NICE announced that artificial intelligence (AI)-derived solutions are available for implementation within the NHS as further evidence is collected to bolster their use in aiding the review and analysis of CT brain scans for patients suspected of experiencing a stroke. The recommended software was
e-Stroke
RapidAI
Viz
These AI technologies require Digital Technology Assessment Criteria (DTAC) approval before application.
These tools must be utilized under the supervision of healthcare professionals, and healthcare centres should adhere to established protocols for scan reporting to mitigate the risk of erroneous outcomes. Additionally, centres must ensure that images exchanged between different stroke facilities can be assessed remotely to facilitate decision-making by healthcare professionals located elsewhere.
Further research is warranted for the following AI-based solutions aimed at assisting in the review and analysis of CT brain scans for individuals suspected of having a stroke:
Accipio
Aidoc
BioMind
BrainScan CT
Cercare Perfusion
CINA Head
CT Perfusion 4D
icobrain ct
Neuro Solution
qER
Stroke poses significant challenges to the quality of life for survivors. Timelier access to treatment could enhance clinical outcomes and overall quality of life post-stroke. AI-driven software, when coupled with professional interpretation of CT brain scans, has the potential to expedite decision-making processes in stroke care, such as determining suitability for thrombolysis and thrombectomy treatments. These software solutions operate on fixed algorithms in clinical settings, with AI employed to enhance algorithmic versions.
However, the clinical evidence supporting these software solutions is currently of limited quality. No evidence meeting review inclusion criteria demonstrate their diagnostic accuracy when used alongside professional review. While certain studies suggest expedited or increased treatment access post-software utilization, the extent of these improvements remains ambiguous. A slight uptick in thrombectomy rates due to AI-driven software, as indicated by the economic model, could render the software cost-effective.
Given their widespread NHS usage, these software solutions must always be employed alongside professional review. Notably, improved image sharing between centres, facilitated by these technologies, aids time-sensitive decision-making. Hence, centres should ensure shared images maintain sufficient quality for remote review. Healthcare professionals should exercise caution in altering diagnoses based on software outcomes, and existing reporting protocols should be upheld.
While e-Stroke, RapidAI, and Viz can be utilized within the NHS as further evidence is gathered to ascertain their cost-effectiveness, other technologies lacking evidence regarding their impact on treatment time or access should be restricted to research settings.
Stroke, a severe and life-threatening medical condition, occurs when the blood supply to a part of the brain is significantly impeded. With over 100,000 cases annually in the UK, strokes affect individuals across various age groups, with a quarter occurring in the working-age population. It ranks among the leading causes of disability and early mortality in the UK.
Treatment for stroke hinges on factors such as the underlying cause, duration of compromised blood supply, and severity of resultant damage. Ischemic stroke, the predominant type, necessitates treatments like thrombolysis and thrombectomy to restore blood flow by dissolving clots or mechanically extracting them. Conversely, intracerebral haemorrhage, a less common type, renders such treatments detrimental.
CT brain scans play a pivotal role in guiding treatment decisions. They aid in confirming ischemic stroke, identifying clots, assessing stroke severity, and evaluating brain tissue viability. AI-driven software, with its algorithms, assists in analyzing these scans, detecting abnormalities, and providing supplementary information for clinical decision-making. The goal is to expedite scan reviews, enhance treatment decisions, and ultimately improve patient outcomes.
Aidoc, for instance, features algorithms like Aidoc ICH and Aidoc LVO, designed to detect intracranial haemorrhage and large vessel occlusions, respectively. Similarly, BioMind, BrainScan CT, Cercare Perfusion, and other platforms offer AI-driven solutions tailored to stroke care. These solutions, though promising, necessitate further research to validate their efficacy and impact on clinical outcomes.
The NICE, in its assessment, emphasized the dearth of evidence supporting several AI-driven technologies. Limited data on diagnostic accuracy, coupled with inconclusive findings regarding treatment time and outcomes, underscore the need for rigorous research. While retrospective data analysis may offer insights, comprehensive studies are essential to ascertain the true benefits and risks of AI-driven software in stroke care.
In conclusion, while AI-driven software holds promise in revolutionizing stroke care, its widespread implementation necessitates robust evidence of efficacy and safety. Collaborative efforts between healthcare providers, researchers, and technology developers are imperative to navigate the complexities of stroke management effectively.