Uniqueness of Medical Data Mining: How the new technologies and data they generate are transforming medicine

Abstract

The collection of healthcare data occurs also at social media giant tech corporations [39]. Data collected on all of us, such as by Facebook, can be and are used for dubious purposes like social engineering and marketing, along with potentially good purposes [32]. Unfortunately we have witnessed huge data breaches (Facebook, Yahoo) that make data available to bad players. Some new technologies can alleviate this problem, for instance, using blockchain technology leaves a trace of every transaction performed, such as sending data from one hospital to another, but this works only for official data transfers, as blockchain cannot prevent data from being hacked or stolen. Because of the latter, we shall talk about P5 medicine, by adding the Privacy-preserving attribute to the predictive, preventive, personalized, and participatory. It is a new task, to develop technologies to better protect health data in the future.New technologies and medical data mining. We describe below how new technologies, several of which came into being since publication of our original paper on the uniqueness of medical data mining [25], that collect large-scale heterogeneous medical data on individuals and populations, are changing the field of medicine. Examples of data generated by those technologies are image and omic (genomic, proteomic, etc.) data. Below, we highlight the biggest changes that have occurred since the Cios and Moore paper [25].Human medical data are at the same time the most rewarding and difficult of all biological data to analyze [101]. Humans are the most studied species on earth and data and observations collected on them are unique and cannot easily be gained from animal studies. Examples are visual, cognitive and perceptive data, such as those relating to discomfort, pain and hallucinations. Animal studies, being shorter, cannot track long-term disease processes, such as atherosclerosis. Since the majority of humans have had at least some of their medical information collected in digital form, this translates into big data. Unfortunately, analyzing human data is not straightforward because of the ethical, legal, social and other constraints that limit their use. Hospitals are the ones who store majority of medical data, however, many of them have little interest in sharing them, which adversely impacts global healthcare by restricting large-scale data analysis and new findings. While the current trend is towards open-access data and collaborative environments, issues related to medical data privacy and market competition are not easily alleviated.On the technology side we have seen creation of disruptive technologies such as blockchain, cloud computing, wearable devices, and augmented reality. A disruptive technology does not have to be an entirely new one: often it has been long-existing but was greatly improved in terms of, for instance, speed or accuracy. Examples of the latter technologies are artificial intelligence (AI) and machine learning [9], natural language processing (NLP) [110], image understanding [...