# Virtual reality volumetric rendering versus cross-sectional imaging for pancreatic cancer resectability assessment: a pilot randomized controlled reader study

## TL;DR

This study compares virtual reality and traditional 2D imaging for assessing pancreatic cancer resectability, finding that 2D imaging performs better in accuracy and agreement among surgeons.

## Contribution

The study introduces a pilot randomized controlled trial comparing VR volumetric rendering and cross-sectional imaging for pancreatic cancer assessment.

## Key findings

- Cross-sectional imaging achieved substantial inter-rater agreement (κ = 0.609) compared to VR (κ = 0.127).
- Diagnostic accuracy was higher with cross-sectional imaging (84.7%) than with VR (79.7%).
- VR users reported significantly lower confidence in their assessments compared to cross-sectional imaging users.

## Abstract

Current imaging assessment for pancreatic cancer resectability demonstrates problematic inter-observer variability, with only fair-to-moderate agreement among experienced raters. Virtual reality technology offers stereoscopic three-dimensional visualization that may improve diagnostic accuracy and agreement. However, optimal visualization strategies for clinical adoption remain unclear.

Ten hepatopancreatobiliary surgeons from two high-volume centers were randomized 1:1 to assess twelve contrast-enhanced CT cases using either VR volumetric rendering or CSI. Primary outcomes included inter-rater agreement, diagnostic accuracy against expert reference standard, assessment time, and surgeon confidence. Statistical analysis employed Fleiss’ κ for inter-rater agreement and two-sided Mann–Whitney U tests on surgeon-level summary measures for between-group comparisons.

CSI display on 2D screens achieved substantial inter-rater agreement for resectability assessment (κ = 0.609) while VR demonstrated only slight agreement (κ = 0.127). Diagnostic accuracy was superior with CSI (84.7% vs. 79.7%), with the most pronounced difference in resectability determination (83.3% vs. 58.3%, p = 0.033). VR users reported significantly lower confidence (4.85 ± 1.15 vs. 6.32 ± 0.77, p = 0.028). Assessment times were comparable between groups (median 313.5 s vs. 327.5 s, p = 1.00).

In this preliminary investigation, our VR visualization strategy demonstrated lower diagnostic accuracy and inter-rater agreement than CSI. However, prior studies suggest that VR systems employing alternative, hybrid visualization approaches may improve inter-rater agreement, indicating that visualization strategy, rather than VR technology per se, is the primary determinant of utility.

DRKS00033932 (German Clinical Trials Register), registered prospectively.

The online version contains supplementary material available at 10.1186/s42234-026-00202-2.