Even before the COVID-19 pandemic had spread across the globe, Artificial Intelligence (AI) had already begun to redraw the landscape of medical science. Indeed, the remarkable predictive power of AI has seen it take centre stage in the unprecedented fight against COVID-19.
In this, the second of our articles relating to COVID-19, we look at innovation in the COVID-19 crisis focusing on this use of AI.
AI is now being employed in a multitude of applications in relation to the pandemic, including vaccine development, patient diagnosis, and tracking real time transmission of COVID in the community.It is being used to predict how the virus may spread during lockdown easing and even predict wider behavioural patterns, such as panic buying.
Unsurprisingly, AI platforms are being widely applied to the identification of COVID-19 therapeutics. The novel coronavirus has demanded a re-examination of our therapeutic toolbox for drugs that may be repurposed to treat the infection. Indeed, the dearth of knowledge on COVID-19 has led many AI companies to utilise a new phenotypic approach whereby drugs modulating biological systems targeted by the virus are evaluated. This differs from typical screening which ordinarily identifies drugs on their ability to bind molecular targets specific to the virus.
One such example is the identification of the rheumatoid arthritis drug barcitinib by the Cambridge based AI company BenevolentAI. By utilising such an approach, BenevolentAI’s software identified that baricitinib's anti-inflammatory activity may be useful in countering the “cytokine storm” often seen in severe cases of COVID-19. Eli Lilly, who own the rights for barcitinib, have subsequently started a large phase II trial in COVID-19 patients.
AI in healthcare services
Furthermore, as COVID places a greater and greater burden on healthcare services around the globe, practitioners are turning to AI to assist rapid assessment and diagnosis of potential COVID patients. A team from the Huangpi People's Hospital in Wuhan, China - the city in which the pandemic is believed to have originated - have developed a deep learning model termed the COVID-19 detection neural network (COVNet).
Other than PCR tests, Chest CT scans appear to offer the most robust tool with which to identify COVID-19. The disease manifests with characteristic imaging features on the lungs. However, COVID-19 shares many traits with other community acquired pneumonias. The Wuhan based team trained their COVNet AI platform to identify CT scans of healthy patients and those with either COVID-19 or other community acquired pneumonias. The COVNet platform successfully diagnosed roughly 90% of CT scans from 450 patients, 20% of which had COVID-19. Thus, it seems deep learning models can be used to accurately detect COVID-19 and differentiate it from community acquired pneumonia and other lung diseases.
Outside the clinic, businesses and governments are also using AI to predict public behaviour in response to the virus. One behavioural phenomenon that has earned much attention during the pandemic is that of panic buying. The nationwide lockdowns established in response to the virus have impacted complex and interlinked supply chains. The American government is presently testing an AI system for predicting outbreaks of panic buying and how to adjust supply chains in response. The Joint Artificial Intelligence Center has developed a platform harnessing a number of data streams used to predict purchasing trends as well relevant logistics and supply chain issues. The platform draws upon data related to food and water supplies, but also for that of ventilators, masks and medical supplies. The predictive output is then used by the government to redirect supplies where needed.
The role of AI in the supply chain
Furthermore, AI is being used to predict how supply chains and logistic operations may change to mitigate the risk of future disruptive events. A recent IBM report highlighted how AI can play a role in helping design low risk supply chains of the future, capable of withstanding unprecedented disasters such as pandemics.
Thus, it seems AI platforms have served to resolve a multitude of problems that have evolved from the pandemic.
Yet, issues concerning the patentability of AI platforms and their outputs remain. AI inherently challenges the notions of ‘originality’, ‘inventiveness’ and ‘ownership’ that are intrinsic to attaining patent rights. Understandably, IP issues surrounding AI have taken a back seat during the current crisis, but it remains to be seen whether some technologies that may arise from the COVID pandemic bring these issues under renewed scrutiny.
For further information regarding intellectual property related to AI and its application in respect of COVID-19, please contact Charlotte Watkins at docketing@secerna.com