Summary

Advice

The technology described in this briefing is artificial intelligence (AI) software for CT brain scans. It is to assess CT images of people with suspected brain abnormalities.
The innovative aspects are that the software automates aspects of detecting brain abnormalities and assists in clinical prioritisation of critical cases.
The intended place in therapy would be to support radiologists in secondary care when they are reviewing CT brain scans of people with suspected brain abnormalities. The technology may be of most benefit when images are not first reviewed by specialist neuroradiologists.
The main points from the evidence summarised in this briefing are from 11 studies. Seven validation studies, including 31,118 CT brain scans, showed the technology to be as effective at detecting intracranial haemorrhages as neuroradiologists. However, study conditions did not reflect clinical practice. Four real-world observational studies including 59,655 CT brain scans suggest the technology may perform well in clinical practice. However, 3 of the studies are reported as abstracts and limited in methodological detail.
Key uncertainties around the evidence or technology are that 6 of the studies in the briefing are abstracts and limited in methodological detail. The evidence base would benefit from well-controlled comparative studies with an appropriate follow-up time to capture patient outcome and time to treatment.
The cost of AI software for CT brain scans is between £8,250 and £80,000 per licence fee every year. The cost of the technology depends on the size of the NHS trust and the number of analyses done. Pay per use is also available for 1 of the listed technologies and costs £45. The resource impact would be greater than standard care. However, this may be offset by faster diagnosis of time-sensitive cases, reducing complications related to delayed treatment.