Stories from the University of Cambridge

OSAIRIS

  • Rajesh Jena[1], Alex Constantinou[1], Andrew Hoole[1], Abigail Bushi[1], Terry Parlett[2], Dafne Chirivino[2]

    1 Department of Oncology, Cambridge University Hospitals NHS Foundation Trust, Cambridge, UK 2 Cambridge Enterprise, University of Cambridge, Cambridge, UK

  • 2011

  • Rajesh Jena:
    rjena@nhs.net

  • Oktay O, Nanavati J, Schwaighofer A, Carter D, Bristow M, Tanno R, Jena R, Barnett G, Noble D, Rimmer Y, Glocker B, O’Hara K, Bishop C, Alvarez-Valle J, Nori A. 2020. Evaluation of deep learning to augment image-guided radiotherapy for head and neck and prostate cancers. JAMA Network Open 3, e2027426.

  • http://www.camradiotherapy.org.uk/rdst/index.html

  • Artificial Intelligence in Health and Care Award, NHS AI Lab Programme (UK), Cancer Research UK RadNet (UK)

  • https://youtu.be/AHchuDX0nn4

ABOUT THE OPEN-RESOURCE

Background 

Radiotherapy is an essential component of cancer treatment, used every year in more than half a million patients diagnosed with cancer. Planning radiotherapy treatment is an extremely time-consuming task, where the patient’s scans are manually marked up by an oncologist, to select precisely the area to direct the radiation to the target. Thinking about how to speed up this laborious task, Dr Rajesh Jena and his team in the Department of Oncology, at the University of Cambridge; and Addenbrooke’s Hospital, in collaboration with Microsoft Research, used artificial intelligence and machine learning to train computers to annotate the scans automatically. The project OSAIRIS (Open-Source Artificial Intelligence Radiotherapy Imaging System) has shown how the computers can perform the preparation around 2.5 times faster than a human, saving time for busy clinicians, and allowing patients to receive more effective and faster treatment. 

Function

It is a machine learning tool to analyse patient scans and speed up preparation for radiotherapy treatment.

Development process

The project was initially developed in 2012 to commercial standards but subsequently released as an open source repository in 2019. “The key thing that we learned was the difference between software risk, which tends to be the focus of people who do software development engineering, and clinical risk of a piece of software that is open source,” says Dr Jena. Even though the original repository from which OSAIRIS was forked had software engineering principles rights, it took more than 18 months of work around the clinical risk in order to turn it into an open source medical device. 

Target user

Hospitals that perform radiotherapy treatment.

Comparison to other technologies

“We believe OSAIRIS is one of the first open source software as a medical device deployment,” says Dr Jena. Even though there are other software development kits available for a wide set of medical images, none of those are yet ready for a clinical implementation. With OSAIRIS the user gets both the code repository, and also the documentation for a technical file that can be used for medical purposes.

IMPACT

Current use

OSAIRIS has been used in the University Hospitals in Cambridge and in Birmingham. The University College London Hospitals (UCLH) are also implementing OSAIRIS.

Successful stories 

OSAIRIS has been running in Addenbrooke’s Hospital since October 2020, and at the moment is the standard system used for radiotherapy preparation for two different types of cancers. Every month, around 120 patients benefit from this technology. Dr Jena explains that this preparation stage, which used to take the oncologist about 90 minutes, now takes around
35 minutes.

Open source choice

Dr Jena is passionate about open source tools and he points out that all the research tools developed by his group are submitted as open source. To the best of his knowledge, OSAIRIS is the first open source artificial intelligence in imaging to be created within the NHS. Furthermore, because OSAIRIS is a data-driven AI technology, it learns from patient data.

GOING FORWARD - WHERE TO IN THE NEXT 3-5 YEARS?

Right now, Dr Jena and his team are working in collaboration with Cambridge Enterprise to understand the bureaucracies behind how OSAIRIS can be offered as an open source software. In simple words, although OSAIRIS is open source and technically anyone can implement it, in practice to have it as a medical device, it requires a level of support that is more than would be achieved just by downloading the code. 

Qualitative evaluation of expert and autogenerated contours on head and neck computed tomography scans. © 2020, Oktay et al., licensed under CC-BY 4.0. Reproduced from JAMA Netw Open. 3, e2027426, https://doi.org/10.1001/jamanetworkopen.2020.27426.

“We believe OSAIRIS is one of the first open source software as a medical device deployment.”

Rajesh Jena