Applications and Opportunities of Artificial Intelligence in Clinical Decision-making 20 May 2020
The explosion in volume and availability of data sources combined with advanced analytical methods, is leading to a revolution in healthcare. Artificial intelligence (AI) and machine learning (ML) are being used across various applications in medicine from drug discovery, biomarker identification, through to clinical trial design and patient enrollment. AI and ML techniques also show great promise for clinical decision making, for example, using algorithms trained on imaging data. By applying advanced analytics to vast image datasets, we can identify patterns and help support the decisions made by clinicians. As Tim Hoogenboom pointed out in a blog for Imperial College (1), the cost, time, and accuracy benefits of these approaches can be immense. In rare diseases, where an individual clinician may have seen very few patients with a particular condition, or in oncology where medics must very precisely assess response to treatment, AI enables us to benefit from the data acquired from a far greater pool of patients. And the benefit grows over time, since as we gain more data from more patients, so the accuracy of the algorithm improves.
With the COVID-19 pandemic disrupting medical provision across the globe, healthcare providers have turned to AI to help improve efficiency and delivery of patient care in non-COVID diseases, as well as for helping with the diagnosis of COVID-19 itself. A recent article in the BBC News (2) discusses how, during the peak of the pandemic in Wuhan, doctors used algorithms developed by Axial AI to scan the lungs of thousands of patients. By analyzing CT imagery, it was possible to assess in a matter of seconds whether a patient had a high risk of viral pneumonia from coronavirus. While this particular algorithm is not currently in use in the UK, the BBC article reported that several British hospitals are deploying other AI tools to help clinicians interpret chest X-rays more swiftly.
In a broader use case (3), Lancashire and Cumbria NHS trust plan on introducing artificial intelligence-assisted (AI) brain scans to speed up stroke treatment later this year. Early results show that outcomes have been improved for patients through the use of the tool. The efficiencies enabled by the technology can help to offset healthcare challenges caused by the coronavirus pandemic.
However, despite the promise, As Kelly et al. point out (4) , there are barriers to overcome to ensure we achieve the full potential of these techniques. Indeed, it is critical that we thoroughly assess how these approaches affect the quality of care and patient outcomes. As with any application of machine learning and artificial intelligence, quality and comprehensiveness of the available data are paramount. We must also ensure the robustness of the analytical techniques to mitigate bias and assure accuracy.
In all cases, we should see such techniques as a way to complement, rather than replace our clinical judgement. ur role as physicians is more important now than ever. Still, such innovations can help us to become more efficient and more accurate, allowing us to deliver higher quality provision and enabling us to respond effectively to major crises such as the one we face today.
(2) https://www.bbc.co.uk/news/business-52483082
(4) https://bmcmedicine.biomedcentral.com/articles/10.1186/s12916-019-1426-2