Artificial intelligence: the impact on clinical radiology and clinical oncology
RCR leading the way
There is much investment in research and development of artificial intelligence (AI) across healthcare, with the hope of reaping great rewards. The RCR – as standard setter for its two specialties of clinical radiology and clinical oncology – will lead in establishing the framework for AI. The RCR has defined four broad work streams:
- Exploring training and development needs
- Establishing frameworks for regulation
- Quality assurance as AI implementation come on stream
- Engagement and partnership working
Training and development
The RCR will explore where the core curriculum for the specialties may need to evolve to accommodate and support the implementation of AI as it becomes more embedded in service delivery.
CPD and continuous learning for members and fellows will also be a focus.
There is as yet no regulatory framework for AI in healthcare.The RCR will collaborate with stakeholders and focus on a number of principles:
- Accountability – where does/should accountability lie for the development, testing, and clinical introduction of AI
- Veracity – arriving at an accepted standard for data validation
- Reactivity – as AI evolves, what are the trigger points to future-proof the framework
- Universal Applicability – the framework should be relevant and applicable across all AI developments – seamless incorporation into extant digital healthcare systems such as PACS, RIS, radiotherapy planning systems and the EPR
Quality assurance must pervade all stages of the development of AI through to implementation and beyond. The RCR will develop standards for validation of data and algorithms to safeguard patient outcomes.
Engagement and partnership working
There are three strands to engagement and partnership working:
- Harnessing AI expertise
- Collating existing datasets for the public benefit
- Breaking barriers to data consent
Through these strands, the RCR will become a source of expertise for the media.
On harnessing expertise, the RCR will provide a platform for all: the NHS and its stakeholder organisations, academic communities, industry and, most importantly, patients. The RCR is free from commercial interest or bias and not within the competitive academic world and so will be an honest broker.
On datasets, there is a need for vast amounts of anonymised, appropriately validated electronic patient data in order to train AI technology. The RCR will bring together the key players to enable equitable access to existing and new datasets so as to protect patient data and ensure the NHS retains the benefits of its intellectual property.