Nine AI technologies can be used to help plan treatment of patients undergoing external beam radiotherapy for cancers, NICE has said in draft guidance.
This is the first piece of NICE guidance to recommend the use of AI to aid healthcare professionals in their roles.
The draft guidance says all contours created by AI must still be reviewed by a trained healthcare professional and edited as needed before being used in radiotherapy treatment planning. Early evidence suggests that using AI is quicker than manual contouring even when including time for healthcare professional review and edits.
Evidence seen by NICE’s independent medical technologies advisory committee suggests that AI technologies generally produce similar quality contours of organs at risk as those carried out manually, with most only needing minor edits. The change to AI automated mark-up is not expected to affect patient outcomes, said NICE.
Clinical experts advising the independent committee estimated a time saving of 10 minutes to 30 minutes per plan depending on the amount of editing needed, while the clinical evidence presented to the committee suggests it may range between 3 to 80 minutes of time saved per plan.
SoR executive director of professional policy Charlotte Beardmore welcomed the draft guidance but said it was not a replacement for staff and caution was needed.
“It is critical there is evidence to underpin the safe application of AI in this clinical setting,” she said. Using AI would still require input by a therapeutic radiographer or another member of the oncology multi-professional team, she added. “Investment in the growth of the radiography workforce remains critical.”
NHS England’s radiotherapy dataset shows that there were 134,419 radiotherapy episodes in England in April 2021 to March 2022, of which a significant proportion require complex treatment planning.
AI technology costs ranged from £4 to £50 per plan and included software and other associated costs including healthcare professional training.
The possible resource benefits calculated by NICE shows that if the lowest time saving of 3 minutes per plan is assumed and 75,000 plans are generated using AI auto-contouring, the time saved would be 3,750 hours. This would increase to 52,500 hours for 75,000 plans and a medium time saving of 42 minutes being assumed. At the higher end of the scale, 100,000 hours would be saved for 75,000 plans and a time saving of 80 minutes per plan being assumed.
Due to a lack of robust data on current practice and other variables such as the costs and time involved, more evidence needs to be generated over the next three years so a full cost/benefit analysis can be carried out by NICE.
A NICE consultation on the recommendations has now begun and comments must be submitted by Friday 25 August 2023. The SoR will be responding to the consultation.