AI in Healthcare
AI is increasingly making inroads in healthcare applications, although the field is still in its initial applications. AI can assist doctors and medical practitioners in various ways. One of the most effective ways that AI can assist is to act as a tool that makes better diagnosis and suggest potential treatments. For example, medical imaging AI has been proven to help in making better diagnosis and analysing the imagery better tirelessly, day in day out. AI can also analyse blood results and other lab results, cross-check them and suggest typical tests while flagging results that may be abnormal or that show patterns of diseases or conditions that have happened or about to happen. Currently AI is currently capable of giving answers without much in the way of an explanation or justification of how it got to those answers. AI needs to improve its explanation capabilities for it to integrate more effectively into the standard toolkit available to medical practitioners, giving an enormous advantage especially in locations when deep medical expertise is scarce or simply unavailable. In keeping with our philosophy that AI should assist people rather than replace them, AI can help people obtain quality medical advice in circumstances that nowadays would be impossible to obtain.
AI is also helping at the forefront in finding typical novel drug discovery methods. There has been a resurgence of interest into what AI can do in helping find drug solutions, for example, during the pandemic outbreaks of the COVID-19 coronavirus, which presented a novel disease that was being discovered incrementally. The incremental discovery and changing nature of the COVID-19 pandemic highlighted one of the main drawbacks of current Deep Learning based AI systems. Each case that is discovered is essentially a new datapoint for the AI system to learn from. In the beginning, there will always be just a few datapoints from which to learn from — but this is the time when rapid discovery can make most impact, before large amounts of people get infected. Deep Learning, and its inability to adapt quickly to rapidly evolving scenarios (called a changed or shift in distribution of the underlying data) will need to be augmented or replaced with different systems to overcome this problem.
What typically happens in such cases is that AI is used together with genome sequencing techniques — that themselves use a lot of machine learning techniques — to look at the structure of the proteins that are on these viruses and tries to find a better custom-made drug that will target specific viruses or infected cells and try kill or inactive them. The internal structure of the disease, which is encoded using RNA or DNA, depending on whether it is a virus or bacteria, is analysed to find a “signature” that allows for its identification with different test techniques. The analysis of the disease pathogen structure gives clues as to how to identify and disable the disease by latching onto it and disabling it or otherwise killing it.
Drug discovery is one major way how AI can help healthcare. The long-sought promise of having personalised drugs tailor made to a person’s genetic makeup and specific conditions is still far from being realised, but I firmly believe that this represents the future of medicinal drugs.
The healthcare of the future is also going to be involved more into continuous monitoring of people. Wearables will become more advanced and become better at monitoring the relevant factors that determine the state of your health together with any risk factors specific to an individual, while raising alerts and helping doctors in making better diagnoses. AI powered wearables and sensors can also detect abnormalities pro-actively and flag up cases where you should go for a check-up by a human doctor, or maybe in the future, immediately bring up a consultation with an AI doctor!
Another area where AI is already helping is in surgery. Robotic surgeons and surgery assistance tools are already in limited use for particular types of surgery and their capabilities will increase over time as AI becomes ever more capable of dealing with the innately imprecise and varied anatomy of people. Robots have always had difficulties in dealing with soft tissues and also the fact that people generally have quite a lot of variance in their internal anatomy, which is something that human surgeons take in their stride but that may confuse robots significantly. Certain interventions, like laparoscopy and keyhole surgery, AI will be of immense assistance as the constrained nature of such surgeries can allow AI better use of its current strengths and capabilities.
Wearables will also be able to enable AI to make better assessment of the effectiveness of daily activities such as walking and also more intense activities such as athletic and sports activities, enabling better human performance optimisation and tailored advice on how to improve training loads and physical activity in general.
I personally think that in the healthcare of the future, everyone will have their own personalised AI doctor that is constantly monitoring you, constantly making sure that you are functioning at your best performance, and flagging up issues that you should discuss further, much before traditional medical practices flag up such concerns. On the other hand, the good application of AI will cut down on false alarms that may be caused by broad matching of symptoms, giving a more surgically precise application of results. I do not believe that AI in health care is going to replace people. Rather, I think that AI should be used as a tool that will assist people in giving them better medical advice and health service at a cheaper cost and with wider availability. I think that by having the highest level of medical expertise available to you in your smartphone or in remote places, everyone will have a better quality of life as a result.
AI can also help manage hospitals, drug factories and other medical related places more efficiently and reduce waste. I think the waste reduction and the potential to reduce cost while making things more efficient will impact all people’s lives very positively. The monitoring and statistical gathering aspects also make sure that issues like diseases are tracked and reported very efficiently and uniformly, ensuring that there are proper statistical analysis and medical models while promoting the safe and private sharing of anonymised data.
One of the ways that AI can also help in treatment discovery, is to shorten the amount of time leading from a new outbreak, for example from the detection of a new disease to discovery of a treatment. One of the ways that this can be done is to make sure that there are proper records of the different types of treatments and different types of substances that are being given and tested on people, especially in an emergency when you do not have enough time to conduct a proper randomly controlled trial, an RCT — which is the gold standard in drug discovery.
By aggregating information from different sites around the world, and applying mathematical analysis to the data together with the appropriate simulated modifications to them, you can effectively have information that is very close to a controlled trial and be able to find effective treatments much more quickly than traditional methods. Obviously, clinical trials always remain the gold standard, and these are indeed the best way to ensure public health safety, but in times when there are emergencies — these methods can lead to lives being saved and faster improvement through knowledge sharing and more optimal protocols being distributed to medical practitioners across multiple sites and countries.
AI can also help in patient healthcare, including areas that are not completely related to treatments and so on. For example, with dementia patients and very old people in general, AI can help keep better records of what has happened — better photos and memories that are used to juggle someone’s memory and hopefully make them remember and cherish those memories, help their memory deteriorate at a slower pace, and even help them get a better quality of life.
Overall, I think AI will be pervasive in the future of healthcare and medicine, leading to a better quality of life, and better, healthier outcomes for all of us.
