29-06-2022 | | By Robin Mitchell
In another demonstration of the power of AI, researchers have recently reported detecting COVID at early stages before symptoms show up just by analysing data from wrist-worn health trackers. What challenges does COVID present, what did the researchers demonstrate, and how will health monitoring benefit from AI?
Now that the pandemic is mostly over, COVID appears to be like any other cold virus that is virtually impossible to eliminate and will play a role in everyday life. As most people on the planet have been infected and/or vaccinated, there is no longer a need for COVID restrictions or the use of PPE, as those that have survived now have antibodies.
But when COVID first came about, it presented a serious threat to the world for numerous reasons. The first was that, unlike the common cold and flu viruses, COVID was a unique virus in that it had never previously interacted with the general population. This means no natural protection methods in the form of antibodies were available, and this meant that anyone who caught COVID had to effectively create their own vaccine via an immune response (and this immune response could be delayed as the body tries to identify the virus).
Secondly, in the act of trying to respond to the virus, COVID has a nasty habit of causing fluid to build up in the lungs. This allows for bacteria to further cause a secondary infection resulting in pneumonia, and it is this pneumonia that often leads to death.
Thirdly, and probably the most unfair, is that COVID is asymptomatic for a large percentage of the population, allowing it to spread extremely fast. Furthermore, COVID won’t even show symptoms in those who are vulnerable to it for a long time (up to a week possible). Thus, COVID is virtually impossible to track and stop without using robust PPE methods, isolation, and frequent surfaces, hands, and clothes disinfection.
Even though the COVID pandemic is now mostly over, it is still possible for new aerosol viruses similar to COVID to spread worldwide, creating new pandemics. While lessons have been learned concerning highly infectious viruses, preventing new pandemics will still be difficult.
In order to be able to prevent the spread of future viruses like COVID, it will be essential to identify those who are carriers and isolate them to prevent further spreading. The use of PCR tests and antigen tests can be used for confirming cases, but as COVID demonstrated, trying to test an entire population on a routine basis is challenging. Furthermore, only testing those with symptoms doesn’t identify those infected but have no visible symptoms.
Recognising the challenges presented by COVID, researchers have recently been devising methods for detecting COVID early, and one team may have succeeded with the use of off-the-shelf health trackers. While COVID symptoms may not be obvious, the body will still experience small changes that can be detected by wrist-worn health trackers with examples of minute changes, including heart rate, sweat, and blood pressure.
Of course, changes in one of these readings cannot automatically indicate COVID and trying to use traditional hard-coded if statements to determine if someone has COVID or not simply doesn’t work. As such, the researchers turned to AI to detect symptoms as AI can train itself to look for patterns that may not be obvious.
To train their AI, the researchers gathered over 1.5 million hours of data from 1,163 people under the age of 51 using Ava Bracelets that operate at night. These bracelets connect to a smartphone that also tracks the use of prescriptions, activity, and consumption of alcohol. Finally, all participants were regularly subjected to PCR swab tests to identify those with a COVID infection. The result of this trial showed that the AI could identify 68% of individuals who had COVID 2 days before symptoms showed.
Now, the AI is being tested on a significantly larger sample of 20,000 people across the Netherlands to identify the accuracy of the AI and see whether it continues to operate after initial infections (as secondary infections may present weaker symptoms and thus be harder to identify).
The use of AI in medical diagnoses clearly presents significant advantages, including the ability to control infections, identify diseases early, and help doctors provide the right treatment. But above all else, AI in medical science will most likely be advantageous in allowing people to self-diagnose and provide self-care.
A good example of where this would be highly advantageous is skin cancer and moles. It is highly likely that the vast majority of the population cannot identify potentially dangerous moles, and it is even more likely that people don’t track the size of their moles or their locations. Instead, people often rely on someone else spotting a mole or a doctor having to do a spotting a mole, or a doctor having to do a physical examination of their body, and anyone who uses the NHS will know that the NHS never does anything without being hassled.
But, imagine a machine similar to a sunbed that could photograph every inch of skin on the body, and this data is then fed into an AI that maps all moles, tracks their size, and compares them to known malignant moles? Such a system could spot problems in their early days and turn a 3-year fight with cancer into a 20-minute session with a nurse and scalpel to remove the problematic mole.
The best part of an AI doctor? Its operation would be virtually free, and it would only get better with time. Readings taken from each AI system can be used to further train the AI to recognise symptoms, and the use of software means that the only real cost is the electricity needed to operate the AI. Of course, additional fees would come in the form of machine hire, processing time, and licensing, but considering that researchers have already demonstrated working AI diagnostic systems, it is likely that such a system would be very cheap to operate.
Overall, AI is an extremely powerful tool that could revolutionise the medical field, and the use of AI diagnostics could cut out the middleman (i.e., GPs) and help patients connect directly to specialists.