Читать книгу The Future of Health - Roberto Ascione - Страница 18

Verum scire est scire per causas, true knowledge is a knowledge of the causes. This was written by Thomas Aquinas, taking up one of the cornerstones of Aristotelian philosophy. It is in this belief—pin the discovery of the cause of a phenomenon—that the scientific method is oriented. Let it be understood, it still does this, but big data analysis has short-circuited this procedure. The science of big data causes, in the opinion of many, a decisive paradigm shift away from the past, based on a style of scientific reasoning that is totally different from what it once was. It is possible to draw a linear path in the history of science: an almost exclusively empirical approach consisting of the mere observation and description of natural phenomena followed the theoretical approach, which was based on the construction of models and the overall explanation of phenomena through general laws. The last phase, born a few decades ago, is the computational phase, which has as its cornerstone the simulation of complex phenomena through computerized models. Along this line, data-driven science stands as a new phase, a new piece of the jigsaw puzzle, a new milestone where experiment, theory, and data processing are synthesized in statistics. The use of big data to do mass screening and prevention is based on the identification of patterns, regularities, cyclical repetitions, and large-scale correlations that make it possible to process a prediction of what will be most likely to be found in the real world in similar contexts. In the science of health care, this predictive technique is made possible both by the enormous advances in the technology and biomedical fields in recent years (and the acceleration in this direction is astounding) and by access to an enormous amount of data, something that was unthinkable until recently. Suffice it to say that we are now able to synthesize an impressive amount of information from various sources. The amount is in the range of several petabytes (one petabyte corresponds to 1,000,000 gigabytes). The computer that allowed the Apollo 11 lunar module to land had a memory in the range of 104 bytes, so just this one comparison should be enough to understand the extent of the acceleration that has taken place in the field of data processing. However, there are some obvious critical issues in this methodology as it is applied to health. The first obstacle is the objective difficulty in governing this huge mass of data, especially when it is used and applied differently from the initial purpose of their collection. Then there is, of course, an obstacle related to privacy—specifically, that people are often unaware of how their data is used. These are ethical and transparency dilemmas that cannot be overlooked, and solutions are required. Artificial intelligence (AI) has entered our lives, especially in health care, with applications and software able to autonomously set up treatment and care plans for patients. AI provides “reasoned” information to health-care professionals to best assist their work, allowing them to prescribe the best possible care.

Several companies around the world have built tools for use in AI-based health care. Deepmind Health, a company incorporated by Google Health, is one of them, and it has developed an AI program that can understand and process thousands of data and medical information in only a few minutes, to then translate it into services.

AI and its applications are experiencing a time of rapid advancement, particularly in the area of machine learning (ML). The most used machine learning, known as supervised ML, is software capable of learning to classify a set of data from the analysis of many similar cases, previously categorized by humans. The source of fuel for this process is big data, the extensive digital data sets made available, for example, by diagnostic equipment or the digitization of medical records. The machine learning approach can be applied in multiple fields, from image detection to understanding genetic data, from diagnosis through digital phenotype to drug development. Based on these factors, it is reasonable to imagine an increasing collaboration between people and artificial intelligence, where the clinical reality will become more and more data-centric, so that every detection, decision, and therapeutic intervention will be codified and recorded. It will be crucial that educational curricula of all medical professions include familiarization with these technologies, which will become increasingly important as day-to-day tools and working partners.

In addition, the use of artificial intelligence is also entering into the health-care sector. For example, in Japan, a new AI-assisted endoscopic system that would be able to identify colon and rectal adenomas during colonoscopy has recently been tested in a clinical setting. To achieve this, the diagnostic method uses an endocytoscopic image of a polyp enlarged 500 times and analyzes about 300 qualities of the polyps examined. Comparing the characteristics of each polyp with those of more than 30,000 reference images has allowed this smart probe to diagnose the presence of a dangerous lesion in less than a second with staggering precision.