Artificial intelligence (AI) solutions have been transforming our daily lives for decades. They are in charge of personalizing the content we see on social networks and streaming platforms; improve web search engine results based on our history; they check the grammar of our emails and protect us from spam. Despite the fact that they often go unnoticed, they are also responsible for facilitating the detection of fraud in banking operations; monitor traffic in GPS applications or allow, among many other applications, voice assistants such as Siri or Alexa to understand and respond to user requests.
In this line, and during the last few weeks, we have seen how ChatGPT surprised and generated concern in equal parts by making clear its ability to maintain “natural” conversations with users and respond to their instructions quickly and accurately, from writing poems , essays or song lyrics, to generate complex lines of code, displaying a behavior that wants to emulate human intelligence and that on many occasions exceeds their abilities.
The health sector has not been immune to the expansion of artificial intelligence, which has contributed to facilitating the design of new drugs and reducing the times and costs of their production; reduce diagnostic errors or improve the prevention and treatment of the most frequent diseases, they affirm in the National Artificial Intelligence Strategy (ENIA) promoted by the Ministry of Economic Affairs and Digital Transformation of the Government.
Thanks to it, genome tests, early detection of cardiovascular diseases, diabetes and skin cancer have been improved; progress has been made in the diagnosis and treatment of mental and neurodegenerative diseases; it has contributed to the research of more personalized medicines and treatments. Precision surgery continues to evolve and monitoring of patients has improved with applications, biomarkers and software based on AI, it is added in the basic document of the Artificial Intelligence Strategy of Catalonia.
But the exponential growth of this technology has also raised, and still raises, difficulties and challenges, both technical, ethical, social and regulatory. On the one hand, those related to their security, privacy and access to data, an essential requirement for the development of AI applications, as well as the risk of generating inequalities due to the bias in the collection of this information, point out in the “Report C: artificial intelligence and health” of the Office of Science and Technology of the Congress of Deputies (Office C). The document also emphasizes the need for legislation, regulation, evaluation and human supervision to facilitate its implementation in professional practice and achieve trustworthy AI.
A trust that has not yet been achieved among the population as extracted from the “Study on the application of artificial intelligence” prepared by the National Observatory for Technology and Society (ONTSI), aimed at evaluating the trust of citizens in relation to the implementation of artificial intelligence in different areas of daily life. Specifically, the study shows that, although 41% of those surveyed are in favor of the application of AI in the medical field, since they see that its implementation would be something very or quite interesting in order to improve diagnoses or follow-up of patients, 26% are not very supportive of its execution.
In addition, 31% express a low level of confidence (value 0 to 4) in relation to the application of AI in the medical field, to which we must add that 33% position themselves at a medium level of confidence (value from 5 to 6), leaving only 12% of respondents who have a high confidence position.
It is undeniable that artificial intelligence is transforming society, but what exactly is it and how does it work? How does it contribute to improving the health sector? As explained by the Health/IA Program of the Generalitat de Catalunya, artificial intelligence is a computer science discipline dedicated to the development of algorithms that make intelligent decisions or behave, at least, as if they had intelligence similar to that of humans.
To achieve this, “systems capable of reasoning, learning from experience, solving problems, contrasting information and carrying out logical tasks in different areas of society” are created. In health, they add, these capacities cover fields as diverse as the automated interpretation of imaging tests, the monitoring and interpretation of patient records, the identification of interactions between drugs or epidemiological surveillance, among many others. In its achievement, automatic learning or machine learning comes into play, a subdiscipline of AI in which the system is capable of “learning” from experience from databases or physical sensors, and does not need instructions or programming to continue doing so. , allowing it to evolve over time as long as it receives new input data.
Within machine learning, the deep learning variant, or deep learning, goes one step further and is capable of quickly analyzing massive amounts of data, prioritizing the criteria necessary to make decisions, and performing complex tasks that may or may not require an minimal human supervision. This data processing technique, based on artificial neural networks with many layers and a function that is inspired by how neurons in the human brain do, is one of the most widely used in the healthcare field, for example, to analyze and detect anomalies. in medical tests such as x-rays or scans.
Although AI was born as an academic discipline in 1956, it has taken decades to have the volume of data necessary to take advantage of it, the computing power to apply it, and the research and development of algorithms and methods necessary for it to evolve. The exponential leap that it has experienced in recent years has allowed the development of artificial intelligence solutions associated with health to be introduced in practically all areas of medicine, from primary care to rare diseases, emergency medicine, biomedical research and public health.
This is detailed by the European Commission (EC) in its report “Artificial intelligence in health care: applications, risks and ethical and social impacts”. However, a survey collected in the Report of Office C of the Congress of Deputies points out that, currently, its development is mainly focused on diagnostic tools, which account for 21% of the total, followed by those for self-care , early prevention and monitoring (14%), or those that function as clinical decision support systems (18%). For their classification, different variables can be used, such as their potential recipients, the environment in which they are used or the practices to which they refer:
Artificial intelligence in clinical practice: it covers a multitude of areas, from the automation of image analysis (radiology, ophthalmology, dermatology, pathology, etc.), to signal processing (electrocardiogram, audiology or electroencephalography, among others). In the field of radiology, for example, this technology makes it possible to identify lesions, prioritize follow-up findings that require early attention, and enables radiologists to focus on the cases most likely to be abnormal.
In the area of ??pathology, some studies show that your analyzes can have a level of precision similar to that of pathologists; and in emergency medicine, they are able to improve the prioritization of patients during triage, as well as help in the analysis of different elements of their medical records. In surgery, artificial intelligence can help make surgical decisions and its use would also contribute to the advancement of radiomics, a field of precision or personalized medicine, they point out in the “White Paper on artificial intelligence applied to health” promoted by by the Center for Innovation of Data Tech and Artificial Intelligence (CIDAI).
Thanks to the development of AI radiomic techniques, they add, it could contribute to evolving screening programs that reduce certain pathologies, improve early diagnosis and predict the most appropriate treatment. In the same area, in “Report C: artificial intelligence and health: potential and challenges”, reference is made to a study that evaluated the use and doses of different treatments in which it was found that patient mortality was lower when the procedure used matched the recommendations of an AI-based wizard.
Another of the most promising applications of AI in clinical practice is its use in cardiology, where new cardiac imaging techniques allow specialists to perform faster evaluations of data extracted from cardiac images, such as cardiac ultrasound, cardiac computed tomography and cardiovascular magnetic resonance imaging. It also highlights its application in home care.
In fact, the 2020 Eurostat report estimated that in 2019 more than a fifth (20.3%) of the EU population was aged 65 or over, and that this proportion will increase two and a half times between 2019 and 2100, from 5.8% to 14.6%. Thanks to the contribution of AI in the home monitoring of this group, they affirm, independence can be promoted and aging at home improved by monitoring the physical space and falls. In particular, tools, software, smartphones, and mobile applications can enable patients to manage much of their own healthcare and facilitate their interactions with the healthcare system.
The CIDAI white paper also refers to this area, pointing out that the relationship between patient and health professional decreases after medical discharge, and that some relapses could be avoided by monitoring their health. This is where the potential of AI comes in, which would help to carry out this monitoring without involving a collapse of the health system, offering solutions that allow monitoring of some parameters.
Finally, in the area of ??mental health, AI applications can support patients through tools that digitally track depression and mood through keyboard interaction, speech, voice, facial recognition, sensors and the use of interactive chatbots. In this sense, the Office C report details that some studies have been able to predict the appearance of episodes of psychosis by analyzing language, with a laboratory reliability of up to 93%.
• Artificial intelligence in research: includes fields such as clinical research, drug discovery, clinical trials or personalized medicine. In the field of medicines, AI makes it possible to extract information from large databases to design new drugs and new solutions to improve their efficacy and safety, in addition to considerably reducing the costs associated with research.
Clinical trials also benefit from this technology, which can facilitate from their design, to the selection of patients based on certain parameters, to obtain samples large enough to be significant or to obtain more precise results.
• Artificial intelligence in public and global health: The potential of this technology in the field of public health ranges from the ability to help identify specific demographics or geographic locations where diseases or high-risk behaviors are prevalent, to contributing to digital epidemiological surveillance to, for example, build early warning systems for adverse drug events and air pollution.
They can also help, especially in low- and middle-income countries, to fill health workforce shortages by reducing their workload, as well as identify disease outbreaks earlier than they would through traditional methods.
• Artificial intelligence in administration: this is one of the sectors where AI can most clearly contribute to improving work dynamics, improving the time spent on administrative tasks and reducing associated errors, but it is also one of the most vulnerable from the point of view of hacking or lack of privacy. In this same area, he points out in the Office C report, artificial intelligence is also capable of composing medical discharges, labeling and generating radiology reports, enriching databases with clinical terminology or helping to optimize resources and medical personnel. in emergency situations or emergency services.
