9. Priel A, Hadida Barzilai D, Tejman-Yarden S, et al. Pre-Operative Planning of a DCR Surgery Using Virtual Reality. Seminars in ophthalmology. 2025. PMID 39028204. https://doi.org/10.1080/08820538.2024.2378341
  10. Colombo E, Regli L, Esposito G, et al. Mixed Reality for Cranial Neurosurgical Planning: A Single-Center Applicability Study With the First 107 Subsequent Holograms. Operative neurosurgery (Hagerstown, Md.). 2023. PMID 38156882. https://doi.org/10.1227/ons.0000000000001033
  11. US Food and Drug Administration. 510(k) K220649, Elucis (Realize Medical, Inc.). Decision 2023-01-17. https://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfpmn/pmn.cfm?ID=K220649

Panel illustrations and the video: 3D models and CT images from the TotalSegmentator dataset (Wasserthal J, et al. Radiology: Artificial Intelligence 2023;5(5):e230024), licensed CC BY 4.0. Video measurements are taken from the dataset’s artery segmentation. https://doi.org/10.5281/zenodo.10047292