Truace
Subscribe

ROECSG report: Real-world implementation patterns and perceptions of artificial intelligence in radiation oncology

Purpose: While international surveys have examined radiation oncology professionals' attitudes toward artificial intelligence (AI) and documented generally low early utilization, actual implementation patterns and the depth of integration among engaged practitioners remain unclear. This multi-institutional survey aims to characterize current AI implementation, identify critical barriers to optimization, and evaluate the psychological factors driving clinical trust. Methods and materials: A de-identified survey c…

By the Truvace desk2 min read1 sourcePeer-reviewedTRV-2026-2202· cite & verify
ROECSG report: Real-world implementation patterns and perceptions of artificial intelligence in radiation oncology
"McArdle Laboratory for Cancer Research" by wackybadger is licensed under CC BY-SA 2.0. To view a copy of this license, visit https://creativecommons.org/licenses/by-sa/2.0/.

In brief

A multi-institutional survey of 80 radiation oncology professionals, mostly radiation oncologists and medical physicists, examined how AI is actually used in clinics. Most respondents reported using AI, but use centered on auto-contouring, with far less in diagnosis, prognostication, and treatment planning.

The findings suggest enthusiasm is running ahead of understanding, since respondents rated their technical grasp near the scale midpoint yet mostly welcomed more AI. Authors call for institutional governance and AI-literacy education, but the sample was self-selected, so the results may not represent the wider field.

Main points

  1. 85.0% (68/80) of respondents reported current AI use, though self-selection likely inflates this figure.
  2. Auto-contouring accounted for 88.2% of AI users, versus 14.7% for diagnosis/prognostication and 23.5% for treatment planning.
  3. 60.3% of users employed AI for task automation without influencing clinical judgment.
  4. Trust correlated with perceived time savings (r = 0.57), clarity of use (r = 0.54), and technical understanding (r = 0.51).
  5. Transparency (38%) and validation protocols (31%) were the main barriers to deeper implementation.

The problem

AI implementation in radiation oncology is concentrated in auto-contouring, with limited use in diagnosis and planning, while understanding is middling and transparency and validation barriers block deeper adoption.

The rundown

The survey used a de-identified instrument covering six domains, distributed through a medical physics listserv and at the 2025 AAPM Annual Meeting. It used 6-point Likert scales to reduce central tendency bias, and analyzed data with Pearson correlations and Welch's t-tests.

Respondents were 46.3% radiation oncologists and 37.5% medical physicists. Mean trust in AI was 3.55 and self-rated understanding was 3.44, yet 88.8% welcomed greater AI integration.

Qualitative responses were analyzed with inductive thematic analysis (κ = 0.78). The authors conclude that the gap between enthusiasm and understanding calls for active institutional governance and standardized AI-literacy education.

Sources

  1. Peer-reviewedJournal of Applied Clinical Medical Physics2026-11-01

The debate