Instruction

• Make a summary of the following article
• Identify each part under its title (Social networks and health, BIG DATA IN SOCIAL NETWORKS AND PUBLIC HEALTH, MEDICAL IMAGING, CONCLUSION).
• Include references at the end and right next to each information in apa style.

Social networks and health

Within the health environment, social networks have emerged with a strong and clear purpose: to share health information and connect with other patients, doctors or health professionals. People have more and more expertise in the use of new technologies, which gives rise to the proliferation of this type of social networks oriented to the field of health. Although there are different types of social networks in health, depending on their orientation or type of registered users, they can be basically divided into three types: professional-oriented social networks (they connect health professionals with each other), patient-oriented social networks (they connect patients among themselves) and mixed networks (patient-physician connection).

A study carried out by AMN Healthcare (7) estimated that approximately one third of health professionals use specific social networks. However, the data reflected in the AMN Healthcare report only talks about the use of social networks by medical professionals, and as the infographic analyzes, it focuses on a very specific use of these social networks by them. However, the use of this type of network by the professionals themselves is more than interesting and relevant in view of the data that is shared on it. Some examples of this type of social networks and the potential that can be derived from the data they contain are initiatives such as Doximity (8), a social network that allows medical professionals to communicate with each other in order to share information about possible diagnoses. Doximity has more than 250,000 members in the United States, representing approximately 40% of physicians in this country.

Another social network of similar interest within the United States is Sermo, which defines itself on its own website as the largest social network for doctors in the country, and currently, in the world. This network also allows, like Doximity, discussions on any topic related to health in an open and collaborative way.

Other social networks with similar themes at the international level are (focused on collaborative diagnosis), SharePractice (oriented towards treatment), WeMedUp or Doc2Doc among many others.

At the national level or that may be oriented towards the Spanish-speaking market, initiatives such as Ippok stand out, a social network aimed at health professionals who can ask for or offer advice or help in terms of treatments, clinical cases or similar, in addition to accessing documents and job offers; Doctor Dice is considered the largest exclusive social network for doctors in Mexico and allows the exchange of all kinds of clinical information within it; MedCenter, which is a portal aimed at the medical community with a priority objective focused on education or Medicalia). Apart from these initiatives, there are many others with similar orientations to those already described.

As far as patient-oriented social networks are concerned, there are different alternatives and with different scopes. Social networks of this type can seek different objectives: from simply sharing experiences with the idea of seeking support or recommendations for certain diseases or treatments, to seeking second opinions or even treatment alternatives based on the opinions or recommendations of other users. In any case, what unites them is the patient-patient interaction, without the role of the medical professional necessarily being present as the main element of the social network.

There are different social networks that share the objective of connecting patients, one of the main and most important being PatientsLikeMe. The objective with which this social network is defined is to allow its users to share treatments or symptoms of their respective diseases in order to be able to follow up and learn from the results of others. This social network has different communities for different diseases, among which some pathologies such as Amyotrophic Lateral Sclerosis (ALS), Multiple Sclerosis (MS), Parkinson's, fibromyalgia, HIV, chronic fatigue syndrome, other mood disorders stand out. among some of the most common, although it also has communities for some rare diseases.

Other social networks with similar objectives, but often very focused on a specific pathology are, for example, tudiabetes.org, stupidcancer, curetogether or Aorana, among others.

Finally, in the mixed group that connects patients and doctors, although this type of social network is more limited, there are initiatives such as RareShare, aimed at rare diseases.

BIG DATA IN SOCIAL NETWORKS AND PUBLIC HEALTH

Public health is the discipline that deals with the protection and improvement of the health of the human population. When talking about public health, we are talking about large populations of individuals, instead of focusing on specific cases or people. Public health can be managed at different levels in terms of population sizes, depending on this precise size, since it is not the same to manage populations at municipal, provincial, state or international levels.

In this context, Big Data fits very well in what refers to the creation of data that can affect large groups of people or large populations. The Big Data concept itself is born from the generation of large amounts of data, but these large amounts of data can be generated in two ways: entities that produce large amounts of data (many or few entities), or a large number of entities that generate small amounts of data. In the end, it all comes down to a scale at the level of data generation: who creates it (how many actors are involved) and what amounts each actor generates.

In the environment, therefore, of public health, Big Data emerges through populations that represent a large number of individuals that generate small amounts of data, but that in blocks represent a large amount of information.

Users, therefore, are small-scale data generators, but the set of all those users who generate data create Big Data. The clearest example is precisely in the previously mentioned DOMO infographic, where the data generated is enormous, but it is generated precisely by different people, by an immense number of people. In this context, the data generation model changes to a somewhat distributed model and where, therefore, social networks are a source of public health information data to be taken into account, as previously analyzed (Evika Karamagioli, 2015; Kass-Hout & Alhinnawi, 2013).

Of the 500 million tweets that are generated daily, it is estimated that around 1 million are related to health issues. In addition, the numbers of health-oriented social networks, such as those described above, have spectacular numbers in the Journal of Biomedical Informatics on biomedical information in social network environments (Rodrguez-Gonzlez, Mayer, & Fernndez-Breis , 2013) and more recently a special track on analysis of medical data and social data has been organized at the IEEE Computer Based Medical Systems (CBMS) conference (27). There are many initiatives that are committed to this line of research. There are different studies and investigations already carried out both nationally and internationally on the analysis of large-scale data from social information sources. Some of these investigations include, for example, the investigation carried out by the Indiana University School of Nursing and ChaCha (28), a question-answer social network with the aim of analyzing through natural language processing techniques and analysis tools. exploration of data in real time those questions and answers related to health and well-being. Other investigations have focused on the link between the data extracted from social networks and the medical records published on these networks (Padrez et al., 2015) or to directly analyze public information from social networks in order to identify possible outbreaks of epidemics or pathological conditions (Asamoah, Sharda, & Kumarasamy, 2015; Capurro et al., 2014; Nambisan, Luo, Kapoor, Patrick, & Cisler, 2015; Paul & Dredze, 2011; Xie et al., 2013 ) as existing platforms such as HealthMap already do through the massive analysis of information from social networks and news, or as GoogleFlu (30) did at the time through the analysis and geolocation of searches related to the flu. Other studies have focused, for example, on the analysis of Facebook groups in terms of well-being and health elements such as healthy eating (Leis et al., 2013), groups related to breast cancer (Bender, Jimenez-Marroquin, & Jadad, 2011) or studies on the use of Facebook to answer medical questions on the page of a well-known medical journal (Rodrguez-Gonzlez, Menasalvas-Ruiz, & Pujadas, 2016) among many other studies of the same nature. Some European projects such as TrendMiner (31) included in their use cases precisely the analysis of social networks in the health environment, in this case specifically for the search for interactions between drugs (Segura-Bedmar, Martnez, Revert, & Moreno-Schneider, 2015).

The growth of both general-purpose social networks (Facebook, Twitter, Instagram, etc.) such as those for specific use (among which those previously analyzed stand out) has given rise to a boom in terms of the generation of information related to health. Social networks for specific use, in which information related to certain pathologies and their treatments are exchanged between patients, are a very important source of data that can be used to extract very useful data such as adverse effects, new treatment options, treatment or even to be able to detect cases of errors in clinical prescriptions among other interesting information to extract.

The social networks of professionals also allow, in addition to extracting information related to that described in the case of patient-patient social networks, the extraction and analysis of diagnoses in a collaborative manner.

Finally, general-purpose social networks are, along with the above, a high-quality source of geographically distributed information. The data to be extracted from these networks will allow the application of public health policies.

MEDICAL IMAGING

Medical imaging is today one of the main sources of information used by doctors for diagnosis and therapy. They can be considered a window on the human body, which, in a practically innocuous way, allows the characterization of different diseases, facilitating their diagnosis and treatment. As a consequence, all the technologies related to medical imaging have evolved remarkably in the last decades.

In this article we are going to refer mainly to 4 of the V's that, from the Big Data point of view, characterize medical imaging: variability, volume, speed of generation and, above all, the value that is extracted from them. The variability is mainly due to the large number of medical imaging modalities.

Regarding the volume, on the one hand, it is necessary to consider the number of images that are generated daily in each healthcare center and that number should be extrapolated to a regional, national and supranational level. By way of illustration, the evolution of this volume of data for two of the most used modalities, Magnetic Resonance Imaging and Computed Tomography, from 2010 to 2014 in Spain. As can be seen, the sum of this number of tests in 2014 exceeded 8,000,000 tests, exactly 8,278,465.

However, to really understand the volume of data that must be stored and processed in the field of medical imaging, it must be borne in mind that a medium-sized 2-D image is made up of more than a million information units (pixels); and that most medical images are 3-D and even 4-D. In this sense, there are studies that consider that the degree of complexity of the processing of medical images, estimated in number of pixels, is comparable to the complexity of the analysis of genomic data.

Regarding the speed of data generation, which in turn is closely related to the volume, a good indicator is the personal experience of any person who goes to an emergency department, but to give some objective data, we will cite the corresponding data. to a hospital in the public network of the Community of Madrid in 2013 (last year for which data are available). The average daily values of recorded Ultrasounds, Computed Tomography, Magnetic Resonance and X-rays were 11, 85, 53, 600, respectively. A very important aspect to take into account is that in order to exploit the information from this volume of data at a speed comparable to that at which it is produced, techniques and tools are necessary that allow automatic or quasi-automatic processing and analysis of the images. Only by having these techniques could it really be possible to extract all the value associated with medical imaging.

Therefore, one of the great challenges is the development of new algorithms and methodologies for the processing, analysis and interpretation of the enormous volume of data that this type of images represent, in order to help health professionals to exploit all the information contained. in the same. Since the extraction of information from these images depends on many factors, such as: modalities, recording conditions, devices, etc., it is not possible to have general procedures to locate and/or identify objects. such as anatomical structures or lesions, but it is necessary to develop particular methods for each type of image and/or pathology. Once the desired information has been extracted from each particular image, the next step consists in associating this knowledge with the images (image semantic annotation), so that, once stored in large databases, it is possible carry out semantic searches of the same or, in other words, searches by concepts. The annotation consists of a set of words capable of describing the information contained in the image. Traditionally, image annotation has been carried out manually by specialized personnel. However, this process has some disadvantages, such as the cost of time and the subjectivity of the operator. Even if the same operator frames and annotates the same image at different times, the annotation will not necessarily be the same. The alternative approach is automatic or semi-automatic annotation performed by an algorithm. Most of the approaches found in the literature are based on machine learning techniques (Menasalvas & Gonzalo 2016).

A problem that has been detected in order to move on to the clinical application of the most theoretical studies is the need to have large databases that make it possible to objectively and rigorously determine and validate which are the best algorithms to use for each application. In this sense, it is of great interest, the competitions proposed in recent years, in which annotated data of real cases have been provided for which the reality-terrain information is known, and a common validation protocol has been established. In this way, it is ensured that the results provided by different algorithms are technically comparable and reliable for their clinical application.To carry out semantic searches of the same or, in other words, searches by concepts. The annotation consists of a set of words capable of describing the information contained in the image. Traditionally, image annotation has been carried out manually by specialized personnel. However, this process has some disadvantages, such as the cost of time and the subjectivity of the operator. Even if the same operator frames and annotates the same image at different times, the annotation will not necessarily be the same. The alternative approach is automatic or semi-automatic annotation performed by an algorithm. Most of the approaches found in the literature are based on machine learning techniques (Menasalvas & Gonzalo 2016).

A problem that has been detected in order to move on to the clinical application of the most theoretical studies is the need to have large databases that make it possible to objectively and rigorously determine and validate which are the best algorithms to use for each application. In this sense, it is of great interest, the competitions proposed in recent years, in which annotated data have been provided.

CONCLUSIONS

The application of the Big Data paradigm to the health environment will mean an improvement of an as yet unpredictable magnitude in the quality of patient care, as well as in the prevention, diagnosis and treatment of diseases, together with a notable reduction in costs. of health. To achieve these achievements, it is essential to integrate all the data from very different sources, as well as the development of new technologies that allow the exploitation of said data. However, the true value of Big Data in health will only be achieved if the different actors involved in the process (public administrations, private companies, hospitals, doctors, research centers, universities,...) commit to this project together to bring the healthcare field into a new era. This can only be carried out within the framework of a Big Data ecosystem in health in which it is integrated together with technology, adequate privacy and confidentiality policies, infrastructures and a culture of data sharing. All this entails a series of challenges that must be faced from different perspectives and degree of depth.