How are you taking part in this consultation?

You will not be able to change how you comment later.

You must be signed in to answer questions

    The content on this page is not current guidance and is only for the purposes of the consultation process.

    Artificial intelligence (AI)-derived software to help clinical decision making in stroke

    Closed for comments This consultation ended on at   Request commenting lead permission

    1 Recommendations

    NICE is aware that companies are reviewing their CE marking in response to changing regulations and advances in digital health technologies.

    Can be used in the NHS with evidence generation

    1.1 The following artificial intelligence (AI)-derived software can be used in the NHS, to support review and reporting of CT brain scans for people who have had a stroke, while more evidence is generated:

    • e-Stroke

    • RapidAI.

    These technologies can be used once they have Digital Technology Assessment Criteria (DTAC) approval from NHS England.

    1.2 The software should only be used with healthcare professional review and centres should maintain existing scan reporting protocols to reduce the risk of incorrect results. Centres should ensure that images shared between different stroke centres can be remotely reviewed to help with decision making by healthcare professionals at a different site.

    Can only be used in research

    1.3 More research is needed on the following AI-derived software to support review and reporting of CT brain scans for people who have had a stroke:

    • Accipio

    • Aidoc

    • BioMind

    • BrainScan CT

    • Cercare

    • CINA Head

    • CT Perfusion 4D

    • icobrain ct

    • Neuro Solution

    • qER.

    See

    1.4 Access to the technologies listed in section 1.3 should be through company or research funding (non-core NHS funding).

    Evidence generation and further research

    1.5 Further evidence is needed on:

    • the impact of the addition of AI-derived software on a healthcare professional's ability to identify people for whom thrombolysis and thrombectomy is suitable (see section 3.7)

    • test failure rate with causes of failures (see section 3.12)

    • the impact of using the software on time to thrombolysis or thrombectomy (see section 3.9)

    • the impact of using the software on how many people have thrombolysis or thrombectomy (see section 3.10).

    Why the committee made these recommendations

    Stroke adversely affects quality of life for many people who survive it. Faster and greater access to treatment could improve clinical outcomes and so quality of life after stroke. AI-derived software used alongside healthcare professional interpretation of CT brain scan images could guide and speed up decision making in stroke, for example, decisions on thrombolysis and thrombectomy treatment. The AI-derived software uses fixed (or static) algorithms in clinical practice, with AI used to derive new versions of the algorithm.

    Clinical evidence on the software is limited in quality. There is no evidence on their diagnostic accuracy when used alongside healthcare professional review that met review inclusion criteria. Some studies on 2 technologies (e-Stroke and RapidAI) in clinical practice suggest that people had faster or greater access to treatment after using the software, but it is unclear to what extent this is an effect of the software. A small increase in people having thrombectomies because of AI-derived software in the economic model would likely make the software cost effective.

    The software is already widely used in the NHS, and should always be used with healthcare professional review. Because an important potential benefit of the technologies is improving image sharing between centres to help with time-sensitive decision making, centres should ensure that shared images are of sufficient quality to allow remote image review. Healthcare professionals should be cautious when changing their findings based on software results. Existing reporting protocols (or those used before the AI-derived software was adopted in a centre) should be maintained in centres using these technologies.

    e-Stroke and RapidAI can be used in the NHS while further evidence is generated to help better determine their cost effectiveness. Other technologies that have no evidence on how they impact time or access to treatment should only be used in research.