Paper Proffered Program Therapy Physics

AI-Driven Web Platform for Radiotherapy Review Training

Abstract
Purpose

Radiotherapy plan quality education is limited by case availability, frequent clinical interruptions, and difficulty aligning schedules. These barriers leave residents underprepared to identify and correct suboptimal plans, potentially compromising future outcomes. To address these challenges, we developed a web-based, AI-driven platform to simulate realistic planning errors and enable interactive correction. This can provide virtually unlimited educational opportunities to support high-volume, active learning globally.

Methods

Multi-modal deep learning models were trained to introduce errors into dose distributions using natural language and to support iterative improvement through user directives. The web platform leverages these to create progressively complex volume-modulated arc therapy dose distributions for head-and-neck and cervix with two error classes: (1) poor OAR sparing and (2) compromised target coverage from over-prioritizing OARs. Generation parameters are randomized so trainees do not review the same example twice, and cases are organized into difficulty tiers to individually tailor learning. The platform can generate or modify examples in approximately 20 seconds, enabling rapid feedback and continuous engagement. Trainees review dose distributions, compare plans, and iteratively improve quality through natural language prompts, mimicking clinical planning decisions. Participants from the United States and Indonesia completed initial assessments and then weekly quizzes while training asynchronously.

Results

Three experts (one physicist and two physicians) reviewed 95 simulated cases and rated all as appearing like human-created plans. Experts agreed with 85% of difficulty assignments; all others differed by one tier, supporting realism and stratification. Trainee plan review median scores increased from 37.5% to 44.8% after at least 3 weeks of training for cervical cancer cases, and from 37.5% to 50% for head-and-neck cancer cases.

Conclusion

This AI-driven web training platform addresses critical gaps in radiotherapy education by providing large-scale, realistic, interactive training to enhance residents’ review skills. Results demonstrate improved plan review skill after 3 weeks of training.

People

Related

Similar sessions

Poster Poster Program
Jul 19 · 07:00
Python-Based Automation Framework for Annual Machine QA Data Archiving In Qatrack+

Annual water-tank measurements help ensure beam characteristics remain consistent with commissioning baselines. However, the lack of a standardized processing workflow and decentralized data storage makes it difficult to analyze...

Syed Bilal Ahmad, PhD
Therapy Physics 0 people interested
Poster Poster Program
Jul 19 · 07:00
User Expectations and Current Availability of HDR Brachytherapy Audits In Europe

The aim of this work was to evaluate the need to implement more dosimetric audits in high‐dose‐rate brachytherapy (HDR-BT) in Europe and to identify which characteristics such audits should meet according to users.

Javier Vijande, PhD Laura Oliver Cañamás
Therapy Physics 0 people interested