Big Data, AI, and Healthcare: A Comprehensive Review

Over the last 45 years, the general activity of making sense of data has evolved from decision support to executive support, to online analytical processing, to business intelligence, to analytics and now to “big data”. One of the terms proposed to define big data is the collection and interpretation of massive data sets, made possible by vast computing power that monitors a variety of digital streams, and analyzes those using “smart” algorithms (Davenport, 2014). Healthcare has become of the key emerging users of big data (Dimitrov, 2016).

Big data analytical tools have the strong potential that allows healthcare professionals to gander upon the clinical data stored within repositories and assists in the process of informed making of decisions. It will be seen very soon that the healthcare industry is making use of this tool for a wide assortment of findings. However, the tools still have to address many difficulties for instance assuring privacy and security, developing the standard and supremacy, and improving the tool technically will be the matter of concern at present. Large findings analytics and its implementation within the healthcare organization are still considered to be at the budding stage; however, with the rapid progress in the technical advancement of the tools will escalate its use.

Application of Artificial Intelligence in the healthcare sector

The term "Artificial Intelligence" has been introduced by John McCarthy within a conference held at Dartmouth in 1956 and explained it as "the science and engineering of making intelligent machines" (Society for the Study of Artificial Intelligence and Simulation of Behavior, 2018). Following that, a period came where there was diminished interest and resource in the field of research concerning AI had been observed also being referred to as AI winter (Crevier, 1993). The concept of AI is a significant part of computer science that seeks to develop intricate models with the attributes of human intelligence. This concept is generally referred to as "General AI" (Copeland, 2016). The concept that operates now at present belongs to "Narrow AI" and here technology can execute assignment superior in comparison to human beings for instance the recognition of face and speech (Copeland, 2016). These techniques have advanced facilities from deep learning and AI for analysis of speech, identification of face and image, and have immense application in the field of natural language processing, autonomous vehicles, and in medicine.

Artificial Intelligence (AI) applications in healthcare have gained attention in recent years. According to B.J Copeland, Professor of Philosophy and Director of the Turing Archive for the History of Computing, University of Canterbury, New Zealand, AI is regarded as the ability of the computer-controlled robot or a digital computer to perform tasks linked to intelligent beings. AI is commonly applicable in project development systems said to be endowed with potential intellectual processes (2019). Eric Topol, the cardiologist, author, and researcher, argues that AI in medicine is beginning to have an impact on clinicians, predominantly via rapid and accurate image interpretation; for health systems, by improving workflow and reducing medical errors; and for patients, by enabling them to process their own data to promote health (Topol, 2019).

AI in the Ophthalmology

On the other hand, ophthalmology is a significant branch of surgery and medicine that focuses on diagnosis as well as treatment of eye disorders (Nelson, 2018). A partial list associated with ophthalmology includes macular degeneration, cataract, dry eyes, proptosis, glaucoma, and diabetic retinopathy. The study of AI and ophthalmology is more pertinent and relevant when handling such areas as early disease detection, pattern recognition, and data management. According to Akkara and Kuriakose, ophthalmologists, and researchers in Kerala, India, AI is said to have made its entry in healthcare and modern life and is showing promising results (2019). In addition, ophthalmology is a growing field that is engaging in measurable data and imaging, which makes it ideal for AI and machine learning applications.

AI has demonstrated its significance in the field of ophthalmology for the interpretation and recognition of patterns or trends within the clinical images for instance the identification of the 3D retinal scans by Google's DeepMind Health. The field of medicine is constantly evolving with the rapid enhancement and development of the computer-based AI tools which had resulted in precision and success in the process of diagnosis across different fields of specialization. With the rapid utilization of AI in radiology, some specialists also have suggested that AI may replace radiologists in the future. However, the question arises that AI will substitute the physicians or boost up their role in their field with the help of AI rather than replacement. Therefore, the main objective of this research is to comprehend the significance of AI in the field of medicine as AI has expanded its role in the field of radiology, cardiology, ophthalmology, and pathology. Thus overall it will enhance the role of the physicians without hampering the physician-patient relationships.

The up-gradation of the healthcare research is based on the analysis of large findings and in the field of ophthalmology, the success of AI tools will depend on the amalgamation of varied aspects of application along with the tool (Patel et al, 2009). With this approach, it is significant to manage the information of the patients stored within the electronic medical records in a holistic approach for analysis. However, the approach creates a problem for the researchers to utilize the AI tool in the aspect of modern health research along with the help of healthcare professionals. The classy tool will allow the healthcare providers to acquire an approach to the clinical records of patients and to formulate better care services to patients. Ophthalmologists are making use of artificial intelligence or machine learning tools which is making the process of detection automated and detecting a variety of problems among the patient with an enhanced level of supervision. According to Andrew P. Schachat, MD, Editor in Chief of Ophthalmology Retina (2019) the significance of AI, profound learning through computers that would augment the care services for the patients can be clearly comprehended. Moreover, according to published scientific evidence on deep learning models that could perfectly project the succession of age-related macular degeneration through the model named DeepSeeNet. This model analyzed color fundus pictures (58,402) and also examined by itself nearly 900 pictures that have been acquired from a longitudinal follow-up investigation of 4,549 sample size participants from AREDS. The findings revealed a comparatively better precise identification of large drusen and changes in the pigmentation which is considered to be as per the diagnosis by a retina specialist. Moreover, the computer could also simplify whether it is fluid or not just like a retina specialist (Schachat, 2019).

Machine Learning, a branch under data science, which combines statistics, computer science, information technology, and data visualization, is a powerful tool with infinite possibilities to enhance clinical practice in identifying certain ocular conditions (Consejo et al, 2019). When combining multiple variables in a study, such as social health factors, environmental factors, and climatic factors through machine learning tools, new insights, and expanded findings can be generated to help in clinical decision making and in patients care.

One of the strengths that make AI ideal for ophthalmology is that ophthalmology is a field of medicine with a lot of imaging and measurable data, thus ideal for AI application. AI has been more relevant in the analytical process of the retinal fundus images associated with diabetic retinopathy. This can be followed by what is referred to as age-related macular degeneration, retinopathy of prematurity, and glaucoma. Based on the findings made by Nelson (2018), major technology companies have taken strides towards AI and ophthalmic use. IBM’s AI, for instance, has the capacity of predicting the visual field data associated with OCT scans. In addition, DeepMind Health, Google’s artificial intelligence business, helps in the diagnosis of eye disease by analyzing medical images. It analyzes 3D retinal scans for signs of major eye diseases, such as glaucoma or diabetic retinopathy. AI can also analyze the scan immediately while patients would ordinarily have to wait days for a specialist to review the images (Esson, 2018).

A recent study in the Journal of Ophthalmology asserted that healthcare has emerged as a significant area at the center of the AI application (Lu et al, 2018). A range of studies has been linked to another branch of Data Science, called Deep Learning (DL). The algorithms in DL are said to be performed at high levels. This has been applicable to breast histopathology analysis as well as skin cancer classification. Other areas of concern as noted by the study include ophthalmology, lung cancer detection, and cardiovascular risk prediction. With a vast range of applications, it is paramount to note that AI in ophthalmology is imminent and unstoppable following the development of AI algorithms and accessible data sets such as Messidor, EyePACS, and Kaggle’s data set. Subsequent observations made in the International Journal of Ophthalmology, indicate that AI can be applied in terms of both (DL) and Machine Learning (ML). This has bolstered subsequent diagnosis of ocular diseases which covers the most leading causes of blindness, diabetic retinopathy, cataract, age-related macular degeneration as well as glaucoma (Du et al, 2018). ML approaches said to have been introduced by Sandrina Nunes and Miguel Caixinha, who have been paramount in monitoring and diagnosing ocular diseases. ML largely attracts small data sets but it can turn out to be cumbersome when it comes to handling visual features. DL, as part of the AI application, is known for having the ability to discover the most intricate structures across data sets even without specifying the rules.

Further observations made by Lu et al. (2018) noted that current studies are putting more focus on machine learning, which can attain satisfactory outcomes. More focus on AI and diabetic retinopathy have attracted towards retinal microvasculature, which leads to damage. More people are essentially affected by diabetic retinopathy, and this has been turned into a public health problem across the world. Large scale screening of diabetic retinopathy is on high demand as far as treatment and management are put into consideration. Practitioners have consistently called for early intervention, which taps into diabetic retinopathy automatic identification. Further attention given to neovascularization detection, microaneurysm, cotton wool spots, hemorrhage, and exudation has raised hopes of AI application. In this case, computers can receive images which can be labeled as diagnostic lesions before identifying the final judgment and input images.

Besides, Ting et al. (2019) noted that the adoption of DL in natural language processing, image recognition and speech recognition has impacted the approach towards healthcare. Apparently, in ophthalmology, the application of DL in visual fields, fundus photographs as well as optical coherence tomography has led to the achievement of robust classification performance. A large number of imaging and image processing techniques available nowadays present new opportunities to develop decision-support tools that assist clinicians with the diagnosis of almost any ocular condition.

In the sphere of Data Analytics, another branch in data science, healthcare has shown the potential to become ‘smarter’ and more effective. The use of artificial intelligence has been enabled by big data, along with markedly enhanced computing power and cloud storage, across all sectors (Topol, 2019). The healthcare industry generates a lot of databases including information on patients, diseases, demographics, and much more. The potential to improve the quality of healthcare delivery and at the same time reducing cost is very promising. The massive quantities of data, also referred to as ‘big data’, hold the promise of supporting a wide range of medical and healthcare functions, including clinical decision support, disease surveillance, and population health management (Ragupathi, 2014). Carol McDonald, a developer in health systems and experts in Java, is also a supporter of utilizing big data to reduce cost, in addition, to improve health outcomes (2019).

Researchers at the Johns Hopkins School of Medicine discovered they could use data from Google Flu Trends, a novel internet-based influenza surveillance system that uses search engine query data to estimate influenza activity, to predict sudden increases in flu-related emergency room visits at least a week before warnings from the Center of Disease Control. Despite predictive flaws in Google Flu Trends, the analysis of Twitter updates was as accurate as official reports at tracking the spread of cholera in Haiti after the January 2010 earthquake (Ragupathi, 2014). The authors used HealthMap, an automated surveillance platform, to measure the volume of news media generated during the first 100 days of the outbreak, and they also looked at the number of 'cholera' posts on Twitter. The study found that online social media and news feeds were faster than, and broadly as accurate as of the official records at detecting the start and early progress of the epidemic, which hit Haiti after the earthquake in January 2010 and has killed more than 6,500 people (Hirschfeld, 2012).

Issa et al (2014), researchers at Georgetown University Medical Center support the use of electronic medical records (EMRs) to create new knowledge in the field of healthcare. EMRs are rich with clinical data that chart patient progression with respect to disease, medications, with other demographic information. Family history, diet, medications, and occupational exposures are just some of the documented information that could be invaluable in determining unique treatments for individuals (2014). EMR information is uniquely positioned to aid in the discovery of new findings when coupled with other datasets. In other words, combining biomedical information with environmental and social contributors would provide a holistic system view of a patient, and will highlight new ways to intervene in enhancing patient care either by early detection of a disease or by a more accurate prevention method. To date, no such platform exists as EMR records have not been transformed yet into the study for clinical medicine, but this is expected to change (Issa et al, 2014). In addition, government agencies worldwide are releasing public datasets about the services provided in healthcare, such as Medicare in the US. These datasets can be queried by multiple classes of users, including hospitals, patients, physicians, and policymakers. However, to realize the true value of the information present in these datasets, appropriate analysis, including classification and clustering needs to be performed. Additional insight can be gained by combining the healthcare data with other data sources such as demographics and epidemiology (Rao et al, 2015).

Furthermore, Yang et al (2015), researchers in Information Technology and experts in Computer Science from China, state that since healthcare covers complex processes of the diagnosis, treatment, and prevention of diseases, EMRs can be used in detecting medical problems at an earlier stage, if the data is collected and managed properly. Moreover, they state that technologies are not solely used anymore for therapeutic purposes, but analysis using big data and cloud computing can reveal trends and can be used in predictive medicine. Existing methodologies for the detection and analysis of medical conditions will have to be revised and extended to discover deep knowledge and deliver enhanced patient care. Cloud computing can support the analysis of big data through innovative technologies and software. Data mining (DM) is the computing task to discover unsuspected patterns from the observational datasets to help users to make better decisions (Zhou et al, 2010). In general, the discovered patterns are novel understandings and represent something hidden in the available dataset that the users did not know before. “The application of data mining algorithms for medical data analysis and utilization can be classified into two categories, i.e., unsupervised (descriptive) and supervised (predicative) approaches. The unsupervised methods mainly concern data clustering, i.e., grouping data into clusters by measuring the similarity between objects or EMRs to discover unknown patterns or relationships in the available datasets. The typical unsupervised data mining approaches cover data clustering, association rule mining, and sequence discovery”

EMRs have been adopted only relatively recently in ophthalmological practice globally; however, take-up has been swift, largely due to the realization that this electronic record management will enable large amounts of clinical data to be used in research. The practice is, however, in its infancy; moreover, clinical data tend to be more complex than other types of data whose management has been transformed by the digital revolution, and thus the expected benefits have not yet been seen in all cases (Boland, 2016).

Nonetheless, EMRs have already proved of significant value within ophthalmological research, offering new insights into matters ranging from disease surveillance through how the health service is utilized, and giving a more detailed picture of outcomes. Furthermore, there has been an increase in the sheer quantity of data available to researchers. It has therefore been recommended that data linkage systems are used to guide future research (Clark et al., 2016).

The overall significance of AI in the healthcare sector:

AI is considered to be enriched with algorithms, huge data obtained from clinical records, health monitors, and computer programs. The market value of AI in the healthcare sector is constantly rising at about 40% and it will reach up to $6.6 billion by 2021 (Frost & Sullivan, 2016). Large records can be stored within cloud and training or learning on algorithms allows gaining perspectives in the field of diagnostics, treatment management, and patient findings (Bresnick, 2018b). AI is well structured to manage the repetitive procedures of work, deals with huge cores of information and helps to make decisions without errors. The study of Frost & Sullivan had highlighted that with the use of AI, patient outcomes can be augmented from 30 to 40% and diminishes the expenditures of treatment by 50% (Hsieh, 2017a). According to specialists AI has immense application within varied areas of health such as dealing with long term diseases and up taking the right judgment (Bresnick, 2016). AI has made rapid progress in the healthcare and medicine sector because of large chunks of data available in the EMR in recent years along with raised efficiency of computing and technology (Pratt, 2018). Therefore, AI could be able to evaluate the findings of patients acquired from varied origins for instance the fitness trackers and home monitors that assist physicians to look after patients which would not have been possible without AI (Pratt, 2018).

Conclusion

As seen from the above statistics and information contained in the proposal, there is a strong need for research into the authorization models that are used for electronic medical records. Without some kind of an authorized access model, health administrators can only take a reactive approach to ensure patient privacy. Creating and implementing an authorized access model for health care systems and exchanges will proactively protect patient data and ensure the continued growth of interconnected health networks. Currently, health exchanges are in their early stages of development. Due to the small size of the networks they can grow without having a sound authorization access model and in its place the networks rely almost entirely on audit logs; a reactive measure. While auditing is a good practice to develop in any information system there must be other measures to ensure complete data control and privacy. Especially as health exchanges and systems begin to grow to meet the eventual nationwide interconnected health network, relying solely on audit logs will not suffice. Shelc 60 Traditional authorization access models are too rigid to conform to the dynamic and ubiquitous nature of the healthcare system. Implementing an authorization model centered on user control of access can help alleviate some of the shortcomings of traditional access models. Although there still will exist barriers before a user-centered authorization model can be implemented this paper aims to begin research into the feasibility of such a system. The movement of personal information into a networked environment has happened in almost all other major industries today. Users can view their financial, social, and personal information online and in some cases control who has access to it. The ability of users having access to their many forms of personal information creates awareness and empowers them with the ability to make informed decisions on authorized access. When it comes to a patient’s health no one knows their medical record better than themselves, allowing users some control over access control will create a system that can allow the digital health revolution to continue.

Big data analytics has the potential to transform the way healthcare providers use sophisticated technologies to gain insight from their clinical and other data repositories and make informed decisions. In the future we’ll see the rapid, widespread implementation and use of big data analytics across the healthcare organization and the healthcare industry. To that end, the several challenges highlighted above, must be addressed. As big data analytics becomes more mainstream, issues such as guaranteeing privacy, safeguarding security, establishing standards and governance, and continually improving the tools and technologies will garner attention. Big data analytics and applications in healthcare are at a nascent stage of development, but rapid advances in platforms and tools can accelerate their maturing process.

The healthcare industry is modernizing its move into research through processing and analyzing big data. The early success of AI in healthcare, including in ophthalmology will depend on the development of integrated environments that allow the merging of knowledge-based tools with other applications (Patel et al, 2009), and thus it seems both worthwhile and medically significant to utilize systems such as electronic medical records, patients’ data, and publicly available data for a holistic data management and analysis. This reality creates challenges for researchers because the implication is that researchers need a collaborative approach with healthcare professionals to be able to leverage AI techniques into modern research. Big data analytics has the potential to transform the way healthcare providers use sophisticated technologies to gain insight from their clinical and other data repositories to make decisions that will enhance patient care.

Ophthalmologists are using artificial intelligence and machine learning to develop programs to automate diagnoses, improve monitoring and help detect a variety of ophthalmic diseases in patients. Several studies published in 2019 show the impact of deep learning, computers and artificial intelligence on improving patient care in the specialty, according to Andrew P. Schachat, MD, Editor in Chief of Ophthalmology Retina (2019). For example, researchers published a study in November 2018 that showed how a deep learning model could accurately predict age-related macular degeneration progression through a model called DeepSeeNet, which trained itself by evaluating 58,402 color fundus photographs and tested itself on 900 images from a longitudinal follow-up of 4,549 participants from AREDS. The result was a more accurate detection of large drusen and pigment changes than how retina specialist would diagnose it. The computer could also diagnose whether there is fluid or not, just as a retina specialist would (Schachat, 2019).

In ophthalmology, AI has shown its decision-making skills in recognizing and interpreting patterns in clinical images. A popular example is the detection of the 3D retinal scans by Google’s DeepMind Health.

The practice of medicine is changing with the development of new Artificial Intelligence (AI) methods of machine learning. Coupled with rapid improvements in computer processing, these AI-based systems are already improving the accuracy and efficiency of diagnosis and treatment across various specializations. The increasing focus of AI in radiology has led to some experts suggesting that someday AI may even replace radiologists. These suggestions raise the question of whether AI-based systems will eventually replace physicians in some specializations or will augment the role of physicians without actually replacing them. To assess the impact on physicians this research seeks to better understand this technology and how it is transforming medicine. To that end this paper researches the role of AI-based systems in performing medical work in specializations including radiology, pathology, ophthalmology, and cardiology. It concludes that AI-based systems will augment physicians and are unlikely to replace the traditional physician–patient relationship.

The term “Artificial Intelligence” (AI) was first coined by John McCarthy for a conference on the subject held at Dartmouth in 1956 as “the science and engineering of making intelligent machines” (Society for the Study of Artificial Intelligence and Simulation of Behavior, 2018). After a period of reduced funding and interest in AI research, also referred to as the AI winter (Crevier, 1993), optimism in AI has generally increased since the low point in the early 1990s. Artificial intelligence (AI) is an important field of computer science that seeks to create complex machines with characteristics of human intelligence. We can think of this concept as “General AI,” which has machines that can think and reason and even see and hear like humans (Copeland, 2016). This concept which can be seen in movies like Star Wars (think C-3PO, a droid programmed for etiquette and protocol) is not something we can achieve at this time. However, what is achievable at this time falls under the concept of “Narrow AI” where technologies exist to perform specific tasks as well as, or better than, humans can (Copeland, 2016). Examples of such narrow AI include speech recognition, facial recognition, etc. These technologies exhibit certain facets of human intelligence. Such intelligence is derived from AI techniques known as machine learning and deep learning which have improved performance in areas such as image classification, text analysis, speech and facial recognition with a range of promising applications such as autonomous vehicles, natural language processing, and in medicine.

AI is poised to play an increasingly prominent role in medicine and healthcare because of advances in computing power, learning algorithms, and the availability of large datasets (big data) sourced from medical records and wearable health monitors. The health care market for AI is increasing at a rate of 40% and is expected to reach $6.6 billion by 2021 (Frost & Sullivan, 2016). Computing power is increasing rapidly due, in part, to the wide availability of Graphics Processor Units that make parallel processing even faster and the availability of seemingly infinite compute resources on demand in the cloud. Big data is also well supported by practically endless storage in the cloud. Learning algorithms are becoming more precise and accurate as they interact with training data, allowing newer insights into diagnostics, treatment options, and patient outcomes (Bresnick, 2018b). The flood of health care data is helping push the development of new AI applications that promise to improve the efficiency and effectiveness of patient care. Healthcare related big data is available from sources such as Electronic Medical Records (EMR) and wearable health trackers, which can be analyzed in new ways. The rise of AI in the era of big data can assist physicians in improving the quality of patient care and provide radiologists with tools for improving the accuracy and efficiency of diagnosis and treatment. AI is well-suited to handle repetitive work processes, managing large amounts of data, and can provide another layer of decision support to mitigate errors. The research firm Frost & Sullivan estimates that AI has the potential to improve patient outcomes by 30% to 40% while reducing treatment costs by up to 50% (Hsieh, 2017a).

Experts predict AI to have a significant impact in diverse areas of health care such as chronic disease management and clinical decision making (Bresnick, 2016). While still in the early stages of adoption, AI algorithms are showing promise in specializations such radiology, pathology, ophthalmology, and cardiology (Hsieh, 2017a). This progress raises a thought-provoking question. Will AI at some point displace certain physicians such as radiologists or will it help make them more effective or will it be a bit of both? This research looks at the potential uses of AI in medicine and considers the possibility of AI replacing certain physicians or at least supplementing the role of physicians.

As pointed out earlier, several factors have come together recently to support the quickening pace of AI developments in medicine (Pratt, 2018). These include the amount of healthcare data collected in recent years, the high-level computing power at low cost now available to process large datasets, the increasing prevalence of EMRs, and overall advances in computing technologies, which have all fueled AI’s advancements in medicine (Pratt, 2018). While AI in medicine is still in its early stages, it is well positioned to make positive impacts in clinical medicine. As an example, AI could collect and analyze patient data gathered from multiple sources such as fitness trackers and at-home monitors and enable physicians to monitor patients’ health in ways that time and resources without AI would not permit (Pratt, 2018). Some of the specializations in medicine where AI is having a positive impact include radiology, pathology, ophthalmology, and cardiology. This section discusses the impact and potential of AI in these specializations.

References Used:

Patel, V.L., Yoskowitz, N.A. and Arocha, J.F., 2009. Towards effective evaluation and reform in medical education: a cognitive and learning sciences perspective. Advances in Health Sciences Education, 14(5), pp.791-812.

Schachat, A.P., 2019. Ophthalmology Retina is now indexed in MEDLINE/PubMed. Ophthalmology Retina, 3(6), pp.457-458.

Crevier, D., 1993. AI: the tumultuous history of the search for artificial intelligence. Basic Books, Inc.

Copeland, M., 2016. What’s the Difference Between Artificial Intelligence. Machine Learning, and Deep Learning.

Take a deeper dive into Autism Spectrum Disorder with our additional resources.