Barcelona Metropolis

Artificial intelligence (AI) is revolutionising medicine and healthcare. There are already developments that can identify tumours more accurately than human experts. A complex surgery such as an aortic valve implantation is hard to understand today without robotic systems supporting physicians. And thanks to data and algorithms, the road to personalised medicine is opening up.

Since its inception, AI has supported medicine and healthcare. As early as 1972, there was already an intelligent tool that could identify the bacteria causing a blood infection[1].  These early tools were called expert systems, because they sought to replicate the knowledge of a human expert in a field of application – medicine, in this case – with the aim of making this expertise accessible to a wide audience, e.g. hospitals lacking a certain type of specialist. However, the clinical practice of medicine is the use of the best available scientific evidence, usually by means of clinical trials, whose results are published in scientific articles. This is known as evidence-based medicine. In this regard, the knowledge of a system must be constantly reviewed, and the systems that support decision making – now known as clinical decision support systems – have become more complex.

Indeed, the clinical milieu has changed significantly since the first expert systems, and clinical information methods now collect a wealth of patient health care records, as well as genetic information and information obtained through new devices, including wearables.[2]. Data availability has allowed this medical evidence to be obtained not only through prospective clinical trials, which involves testing individuals, but also testing hypotheses using retrospective trials and information stored in our health systems. In addition, machine-learning methods are applied to data from health records to identify, for example, which individuals are at high risk of cardiovascular disease. With this information, physicians can recommend actions to prevent possible health deterioration.

In addition, genetic data brings us closer to what is known as precision medicine, which allows us to design specific drugs for patients with special health conditions, in accordance with their genetics. In fact, the generation of new proteins that allow the synthesis of new drugs is one of the focal points of the use of AI, which mainly uses techniques known as deep learning.[3].

As far as new devices are concerned, nowadays we have complex monitoring systems, medical imaging, assisted robotics… If we picture an ICU, we will see patients connected to many devices that monitor their vital signs. This amount of information is difficult to control for the person in charge of the patient, who is helped in the process by AI systems, activating alerts if risk situations are detected. These monitoring systems are developed on the basis of the case histories of patients who have experienced similar scenarios, some of whom have recovered while others, sadly, have not. For instance, there are systems that allow early detection of sepsis, a serious medical condition that can rapidly lead to death[4].

It is worth noting that only 17% of the algorithms are aimed at improving management, while 60.53% are dedicated to clinical practice and surgery. In the latter area, the important role of robotics is noteworthy. Robots such as Da Vinci or Bitrack[10] are supported by medical imaging systems that show the physician the most relevant information in each case. A complex surgery such as transcatheter aortic valve implantation (TAVI) would be difficult to understand today without the help of AI systems that provide support to clinicians at all times[11].

All these systems are also propelled by a reality: we have a very ageing population. Medicine must provide solutions for people who reach an advanced age with chronic diseases (such as respiratory diseases or hypertension), with comorbidities (suffering from more than one disease) and with limited healthcare resources to care for them. However, due to advances in medicine, there are many people who reach old age with a state of health that allows them to live autonomously at home. In this context, smart homes are emerging, equipped with sensors that allow a person’s life to be monitored and can detect if the person falls or is unable to get up[12].

Robotics is also accompanying these elderly people by providing them with carers to look after them in the realm of routine tasks and allowing them to share leisure time with their family members, who, thanks to this assistance, will not be worn out. The film Robot and Frank, which illustrates this future that lies ahead of us, is a case in point.

The opportunities for caring for the elderly are highly inventive. For example, the company Sound Intelligence has developed software to monitor noises in a nursing home: if the observed sequence corresponds to normal behaviour (such as a bed movement, followed by the noise of a door opening, water falling, a door closing and another bed movement), the system does not disturb the potential caregiver; but if the sequence of noises is abnormal (a bed movement and a loud bang), an alarm is set off[13].  And the system has to work on a stormy day!

In conclusion, we can say that one of the drivers of AI in medicine and healthcare is the availability of data. However, there are major challenges. Firstly, in terms of data quality, the occurrence of bias and privacy. Generative AI – ChatGPT being one of the most prominent examples – can open doors to new drugs and ultimately generate new hypotheses and treatments, but it poses huge challenges regarding the body of data on which the results are generated, which can lead to significant biases[14].

Secondly, with regard to the transparency of the systems: the solutions provided by AI algorithms must be comprehensible to merit the trust of professionals. There is still a road ahead of us that needs to be tackled in terms of legislation, with the corresponding regulations for use.