The combination of artificial intelligence and machine learning technologies promises a future of more reliable and personalised diagnoses to more effectively treat diseases such as cancer, Alzheimer's and stroke
14 July 2022
An X-ray every two years for all women between the ages of 50 and 69. Since 1990, this has been the National Health System's biggest screening challenge, and its aim is to prevent one of the most common cancers in Spain, breast cancer. The method consists of X-rays that detect potentially cancerous areas; if anything suspicious is observed, this test is followed by more tests, often with a high chance of false positives, which are harmful and costly. Their tortuousness is the main reason why screening is limited to the highest risk groups. By adding predictive algorithms to mammograms, the risk areas of the patient's breasts would be narrowed down and the reliability of diagnoses would increase to 90%. Therefore, they could be done more frequently and the age range of targeted women could be widened.
It is a process that already exists, which uses artificial intelligence, and is being developed by a team at the Spanish National Research Council (CSIC), specifically at the Institute of Corpuscular Physics (IFIC). It forms part of the field of machine learning in precision medicine, and of a network of research that seeks to increase the efficiency with which each patient is treated and to optimise healthcare resources.
To understand how, one must first understand the concepts that come into play. The first is artificial intelligence. "The ability of a computer or robot to perform tasks generally associated with intelligent beings", according to the definition of Sara Degli-Esposti and Carles Sierra, authors of the CSIC white paper on this discipline. In other words, they are the procedures used to replace the work of human beings with that of robots, with the aim of having them carry it out with greater precision and efficiency.
And where can artificial intelligence work in medicine today? On several fronts," says Dolores del Castillo, a researcher at the CSIC's Centre for Automatics and Robotics, "from the administrative to the management of clinical documentation. And, in a more specialised way, in image analysis, or in patient monitoring and follow-up". And where are there still enormous limitations? Above all, "in the field of healthcare, in the legal and ethical aspects of dealing with critical cases". Del Castillo, who is working on projects on neurological movement disorders among others, points out that there is still a lack of training for a large part of the healthcare staff.
With its advantages and shortcomings, the second concept, machine learning, is a subfield of artificial intelligence. This could be translated as machine learning. That is, artificial intelligence that works through computers that detect patterns in population groups. These patterns are used to make predictions about what is most likely to happen. Machine learning translates data into algorithms.
Precision medicine adapts to the individual profile of each patient. / ISTOCK
And after artificial intelligence and machine learning, there is a third concept: precision medicine. That which adapts to the individual, to their genes, to their background, to their lifestyle, to their socialisation. A model that must first be able to anticipate the disease; secondly, according to Francisco Albiol of IFIC, to "assess each patient, apply the best therapies based on clinical evidence, discern the most complex cases and assess their inclusion in management programmes". It makes sense for high-impact diseases, and does not make sense for serious diseases; for example, to distinguish in primary care a flu from constipation, because the benefits would not compensate for the effort required.
The key to the use of artificial intelligence in medicine is also cost optimisation, which is very important in public health. The Spanish population has increased from 42 to 47 million people between 2003 and 2022, i.e. by more than 10%. And from 2005 to 2022, the average age of the population has risen from 40 to 44. We are getting older and older. For this reason, "the most highly rated projects, and therefore eligible for funding, are those that incorporate artificial intelligence technologies to address the prevention, diagnosis and treatment of cardiovascular diseases, neurodegenerative diseases, cancer and obesity," says Dolores del Castillo, but "special attention is also paid to proposals on personalised and home medicine, care for the elderly and new medicines". "The need for healthcare has increased because of our demographics, and the aim has to be to reduce and simplify the challenges with technology. We are trying to do this with machine learning," summarises Albiol.
Albiol is one of the scientists who has led the programme to improve breast cancer detection through algorithms. He argues, like other researchers, that if we mix machine learning with precision medicine, what we should be talking about is 4p medicine. The one that conglomerates four characteristics: "Predictive, personalised, preventive and participatory". Because the purists limit precision medicine to the field of patient genetics, and would not include in the bag that which takes into account more characteristics. Those who do argue that we are talking about something much broader: "Applied to precision medicine, machine learning allows us to analyse large amounts of data of very different types (genomic, biochemical, social, medical imaging...) and model them together to be able to offer diagnoses adapted to the individual, more precise and thus carry out more effective treatments," summarises Lara Lloret Iglesias, a researcher at the Institute of Physics of Cantabria (IFCA, CSIC-UC).
Lloret is part of a network of scientists who, like Albiol and Del Castillo, are dedicated to projects on machine learning and precision medicine. One of the projects developed by his team, which he leads together with the physicist Miriam Cobo Cano, is called Branyas. It is in honour of María Branyas, the longest-lived woman in Spain who managed to overcome the covid-19: she did so at the age of 113. It brings together the casuistry of more than 3,000 elderly people, not only genetic: "Machine learning techniques are used to establish risk profiles for becoming ill or dying as a result of the coronavirus. We have obtained data from the analysis of three risk profiles: a socio-demographic one, a biological one and an extended biological one, which will add information on issues such as intestinal microbiota, vaccination and aspects related to immunity".
Josep Lluis Arcos, from the Artificial Intelligence Research Institute, also explains this. The diseases usually associated with precision medicine are cancer and Alzheimer's, but they have stood out with the Ictus project. Launched in the middle of the pandemic (which has made things difficult, he acknowledges), it has dealt with patients at the Bellvitge hospital in Barcelona who suffered a stroke and, after the critical and acute phase, have become chronic. Specifically, with those who have movement difficulties in one or both arms. They have conducted more than 700 sessions in which they asked patients to play the keyboard of an electronic piano. They then translated the analysis of the finger movements into results to see what the patterns of difficulties, and improvements, were. And they have had a particularly positive response from users "because it's not just an exercise, but it affects a very emotional part". The aim now is to extend it to hospitals in the UK.
What about the future? "I think the challenge of artificial intelligence in medicine is to incorporate the results of research in a generalised way into daily practice," says Dolores Del Castillo, but always without forgetting that "it is the experts who have the last word". To this end, "it is necessary that doctors trust these systems and can interact with them in the most natural and simple way possible, even helping in their design". Lara Lloret believes that we will have to be able to create "generalisable prediction systems, that is to say, that the efficiency of the model does not depend on superfluous things such as which machine the data was taken from, or how the calibration is". Francisco Albiol focuses on a problem that "will have to be solved in the long run": for the moment, "large hospitals are favoured in these techniques compared to small cities or towns. Facilitating and reducing costs also has to do with reaching everyone.