Artificial Intelligence in Radiotherapy - From concept to clinic and back again

Artificial Intelligence in Radiotherapy - From concept to clinic and back again

by Professor Raj Jena

Applied Sciences & Engineering Medicine & Health MJCLOncologyMJCL1Radiotherapy
🎙 Professor Raj Jena 👥 2K 📅 November 3, 2025 ⏱ 67 min 👁 263 📄 expert opinion 🧭 2026-08-16
Available in: English (current) Français

Keywords

radiotherapyAIdeep learningsegmentationclinical translation

Summary

Professor Raj Jena, a clinical oncologist and AI researcher at the University of Cambridge, delivers a talk on his journey of integrating AI into radiotherapy. He begins with his early interest in neural networks, inspired by a paper on backpropagation, and his later collaboration with Microsoft on the InnerEye project, which used deep learning for medical image segmentation. He then discusses the challenges of crossing the ‘valley of death’ between research and clinical practice, highlighting his experience with a commercial venture that was eventually open-sourced. He details the OSAIRIS project, an NHS-funded initiative that successfully deployed an AI-powered segmentation tool in the hospital, achieving significant time savings and demonstrating rigorous clinical evaluation. He also touches on his current work with Microsoft on discovery research and the STELLA project, which aims to develop an AI-powered radiotherapy machine for low- and middle-income countries. The talk provides insights into the practicalities of clinical AI deployment, including regulatory hurdles, automation bias, and the importance of robust evaluation.

163 words

Critical Evaluation

Value of the Information & Strength of the Argument

The talk offers valuable insights into the practical challenges of translating AI research into clinical practice, drawing on the speaker’s extensive personal experience. The argumentation is coherent and well-structured, tracing a clear narrative from concept to clinic and back. The speaker provides concrete examples, such as the OSAIRIS project, to illustrate his points, and he addresses important issues like automation bias and regulatory compliance. However, the talk is primarily anecdotal and lacks rigorous scientific evidence or comparative data, making it more of an expert opinion than a systematic review.

Scientific Rigor, Source Quality, Title Accuracy

The speaker is a credible expert with direct involvement in the projects discussed, lending authority to his account. However, he does not cite specific sources or references during the talk, and the description provides no links to publications or further reading. The title accurately reflects the content, which is a personal narrative of AI in radiotherapy. The talk is not a formal scientific presentation but rather an engaging lecture for a general audience, so the lack of formal citations is understandable.

185 words

Title / Content Match

The title accurately reflects the content, which traces the speaker's journey from AI concept to clinical deployment and back to research.

Quality & Reliability

8/10

The speaker is a clinical professor with direct involvement in the projects described, providing first-hand experience. The talk is an expert opinion with practical examples, but lacks formal citations or peer-reviewed references.

Key Moments

Markers derived by PSI from the transcript: the creator did not define chapters.

Cited Sources

  • No specific sources cited in the video — The speaker mentions the BRATS challenge and the InnerEye project but does not provide direct references.

Concurring Sources

  • InnerEye project — The speaker's collaboration with Microsoft on medical image analysis.
  • OSAIRIS project — The speaker's project on AI-powered radiotherapy segmentation.

Contribution & Novelties

The talk provides a unique first-hand account of the entire lifecycle of an AI medical device, from research to clinical deployment and back to research. It offers practical insights into the challenges of regulatory approval, clinical evaluation, and the ‘valley of death’ that are rarely discussed in academic papers. The speaker’s emphasis on automation bias and the need for rigorous evaluation is particularly valuable.

Pour aller plus loin :

  • Deep learning in medical imaging — Overview of deep learning applications in medical imaging.
  • Technology Readiness Level — Framework for assessing technology maturity.
  • Automation bias — Cognitive bias relevant to AI in clinical settings.

103 words

Radar Profile

The radar profile shows high scores in information quantity, quality, and reliability, reflecting the speaker's expertise and detailed account. The technical level is moderate, making the talk accessible to a broad audience. The overall high scores indicate a valuable and credible presentation.

Reliability 8/10

💬 No comments were provided for analysis.