Artificial intelligence (AI) is the use of technology and machines to work and react in place of humans, conducting functions that were previously thought to require human reasoning and problem solving skills. That is the ideal definition at least. However, to this day, most AI applications have been only successfully programmed to carry out specific tasks or solve pre-defined problems. AI is not something new, but there have been significant advances made in the field these past few years. It is believed that AI could be the answer in helping to combat important health challenges, such as how to meet the healthcare needs of an ageing population. We are seeing an emergence of world-renowned technology companies, including Google, Microsoft and IBM that are investing in AI research and development for the future of healthcare.
Applications of AI in Healthcare and Research
AI could potentially be used for planning and management of resource allocation in health and social care services. Essentially, it would match patients with healthcare providers and design personalized care plans that meet their needs according to their allocated care budget. AI is currently being used in some hospitals around the world in order to improve patient experience and satisfaction. For example, the Alder Hey Children’s Hospital in Liverpool, UK worked with IBM Watson to create a ‘cognitive hospital’, in which an app was created to facilitate interactions with patients. The main aims of the app are to provide patients with a medium to declare their symptoms and chief complaints prior to a visit, provide information on demand, and provide health practitioners with the necessary information to help deliver the most appropriate and effective treatments.
A sector of the healthcare system which is currently slowly transitioning to AI is the medical research community. Healthcare data is complicated. The benefit of AI in this case is that it can be used to analyze large pools of complex information and identify patterns within datasets faster and more accurately than possible by a human. It can also be used to search scientific journal databases for relevant studies pertaining to the research topic of interest in order to combine the data and aid discovery. For example, the Institute of Cancer Research uses AI to predict cancer drug targets by using their canSAR database to combine the patient’s genetic and clinical data with information from scientific research. Researchers have also developed an AI ‘robot scientist’ named Eve in order to make the rigorous process of drug development faster and more affordable. AI systems could also be helpful in medical research by matching patients to appropriate clinical studies.
Using AI to analyze clinical data, research publications, and professional guidelines; it has the capability to assist in the diagnosis of disease and to formulate personalized treatment plans for patients. The fields in which AI will produce the greatest waves of impact include: medical imaging, echocardiography, screening for neurological conditions, and surgery.
- Medical imaging – stored collections of medical imaging and scans are used to train AI systems to detect conditions such as pneumonia, breast and skin cancers, and eye diseases. The benefits of such practice include not just the reduction of time and costs in analyzing scans, but also the increase of accuracy in diagnoses.
- Echocardiography – detection of irregularity patterns in heartbeats, such as in coronary heart disease, can also be administered by AI systems.
- Screening for neurological conditions – speech patterns are analyzed and processed by AI systems to predict the onset of psychotic episodes and monitor signs of neurological conditions, such as Parkinson’s disease.
- Surgery – robotic tools controlled by AI are increasingly assisting microsurgical procedures to help reduce surgical mishaps and malpractices.
Infectious disease outbreaks and sources of epidemics are major global health concerns and national security threats that know no borders. AI could potentially aid in their detection, isolation, and help achieve disease eradication goals.
The Limitations of AI
AI is only as good as the quality of data it is trained with. In other words, inconsistencies in the availability of data could hinder the learning process and restrict its potential. Also, keep in mind that with the analysis of large and complex datasets comes a significant amount of computing power that is required. An expected challenge in training AI systems is the fact that many healthcare systems around the world do not consistently digitize their medical records. Within these healthcare systems, there is also a lack of standardization in IT systems, digital record keeping, and data labelling.
Another concern is more on the human level. Humans have certain attributes, such as compassion, that cannot be learned by a machine. There are certain traits used in clinical practice by physicians- like reading between the lines and reading social cues, that are beyond the current comprehension level and replicative ability of AI. This presents the debate: are there some human traits that cannot be taught, and is there an algorithm for tacit knowledge that is normally instinctively understood by humans? Another claim is the question of autonomy: will AI be held on the same standard as an autonomous human, in which by definition cannot be held by a machine?
Ethical and Social Issues
Many of the ethical issues and concerns presented pertaining to AI mirror those raised about data usage, automation, and the overall dependence on technology on a broader scale in our daily lives.
Data Privacy and Security
While AI has the potential to pave the way for many good things in the near future, it can only do the work according to the intentions of the programmer; meaning, if the programmer had any malicious intent while programming, then AI could also be used to fulfill those malicious purposes. For example, there is fear that AI could be used to monitor behavior and detect patterns by tracking people’s biomedical sensors, such as smartwatches, activity trackers, etc. which could reveal information about a person’s health that can then be sold to health insurance companies.
Reliability and Safety
Reliability and safety are important issues to be discussed because AI is not only used to control the equipment, but is also used to make decisions on the necessary treatment plans for the patient. AI – just like any other machine – is only as good as the creator, but in this case, the trainer. AI too can make errors, and if the error is buried in a pile of endless data, it can easily go undetected and have serious implications. AI applications in healthcare also use data that is sensitive and private. In other words, if an AI system were to be hacked in order to obtain sensitive patient data, this could happen without being detected.
Transparency and Accountability
It is difficult to hold a machine accountable, let alone AI – a type of machine learning technology. There is a gray area when dealing with the transparency and accountability of machine learning technologies because they have the ability to constantly change their own parameters and rules as they learn. This creates the problem of validating the results and catching any errors or biases in the data obtained from AI systems.
Effects on Patients and Healthcare Professionals
AI systems that are used to help support people with chronic health conditions or disabilities have been found to positively impact patients by giving them a sense of dignity and independence. This is because AI health apps, for example, give the patient the ability to evaluate their own symptoms and learn ways to take of themselves from home without having to be admitted to a care institution for long periods of time. However, AI systems can also have negative impacts on patients. For example, there are concerns that if AI technologies are used to replace physicians and healthcare professionals, then there would be a loss of human contact – an essential element of physician-patient relationships. The patient’s freedom to make informed decisions for themselves about their health could be restricted if physicians are unable to explain to the patient how the AI system arrived to the diagnosis or the treatment plan.
Introducing AI systems into healthcare may cause healthcare practitioners to feel threatened by this new form of technology, especially if their expertise, autonomy or authority is challenged. The automation of tedious tasks, such as paperwork and computerizing data, could free up time for health professionals to spend engaging more directly with their patients. But this raises the issue that AI systems could be used as an excuse to employ less qualified staff since less expertise is needed. This is problematic because if the technology were to fail, the staff would not be able to recognize any errors. There is also the problem of complacency, in which healthcare practitioners would rely too much on the AI results and not challenge the results or check for errors.
AI technology is being trialed and used around the world in healthcare and research for many difference purposes ranging from the detection of diseases and management of chronic conditions to the delivery of healthcare services and drug discovery. AI has the potential to create new solutions for important health challenges, but unfortunately does not possess the ability to express key human characteristics and is limited to the quality of health data available. The use of AI has also raised some ethical and social concerns; such as, the ambiguity in transparency and accountability of machine learning technologies and the extent of data privacy, reliability, security and safety. We need to ensure that AI is trained and used in a way that is transparent and compatible with the interest of the public but also not overly regulated to the point of restricting further innovation in the sector. Ultimately, it is up to politicians, policy makers and most importantly, the general public, to decide the future course of artificial intelligence in our healthcare systems.