The Role of AI, Big Data, and IoT in Health Care

Big data analytics has a tremendous ability to change medical models and business into intelligent and efficient care. This allows anonymous collection of health information to allow secondary use of data. Authentic decision making can also be facilitated by recognizing patterns and decoding associations. In clinical practice, big data analysis can contribute to the primary detection of diseases, the precise prognosis of disease courses and deviations from healthy conditions, the detection of symptoms, and the uncovering of fraud [14]. Management and analysis of large health data is critical to ensure patient-centered care. As the type and size of data sources have grown over the past two decades, obsolete data management tools have become inadequate.

Modern and innovative big data tools and technologies are needed that can complement the ability to manage large health data. Forecast for a research study on global big data expenditure in health care to reach an average compound interest rate (CAGR) of 42% in 2014-2019 [15-17].

Figure 3.1 shows the Google trends from 2004 to 2019 for the analysis of big data, AI, and IoT in health care. Google Trends is a distinctive, freely available online portal from Google Inc. that enables consumers to collaborate with Internet hunting

data, providing an in-depth view of human activities and health-related events. Google flu and dengue trends are now being used to broadcast the spread of flu and dengue-like diseases. Google Trend has been used in a number of publications [18, 19]. Figure 3.1 shows that the term “big data in health care” first was used in early 2013. The growing attention in this topic can be linked to a widespread McKinsey & Company report that was published in early 2013 [20].

AI and big data in medicine are still in the early development stage but are growing rapidly. Although AI algorithms manage big data effectively and access data quickly to solve healthcare issues, their development is an open question to answer in the future. However, AI algorithms will not grow well unless IoT devices generating large amounts of structured data actively help them to do so. AI has become more powerful due to big data generated by IoT and historical medical images. IoT- generated big data has different characteristics than common big data because of the variety of data it collects. Figure 3.2 depicts the relationships among AI, IoT, and big data.

The interconnection of health care systems with IoT and the emergence of big data require AI to automate the clinical processes and provide rapid data analysis to identify current, new, or future issues [21,22]. System AI should be activated so that the machine can automatically forecast the outcomes from its conjugation of experience and determine them. In order to activate systems AI, natural language processing (NLP), knowledge representation, automated thinking, and machine learning must exist in the system. NLP helps to make computers/machines intelligent in the same way as humans are in understanding language [23].

Image-Based Diagnosis

With megapixels of megapixel data to the results of X-rays, CAT scans, MRIs, and other test methods, combining high-resolution images can be a challenge even for the experienced clinician. Artificial intelligence has already shown that it can be a valuable partner for radiologists and pathologists to increase their productivity and improve their accuracy [24]. Advanced AI algorithms, especially deep learning,

have shown remarkable advances in image recognition tasks. Numerous medical experts, including those in ophthalmology, dermatology, radiology, pathology, and neurology, rely on image-based diagnostics [25]. Historical picture databases are mostly stored by radiology departments in an image collection and communication system, which usually contains thousands of examples of training networks [26].

The convolutional neural network trained on more than one lakh clinical images has achieved dermatologist-level precision in diagnosing skin malignancies. In a comparison of AI algorithm prediction against 21 dermatologists evaluating a set of photographs and dermoscopy images, the performance of deep learning algorithms was better than that of ordinary dermatologists. With the emergence of deep convolutional neural networks, AI may be useful in detecting prostate cancer from biopsies, identifying breast cancer metastases in lymph nodes, and mitosis in breast cancer [27].

Big Data Analytics Process Using Machine Learning

Modern machines with large computational capabilities can analyze the volumetric data center and use approximate correlation models. Big-data technique has the ability to accurately affect machine learning skills and enable instantaneous decision making to improve overall operational performance and lessen redundant costs [28]. Figure 3.3 shows the process of bigdata analytics in the Medicare sector. Machine learning has its own unique importance in sensing various parameters and correlating them with diseases. It is an effective method in which computers use machine learning algorithms to analyze huge amounts of data presented in a nonlinear manner, categorize patterns, and make verifiable and validated predictions.

Existing data and information from past experiences are predicted during supervised learning. In addition, patterns of unknown effects can be forecasted from the data during unsupervised learning. [29]. Machine learning, deep learning, and cognitive computing are all different steps toward higher levels of AI, but they are not same. Deep learning is a subset of machine learning. It uses artificial neural networks, which simulate human brain connections that are “trained” with the time to answer questions with almost 100% precision [30,31].

Big data in health care can be in an unstructured, semistructured, or structured format [32] and can be obtained from primary sources and secondary sources. Clinical decision support systems, computerized physician order entry (CPOE), electronic patient records (EPRs), etc. are just a few examples of primary sources and insurance companies, laboratories, government sources, health maintenance organizations (HMOs), pharmacies, etc. [33] are just a few examples of secondary sources.

EPRs, image processing, social media, smartphones, and w'eb databases are the core sources of big data in the Medicare sector. The digitization of medical history in the past has provided hospitals with a basis for medical records of patients. EPRs data is received from doctor records, ECGs, lab reports, scans, health sensor devices, X-rays, medical prescriptions, etc. [34]. This dataset is the foundation of personalized medicine and extensive cohort studies for hospitals. In image processing, medical images are also one of the main reasons for data analysis. Photo acoustic imaging, computed tomography (CT), mammography fluoroscopy, molecular imaging magnetic resonance imaging (MRI), ultrasound, positron emission tomography-computed tomography (PETCT), and X-rays are just some examples of well- recognized imaging procedures within medical systems [35].

Twitter, YouTube, Facebook, Instagram, and Linkedln are the most popular social media platforms for collecting health care data. Social media offers health care professionals with advanced tools to share information, discuss policies and practice health issues, interact with the public, promote healthy behaviors, and communicate with patients, colleagues, students, and nursing staff [36-38].

Data from medical social journals is generally used to analyze the spread/trans- mission of diseases. For example, gathering information on Twitter about a person affected by the flu is faster than the traditional method [39]. Applications are the most vital data sources in the field of Medicare self-management. Currently, smartphones have health related applications like the adjustment bit, pedometers tjat produce a lot of step data, the number of steps climbed, and the number of calories burned [40]. Another mood panda application is used to measure individual moods as well as everything related to emotional, mental, and physical aspects as well as social and environmental outlook of everyday life. We also have mobile applications such as the glucose buddy, diabetes connect mysugr, and sugar sense for diabetes management iBGstar (sugar and glucose meter app) to monitor blood sugar. These applications generate large amounts of data 24/7, which contributes to health research. Popular websites that document health data include 23 and Me and uBiome. 23 and Me is a DNA study service that provides evidence and tools for individuals to get to know and research their DNA. uBiome is a microbiome sequencing facility that provides people with evidences and tools to discover their microbiome [41].

The role of storage is very essential and effective in big data. In Medicare, data is growing rapidly, so a streamlined and heavy storage is necessary, such as cloud computing technology [42]. It provides elasticity and competence to access data, create awareness, accelerate the ability for scalable analysis solutions and increase value. This technology is a robust technology for storing large amounts of data and for carrying out comprehensive and complex data processing. This eliminates the need to get expensive computer hardware, software, and dedicated storage space [43]. The need for cloud for big data research in health care is understandable for the following reasons: it may be important to invest in big data research and increase the ability to have an accomplished and affordable infrastructure. Big data in health organizations is a mixture of internal and external bases, as they often store the patient's most sensitive data internally. Gigantic amounts of big data made by public providers and third parties may be outside [44]. Data management in health care comprises cleaning, organizing, retrieval, data governance, and data mining. It also includes methods for checking and verifying scrap data or lost values and such data must be eliminated [45].

Data analytics is a splendid way to transform raw data into information. In the health care industry, big data analytics can be done by five different analytics named as descriptive, predictive, diagnostic, prescriptive, and cognitive. Current research in industry and academia suggests that by integrating big data into analytics, retailers can achieve a return on investment of almost 15% to 20%. Descriptive analysis examines former performance based on historic data. Descriptive analysis is also called unsupervised learning. It briefly describes w'hat is happening in health management and clearly explain the effects of parameters on the system [46,47].

Diagnosis analytics uses historic data to estimate and identify the core reason of the problem. Predictive analytics analyzes both historic and real-time data, also referred as supervised learning. This may merely predict w'hat will happen later, because all forecasting is probabilistic in nature. It cannot predict the future. What does it expect? What are the futuristic trends? Prescriptive analytics automatically integrates big data and suggests before deciding for multiple feasible solutions. This precious information taken and processed by strategic decision maker. This analytic recommends that what we should do. What is the optimum result and how it can be achieved easily [48-50]?

Cognitive analytics is an innate development of data processing and visual analysis. Cognitive analysis takes humans out of the loop and works fully automatically. Cognitive analytics builds on advances in many areas and integrates computer and cognitive science approaches [51]. In cognitive analytics, extensive data come from heterogenous sources and having structured, semistructured, and unstructured data. Moreover, it uses knowledge structures including taxonomy and ontology to facilitate reasoning and inference. For cognitive analytics, it is very important to extract low'-level as well as high-level information. Unlike all other analytics, cognitive analytics makes several responses to a query and allots a certain level of trust for every suggestion [52].

Data visualization provides the results of analyzing health data from numerical to graphical or pictorial form so that complex data is easy to understand and decision making can be improved. It can be used to understand the relationship between the model and the data. IBM Watson analysis, Graphviz R and Cytoscape are visualization tools used in health care. These tools can be used for predictive analytics, advanced analytics, reporting, instantaneous analysis, dashboards, and large data visualization [53,54].

< Prev   CONTENTS   Source   Next >