Clinical Implementation of AI-Generated Contours for Increased Efficiency in Radiotherapy Treatment Planning

Loading...
Thumbnail Image

Contributor Profession (Non Medical)

Publication Date

Item Type

Conference Abstract
Conference Poster

Language

en

Subject

RADIOTHERAPY, ARTIFICIAL INTELLIGENCE, ALGORITHMS

Subject Headings

Journal Title

Journal ISSN

Volume Title

Abstract

Background: In Radiotherapy planning, ‘contouring’ involves outlining organs on CT/MR images. Historically, this has been a manual and time-consuming process. Recent guidance has been published supporting the use of AI-generated ‘auto-contours’. [1,2] In 2023, LTHT partnered with RaySearch Laboratories to help develop a commercial AI model. A multi-disciplinary team (MDT) safely implemented this novel technology into the clinic to help combat increasing workforce pressures. Aims: 1) Evaluate the accuracy of AI contours. 2) Safely integrate AI technology into a complex clinical pathway. 3) Quantify workflow efficiencies. Methods: A retrospective sample of approx. 20-30 patients was selected for thorax, abdomen, head & neck treatment sites. AI contours were compared against gold standard manual contours. Contour quality was assessed visually, geometrically (DICE), and dosimetrically (dose differences). Pre- and post-implementation audits compared manual versus AI-assisted contouring times. Results: AI contours were consistently geometrically accurate (DICE >0.7) and dosimetrically comparable (<2% dose difference). Occasional clinically significant differences were attributed to differences in contouring definitions (e.g. sup/inf extent of larynx), unusual anatomy, or imaging artefacts. Auto-contouring has been commissioned for thorax, abdomen, and head & neck. This required strong MDT communication and a robust framework for identifying systematic differences and exclusion criteria. All AI contours are reviewed and edited before use. Staff feedback has been positive. Of 6,550 patients treated in 2024, 4,200 (65%) used auto-contours. Mean contouring time was reduced from 5-60mins to 3-28mins, saving approx.16.5 minutes per patient (155 person-days annually). Greatest gains (>45 minutes) were seen in Oesophagus and Liver cases. Conclusion: AI auto-contouring has been successfully and safely implemented into the Radiotherapy pathway. Increased efficiency has helped help combat increasing workforce pressures and enabled staff to focus on higher-value aspects of patient care. The Radiotherapy department will continue working with commercial partners to expand auto-contouring to additional sites (e.g. MR brain).

Journal

Link to Publisher Site (DOI)

Endorsement

Review

Supplemented By

Referenced By