Medicine, Big data and Artificial Intelligence

.

Fernando Lopez

Telemedicine is rapidly catching up with medical practice, with tools that are worldwide applied.​ Although years ago, this concept was merely hypothetical futurology, nowadays, it has become a set of real operational options, which is increasingly accessible to anyone’s reach. Below we will depict a general panorama of artificial intelligence technology which complement and boost telemedicine. More specifically, we shall focus on text analysis of the medical jargon domain.
 
Data Science provides management applications, packs of methods based on quantitative models, and complex algorithms which until recently, only researchers had access to this state-of-the-art knowledge. Accuracy of diagnosis was limited to the expertise of manual evaluation and conclusions were drawn from a data volume which was impossible to process by a single individual or a traditional computer system.


Image classification was one of the most spread machine learning. A model is trained using examples labelled positive or negative in connection with a specific characteristic. The outcome is a “pattern,” a classifier capable of working out the presence or absence of whatever it is we are looking for. We might be searching the pattern of something we know, or check the opposite, that there is no pattern, in which case there is an “anomaly.” For example, there are several applications for mobile phones which can detect skin cancer from an image.
 
Also, Genomic Analysis is another example of Big Data, as is any product resulting from sequencing ADN or ARN. At any rate, each analysis involves files of dozens and hundreds of gigabytes. Next Generation Sequencing (NGS) technology has dramatically reduced the cost of the basic genomic analysis due to their platforms for bioinformatic analysis, that will soon be available to all patients.

Odoo image and text block

Text treatment, meaning extracting knowledge from available literature has been widely developed and multiple specific lines of work have been created. Working with a text written in natural language implies being acquainted with the sources and the application area as well. For instance, there are medical records where a patient’s symptoms have been described, past treatments and even diagnosis or academic articles, which may be available in several languages. Here, the first problem is to agree on terminology: the correct name for a procedure, disease, symptom and vice versa, how to tell when a text is referring to a specific concept. It has been extremely important for Uruguay to be integrated to Snomed CT which gathers ontology collections of medical terminology with more than 400.000 concepts, available in several languages, and has local terminology too. The first version of Snomed CT was released by SALUD.UY in 2016.
 
Additionally, there is a continuous flow of publications with the latest findings in medical sub-disciplines. PubMed of the National Institute of Health of the United States is one of the best examples, because it contains hundreds of articles, hypothesis, conclusions, and methods which shape the state of the arts' evolution of medicine. In Latin America there is one too, Scielo, a scientific library on line, with many articles and proceedings taken from conferences.
 
It is impossible for any individual to have access to all this literature. It is here when Natural Language Processing (NLP) plays a leading role, supported by the latest tools of artificial intelligence (AI). When research is on progress, NLP + AI extract all the relevant information, identifying the best sources, and can even suggest conclusions. Needles to say, it is the medical team’s responsibility the decisions and the diagnosis they make. In other words, it all boils down to providing medical doctors with relevant information, drawn from all the sources and news available at that moment.
 
Information may also be obtained by using natural language, actually talking with the tool, using a conversational interface (chatbox) or a "virtual assistant,” its more sophisticated version. We can hold a written or spoken conversation by means of voice to text technology.


Finally, despite all the technological progress made, like having electronic clinical records, medical doctors, and staff, are irreplaceable  to integrate the analysis and diagnosis process, from beginning to end, and responsible for the decisions made with respect to diagnosis and treatment.

Fernando López Bello @fer_lopezbello

Ingeniero en computación, PMP
Big Data Expert