Present and future of AI in health - the interactive doctor

Talking about the medicine of the future means talking about technology, since inevitably artificial medicine and intelligence (AI) will go hand in hand from the moment we talk about the digitalization of processes.The complexity and increase of data used in medical assistance make this technology many and many applications, especially in clinical decision making.

But not only, since we have examples of different types of employees to make diagnoses, establish treatments or improve the participation and adhesion of patients who have shown that they can function equally or better than with human intervention with.In this scenario, the current pandemic has meant an impulse in the development of AI while digital health is the one that is helping to build post-covid reality.However, its application is not exempt from difficulties that have to do with ethics in their use and safety in their functionality.

The beginnings of artificial intelligence date back to the 50s, when Alan Turing, considered the father of artificial intelligence, wrote about ‘Computing Machinery and Intelligence’ in Mind magazine.Since then, its use in health sciences has been the most fruitful.The pioneering system in this use was the Mycin, from Stanford University (EE.UU.), developed for the diagnosis and treatment of bacterial diseases.From there it has evolved having ahead of the potential to transform many aspects of patient care.

To explain some of the applications that are already working in the field of health, it must.

Smart machines or Machine Learning

The automation of processes through the use of robots to perform human tasks is one of the most developed health fields.For this, not only AI is used, but one step further with the Machine Learning or automatic learning is going.

It is based on the use of machines capable of learning from the experience of its own functioning for the development and optimization of the tasks assigned to it.And they do so and finding patterns in large data sets independently that allow improving their prediction precision.An innovation that can lead to more effective and holistic care strategies that improve health results in patients.

Deep learning

Deep or Deep Learning learning is one of the most common forms of AI and its most frequent health use focuses on precision medicine, since it predicts which treatment protocols may be more successful in a patient based on their circumstances andThe therapies available.It is a very used technology in radiology and oncology to detect potentially carcinogenic lesions.

The processing speed and cloud infrastructure allow automatic learning applications to detect anomalies in images, beyond what the human eye can see how experienced it is, which helps to diagnose and treat diseases.It is estimated that any AI system, in order to get predictions, must examine, at the outset, at least two million data.

It is a model that has been used to diagnose more than twenty skin conditions, detection of diabetic retinopathies, as well as metastatic breast tumors;predict whether prostate cancer will be aggressive or if one is evil in the lung.In the opinion of Andrés Visus, professor of ESIS, we are facing the most important activity of AI, since it identifies, diagnoses and predicts with an accuracy much greater than the human.

Devices that listen as doctors

Other technologies in which large companies such as Amazon or Google have made large investments are in the field of natural language processing or NLP).Includes applications in voice recognition, translation and text analysis;and other objectives related to language with two basic approaches: statistical and semantic nlp.

In health, this technology is used in the creation, understanding and classification of clinical documentation (notes, reports ...) and published research.Also to transcribe patient interactions through voice assistants such as Alexa.You can already find these devices in medical centers to take note of patient requests and locate them in the corresponding specialty or for other administrative tasks.

Virtual Health Reality

Among other new technologies that are also being applied to health is virtual reality.It emerged at the beginning of the 90s, with the appearance of video games, and in Medicine it already has numerous projects.According to a Goldman Sachs report, health virtual reality projects will invoice the 5.000 million euros in the next five years.

From the Association of Researchers in Esalud (AIES) they recognize that it is a technology used in many hospitals for the training and training of health professionals in matters of prevention and carrying out different procedures from simulation systems.Other applications focus on use with patients who have post -traumatic stress, for pain control or even the diagnosis of pathologies such as care deficit and hyperactivity disorder (ADHD).

Despite being a widely used technique and offers an infinity of applications, from AIES they recognize that it is necessary to carry out more studies that validate their benefits in the health field.

In the Oncological Clinical Practice

Presente y futuro de la IA en salud - El médico interactivo

During the practical application of these technologies in the different medical disciplines, the Vall d'Hebron Radio Groupto treatment.In this way, they manage to identify which patients with tumors will be responders to immunotherapy.In the words of the principal researcher, Raquel Pérez-López, this technology allows to extract information "infinitely greater than that obtained only with the observation of a specialist".

For genomics in the field of precision oncology, this technology is being very useful, since each tumor is unique and does not always fit a common protocol.With ia techniques, totally personalized treatments and diagnoses can be developed.

The massive sequencing processes in this specialty have the advantage of being able to develop more effective and less toxic personalized therapies.The problem arises when analyzing the large amount of outgoing and variable data, an area in which both Machine Learning and Deep Learning are able to automate their processing and draw conclusions from their analysis, using computer models and artificial neural networks.This is summarized.

Mass sequencing platforms and algorithms designed to detect and interpret tumor variants are designed to choose the drug and the most indicated treatment in each case.However, for this, the predictive tools must have large amounts of data from patients with clinical follow -up that allow the training of models.

Alzheimer or Parkinson's detection

In that same sense, in the field of neurology, there are many AI projects that are oriented to the diagnosis of diseases.Know what people can develop certain neurological diseases through the use of automatic learning that analyzes the images obtained by magnetic resonance.Or, models based on the use of biomarkers that act as indicators of the disease that are used in the diagnosis of mild cognitive deterioration (DCL) or Alzheimer's disease.

In the case of Parkinson, although it is a more complex ailment to diagnose, especially in the early stages, there are programs that use AI oriented to the analysis of voice or pituitary alterations.There are even classification algorithms designed to positively discriminate against people with this pathology against others with similar symptoms, such as progressive supranuclear paralysis (PSP).

Prevent the recurrence of the stroke

At the Vall d’Hebron University, recently, they are working on the use of AI to identify the most important ICTUS factors in order to prevent the recurrence of strokes at three and 12 months at the individual level.Knowing this risk has great clinical value for doctors and patients, since they can focus on prevention of it and the rapid response to the disease.

It is also working on automatic learning models in the management of epilepsy, amyotrophic lateral sclerosis (ELA) or headaches.And, as in other fields of health, this technology looks towards personalized medicine with projects that use large databases and Machine Learning to generate a mathematical model that allows predicting the therapies that will best work in a given patient.

Applications in other specialties

But AI is not exclusive to neurology or oncology, but verified algorithms have also been developed for use in dermatology and gastroenterology.And there are teams in the cardiovascular area working in the discrimination of individual myocardiopathies, in the early diagnosis of atrial fibrillation, heart failures and valvulopathies.In fact, Machine Learning can be essential in the prognostic evaluation of cardiovascular diseases and determine the risk of new events with a lot of precision.

In the field of assisted reproduction, the first steps have been taken to create algorithms that allowed to determine, through previously defined objective criteria, what is the best embryo to transfer.

Harmonization and use of data

In all these projects, it is necessary to think about the use of large personal medical databases and the need for anonymization, harmonization and computational platforms agreed to process them, the Big Data.Dr. Xosé Ramón García Bustelo proposes, in the aforementioned monograph, the use of cloud and supercomputing systems that allow the safe and processing storage of information efficiently.

However, this that a priori may seem simple has serious complications in practice, since formulas that allow information between researchers and companies will have to be established.In addition, regulation and regulatory change will be necessary to access these data that may contain genetic references to the identity of patients.

Ethical problems

In this sense, the implementation of this technology must also face the resolution of the ethical problems that it can pose.Fernando Abellán-García, Director of Advisors Health Law, emphasizes that the rights and values of citizens must be respected.

The current context of the pandemic has triggered the use of videoconsultation services, "which makes it necessary to believe new action protocols that take into account respect for personal dignity," he says.Likewise, work systems and processes must guarantee human supervision and control, allowing people to adopt certain vital decisions on their own and not depend on machines, the expert alleges.

Abellán-García concludes that the regulations on damage due to defective products and civil liability regulation are insufficient to guarantee adequate coverage of responsibility for damages caused to patients following healthcare through AI systems."In this sense, as the European Commission points out, the European and national legislator must face modifications to contemplate this new reality".

For his part, Federico de Montalvo, president of the Bioethics Committee of Spain, has stressed that we are legally prepared to face the risks that new technologies can raise, such as, for example, cybersecurity.And indicates a governance framework that involves the industry, public authorities and patients ”.

Implantation complexity

Regarding its implementation in the National Health System (SNS), in the opinion of Dr. Jaime del Barrio, president of the Digital Health Association (ASD), the shortage of human resources specialized in the use of these technologies and the technical risks forIts implementation in current medicine are the main barriers facing AI in its implementation.

"The solutions based on AI - explains - play a crucial role in the organization of modern health systems and expose their maximum potential in times of crisis", so an infrastructure is needed that connects all parts of the system, facilitates theinteroperability and allow the flow of clinical data, ensures.

From the neighborhood, it should not forget that incorporating AI to health care involves introducing new applications that affect patient care processes and that they should not affect the service.They can improve, optimize, integrate orautomatize, but not negatively affect or interrupt the service.

“A wide number of AI solutions is given by the coding of existing guides and clinical protocols through a previously established system.From there, we can start a path from which they can advance the models after the time they learn from the data they get, ”he says.

Functionality security

The truth is that AI is increasingly present in our lives and, as its use increases, we will depend more on it, so a good management of this technology is essential.Its functionality and safety are essential aspects to take into account, especially in the health field.

Incorrect use can lead to the opposite for what has been designed, an idea that scares only thinking about it.To give an example, an artificial vision algorithm could be designed that processes data to know its own operation.It is what is called ‘antagonistic sample’ and, in addition, allows it to be manipulable.From there, it would be easy to create erroneous patterns to dominate AI in order to detect something that does not exist or do something that should not.

For Amparo Alonso Betanzos, president of the Spanish Association for Artificial Intelligence (AEPIA), the first thing we should do is question what type of intelligent system we want in Medicine.Responding to your own question, he points out that "it must empower clinicians, allow to increase the speed and quality of care and reduce costs".Without forgetting that this technology would have to serve to increase the quality time that clinicians dedicate to the care of their patients.

To do this, a technological change in hospitals is needed, in addition to a cultural and social change that ensures a positive implementation of technology in professionals and patients.But it is also necessary to advance in issues of confidentiality and privacy in the use of data."We must focus on the development of AI systems from a multidisciplinary perspective and with a clear alignment with ethical, legal and social aspects," summarizes Alonso Betanzos.


What has the COVID-19 be for AI and vice versa?

The pandemic caused by the SARS-COV-2 has meant an impulse for artificial intelligence by putting on the table the need to achieve results in record time: vaccine, diagnostic and treatment methods, etc..On the other hand, Pandemia itself, in turn, could say that it was predicted by an AI system.The Blue-Dot company that Machine Learning uses to detect sprouts of infectious diseases, warning about the unusual increase in cases of pneumonia, China;Although, in this case, the prediction did not serve much.

However, the accumulated data of patients infected with COVID-19 of four nationalities have served for the Big Covidata project that has used a methodology of analysis based on artificial intelligence.This is the first international study, promoted by Savana, which applies Big Data, automatic learning and natural language processing to define the clinical characteristics and predictive factors of the evolution of patients.

For Dr. Ignacio Hernández Medrano, medical director and founder of this company, “we tend to think that Machine Learning in Medicine helps us classify for the diagnosis and screening of patients, but it also helps us predict and anticipate, stratify the riskWith a precision that, otherwise, we could not do, ”he argues.

Big Covidata Results

Big Covidata is one of the broadest studies carried out so far and has provided very relevant data on the impact of this infection in different population groups with chronic respiratory diseases.

In addition, thanks to the study, the characteristics of the profile of a COVID-199 candidate to be hospitalized in the Intensive Care Unit (ICU) have been identified, once diagnosed as positive."We can predict which patient is going to end at the ICU from the moment he called the Health Center," says Dr. José Luis Izquierdo, head of Pneumology at the Guadalajara Hospital and one of the specialists who has participated in the study.

Medrano, who is also a neurologist and deputy director of the Ramón Institute and Health Research Cajal (Irycis), highlights that the advantage that Big Covidata has contributed is that the clinical trial has good efficiency, but with bad external validity, since notalways reflect the health reality of the population.

Strengths of the use of AI

In that sense, AI can provide additional advantages that have been demonstrated with the Big Covidata project.Medrano refers to three main strengths: the speed of analysis of a large amount of data almost immediately, the reuse of scientific evidence generated by clinicians and the detection and association of exploratory variables, opening new lines of work.

The project is open to a next research phase where it is intended to advance global clinical knowledge and the development of effective treatments against the Coronavirus, using new technological tools.

Other apps

Likewise, four Madrid public hospitals managed by Quirónsalud - the university hospitals Foundation Jiménez Díaz, King Juan Carlos, the Infanta Elena and the General of Villalba - have implanted a method based on AI that allows to realize in real time the progression of many patients affectedby Covid-19.A tool that, in addition to improving clinical results by speeding up medical action, allows optimizing efficiency to provide for the resources that the center will need.

Another project is Bigsalud2, developed by the Technological Institute of Informatics (ITI), in which they investigate how to detect COVID-19 through the medical image of thorax through artificial intelligence.