Recent developments in Artificial Intelligence (AI) have fuelled the emergence of human-AI collaboration, a setting in which AI is considered a coequal partner. Especially in clinical decision-making this has the potential to improve treatment quality and reduce time and medical errors. However, while human-AI collaboration shows great promise, there are certain challenges to the widespread adoption of AI systems in healthcare, which has attracted increasing research attention recently.
Human-computer collaboration in healthcare is not a new concept. For instance, computer systems have been used to aid clinical decision-making for at least a couple of decades (e.g. CHESS, an expert system developed by Gustafson and colleagues in 1992). However, while the computer systems back then were rather simple and interpretable for the practitioners, the systems are now growing increasingly complex, and recent advances in AI have prompted human-AI collaboration in healthcare. Here, human experts work together with sophisticated AI systems to make decisions, complementing rather than replacing each other.
Although human-AI collaboration shows great promise it has still not been widely adopted in clinical decision-making and healthcare more broadly. Many factors have been cited as a reason for this, including the lack of trust, missing human-centred design, and opaqueness of the algorithms used.
Adopting AI systems in clinical settings has great potential. AI systems can support human beings with tasks that are cognitively demanding (such as analysing large volumes of records), help overworked practitioners, and improve the accuracy and efficiency of clinical decisions. The AI systems being developed can operate on many different data types, including language, medical images, and medical records, and thus, the systems can be deployed widely across healthcare. For instance, we already see how AI can help accurately diagnose cardiovascular diseases, detect skin cancer from images, and segment tumours in MRIs.
Consequently, understanding what hinders the adoption of AI systems, and how this can be addressed, is critical. This has recently attracted significant research interest in the artificial intelligence community, and researchers are studying concepts such as safety, reliability, and trust in connection to AI, paving the way for a greater use of AI in healthcare.
Healthcare is a highly regulated space, and developing AI technologies to meet regulatory requirements is a challenge. AI technologies have to be approved by FDA to be deployed in a clinical setting in the US, and depending on the computational approach used, this can be a long process.
Big corporations like Google and Amazon are also moving into healthcare, a multi-trillion dollar industry, eying how modern computational approaches, including AI, can be used to modernize the industry. Moreover, hospitals are also increasingly looking into the deployment of AI, with hospitals like Mass General Hospital having their own dedicated research groups.
A lot of research is being done on 1) AI applied to medical problems and 2) how to make AI technologies safer, trustworthy, and human-centred. This research is spread across the US, with strong research centres at e.g. Harvard Medical School, MIT, and UC Berkeley.
There are many AI start-ups across the world moving into healthcare, providing technologies ranging from wearables, natural language processing (NLP) technologies for medical records, to tools for clinical decision-making. Alone in the US there is more than 500 companies US raising money.
AI in healthcare is expected to be a massive market going forward, with a projected growth from US$ 6.9 billion in 2021 to US$ 67.4 billion by 2027. In 2022, more than 500 companies in the US have raised funding for AI in healthcare, collectively raising more than US$28.9B in pre-IPO funding.
Please reach out to Science attaché Torben Orla Nielsen at firstname.lastname@example.org for any inquiries. We offer our services to corporates, SMEs and academic partners looking to dive further into the area of human-AI collaboration and human-centred AI.