Social sectors constitute an important part of the economy for gauging the direction and speed of its growth. A lot of emphasis has therefore been given to develop these sectors for an all-around development of the economy. Technology has played a crucial role in digital transformation of these sectors with unparalleled outcomes.

An important social sector among these is healthcare which has predominantly received worldwide attention with focus on safe and healthy living using technological inputs. Parallely, issues have arisen about the acceptability of applying AI in healthcare delivery, more particularly in clinical practice. It has been stated that the most powerful use of AI in healthcare is to enhance human capabilities, not replace them.

International and National context:

With the rise in demand on delivering quality and cost effective healthcare, Applied Artificial Intelligence (AAI) has received tremendous attention the world over for specific and goal oriented development programs in the healthcare. Countries are investing large sums and efforts to develop and deploy AI technology based applications for healthcare. US, Japan, China, Russia, UK, as notable among them, have promoted major development programs in the area. Research on applying AI to healthcare has been an area of continuous focus.

Some statistics to reckon with the global AI healthcare market are;

• AI healthcare market is expected to be US$ 13 billion by 2025 (Frost & Sullivan) and grow to US$ 36 billion by 2030 with CAGR of 24.4 %. As per another estimate, this market revenue is expected to grow to US$ 35 billion by 2025, and to US$ 51 billion by 2027 with a CAGR of 41.4 %.
•  The AI implementation in healthcare raises the potential to improve healthcare outcomes by 30-40 %.
• For the Venture Capital investments, AI healthcare has been among the top - 50 of the top Companies have received funding of US$ 8.5 billion (McKinsey).
• By 2035 there will be global deficit of 12.9 million healthcare professionals, the impact of which will be reduced with AI implementations across the sector and availability of AI skilled workforce.
• Concerns have been on the waste in healthcare systems – in US alone it is US $ 750 billion annually due to long diagnostic time and patient treatment wait. This is expected to be minimized with AI implementations resulting in a significant savings in healthcare spends.

So what are the growth drivers for AI in healthcare?

• Growing need for patient centric care that will be enabled by AI technology and deep learning with the effective use of medical data.
• Growing applications of AI in healthcare as a result of extensive research and successful pilots to find solutions to the problems in healthcare. 
• Shortage of healthcare work force that will be overcome with automation of patient diagnostics and management of back end work flows.
• Increasing complexity in clinical decision making that will be handled with the use of AI assisted diagnostics.
• Progress in health care informatics for management of back office administrative functions and management of a vast amount of medical data, mostly unstructured.

However, the AI healthcare market restraints are:

•  Lack of clearly defined regulations by the Administrations in use of results from AI implementations.
• Data security and data privacy concerns as the AI implementations require extensive use of patient data across platforms.
• Barriers to Physicians’ adoption of AI applications due to lack of exposure to AI tools and their utility and the technical infrastructure.
• Legal implications of use of AI systems are not well laid out.

Nevertheless, the market opportunities of AI healthcare are:

• Capitalizing on unmet demand from an emerging gigantic healthcare sector.
• Enabling patient care at home through remote monitoring- Teledoc and Telemedicine.
•  Leveraging rapid developments in effective use of AI tools - deep learning, data analytics, natural language processing etc
• Unlocking combined potential of AI and Robotics for a complete AI driven solutioning of health care problems hitherto intractable.
•  New concepts of Digital Twins enabling more accurate and faster diagnostics for patient care through the applications of AI.

Not lagging behind, India has instituted a national program on AI, identified as a key enabler for economic development, and a specific program in healthcare, called ‘Ayushman Bharat’, which derives significant benefits from the use of technologies. Healthcare, as a gigantic sector, has accordingly been allocated a significantly risen investment of 2.7% as a percentage of its GDP. Industries and development organizations are working tirelessly to build and deploy AI based solutions for the Indian healthcare sector.

Areas of AI intervention in healthcare:

In healthcare, it is the enormity of data that makes it possible for AI based computer aided diagnostics and predictive analytics for a better healthcare delivery. Also, it makes it possible to affect enormous savings in healthcare spend by helping front line clinicians to be more productive and automating back end processes to be more efficient. Acceptance of AI in clinical practice is still an issue, and is the reason why it is mostly confined to aid/augment healthcare delivery. This is the area where  much of research is needed.                                              

The AI applications to healthcare range from Robot assisted surgery to virtual nursing assistant, admin work flows, fraud detection, dosage error reduction, clinical trial participation, preliminary diagnostics, automated image diagnosis, cyber security etc. Thus, following are the ways AI is predicted change the healthcare scene globally going into the future.

•  In its applications, AI will access multiple sources of data to reveal patterns and therefrom aid in patient centric treatment and care to be more accurate and fast.
•  AI healthcare systems will be able to predict an individual’s risk of certain diseases and suggest preventive measures.
• AI will help reduce waiting time for patients and improve efficiency in overall healthcare delivery thereby affecting significant savings in healthcare spend.

Derived from this, the areas of AI intervention in healthcare can be mainly grouped under the following categories;

• Patient Care: covers automatic diagnosis and prescription, personalized medication, patient data analysis, real time case prioritization.
• Medical Imaging & Diagnostics: covers diagnostic error prevention, medical  imaging insights, speedy and accurate diagnosis.
• Management of back office administration: covers appointments, insurance claims, settlements, fraud detection, patient data record tracking, voice to text transcriptions and the like.
• Research & development: covers gene analytics, drug discovery in the discipline of computational biology using high performance computing,  compilers and libraries.

Research & Development as a key activity:

There has been a felt need for technological inputs into the extremely important sector of healthcare that concerns the Society as a whole. While there have been several developments and the Industry are seized with building world class applications to service the gigantic healthcare sector, there is a felt need for a continuing thrust in the sector by conduct of quality research to help solve some of the demanding problems in the sector.

The researches are also to   lead to knowledge building for public good, create intellectual property and usefully commercialize and build economic value therefrom for the Country. As a spin off, skills in the areas, for which there is a woeful shortage, are generated that are needed in numbers across the length and breadth of the Country, for development and deployment of technology driven solutions in healthcare.

The area involves multidisciplinary fields of science and technology with experience and knowledge in applied mathematics, statistical techniques, cognitive science, data science, advanced computing and software which are useful in conceptualizing and developing AI applications and putting them for practical use.

Technology components of AI in use for healthcare:

The technology components within the realm of AI that are predominantly used for research and in building the applications in healthcare are;

Machine learning, Data Analytics, Natural Language Processing, Computer Vision, Deep learning, High Performance Computing and Cloud.

• Machine learning (ML) is an integral part of AI technology implementation. This refers to the analysis of data sets to study any patterns within it, and using this information make any prediction. In this way we unearth hidden values or knowledge within data sets that helps to carry out predictive analysis using the learnings by machine.
• Data Analytics refers to the use of an algorithm to extract knowledge and insights from the data that is available mostly in an unstructured form. Companies working on such Analytics monetize the Data to their advantage and to service their Customers more efficiently.
•  Natural language processing (NLP) is a scientific process that is computationally implementable for a host of applications such as text mining for information retrieval, machine translation from one language to another, speech recognition in various languages, or entity recognition. It involves automatic manipulation of natural language, text or speech or both, using software for extracting useful information from the data, integrating machine learning based approaches with scalable processing. Sifting through the unstructured data, NLP helps to identify key ideas, uncover trends, analyze sentiments, and identify correlations between words.
• Computer vision is a scientific field that deals with creating software to enable a computer to gain high level understanding from digital images or video clips. In the process it seeks to automate tasks that normally a human visual system has to do tirelessly. And the computer system gains insight into the visuals to unearth valuable information that is generally not visible to the naked human eye.
• To serve major AI technology based implementations, a high performance computing environment is now a reality with the availability of hundreds of Petaflops of computing powers and very low, of the order of nano seconds,  inter-processor communication latencies. These high performance computers provide a platform for modelling, simulation, cloud, data analytics and graphic virtualization etc.
• Deep learning, as a broader family of machine learning methods, is based on artificial neural networks where we could uncover every layer of understanding the data. It is used to perform tasks such as recognizing speech, identifying objects and making predictions. Very high level of computing powers are needed for deep learning applications.
• Cloud has enabled accessing of high power computing resources, data and software tools and application modules by multiple teams without geographical restrictions to enable collaborative research and offer services.

Research areas for use of AI in healthcare:

Below we present some of the key areas of research seeking solutions to diverse requirements in healthcare with the implementation of AI.

- Development of an AI and NLP powered Chatbot model providing prescriptive diagnosis and Patient queried information on a two way interactive platform.

Chatbot is an interactive web platform that offers on line services aiding  medical diagnostics and access to Patient data and a host of other healthcare related information. Powered by AI and NLP, it aids such functions as providing  informational support, assistance in pharmacy, provide medical assistance,  collect patient data, schedule appointments and physician consultations. Development of an efficient, powerful and user friendly model for use by the diverse cross section of the society is a continuing need.

- Medical diagnostics covering processing and analysis of Images (MRI, CT Scans and X Rays) to pin point disease type and location, by correlating with the available researched data from such images so as to minimize time in the analyses and reduce errors in Physician intervention using AI algorithms.

AI tools are applied to relate the symptoms and patient’s health parameters with a data pattern to pin point the cause of disease and prescribe a treatment. The diagnostics cover analysis of MRI or CT scans for disease pattern and  correlation with the researched data from such scanned images, so as to minimize time in such an analysis and improve accuracy and speed of Physician intervention. Developing a predictive model of such a diagnostics is an area of continuing research.

- Develop a video analytical model by using video feeds from a given geographically marked area to determine faces, symptoms, social contact and mask usage etc. that do not conform to the laid down parameters.

One of the interventions required towards public health is monitoring of the locations and conditions of the infected persons form certain diseases in real time in a geographically marked area. For this, it is important to have real time video surveillance in busy and crowded areas to determine AI enabled face detection, symptom detection, and monitoring social contacts, mask usage etc., and raise alerts as required from the given hot spot zone and refer to the Administration and hospital services providers from the given geography. Researches are being conducted for a powerful video analytics software that would make this possible.

- Study of genomic data for determining molecular structures that relate to neurological disorders or other infectious diseases is an area of intense research. For this, high performance computing with a host of performance libraries, compilers and a GPU computing environment, hosting deep learning training & development tools and Application framework and libraries are needed.

Our understanding of a host of diseases and infections, neurological disorders etc. through molecular modelling and discovery of new drugs for precision treatments are made better with the help of high performance computing. With the evolution of Genomic sequencing, the available data on human genomes is so huge that it has made possible to analyze the data and determine trends, and discover the drugs molecular structures specific to the particular disease. In the Indian context, an enormity and diversity of data is of particular significance and advantage.

- Emergence of new Medical Technology leading to the development of smart medical devices with the use of AI, device miniaturization, wireless connectivity that  tap into the enormous amount of data to help improve the health of the Patients with chronic health conditions.

The development of these smart medical devices is enabled by a combination of miniaturization, powerful computing, advanced computational modeling,and sophisticated data analytics techniques for programming AI algorithms. Fueling such smart medical device development is the concept of the “Digital twins,” which is a computer based model of a physical device and the Patient. Such a modelling helps to see how certain devices work on certain patients with specific medical conditions. Parallely, the interaction between the real device and the real Patient generates more data, which further helps to refine the device’s performance with the resultant improved and speedy healthcare services. In this process, medical devices “learn” and adapt their performance to the needs of a specific patients that will contribute to amore tailored, individualized response to healthcare needs.

Concept of Digital twins has been used in Aerospace and Defense, Oil and Gas, Manufacturing etc to improve the performance of the machines and productivity while cutting costs and reducing development cycles. Its application to healthcare is of a recent origin. Researches are being carried out dealing with the many variables of a biological system in a modeled environment of a virtual patient and a virtual device using computer simulations and the data with a view to reducing from many choices to the most promising ones, thereby reducing the time to diagnose and improving the accuracy of treatment.

• For the treatment of Diabetics, researches are being carried out using data on diet, exercise and blood glucose patterns of a large number of patients, and building a model of metabolism and design an algorithm that will enable an Insulin pump to automatically administer an accurate amount of insulin to keep the glucose levels of a patient within a healthy range. In this process the algorithms can be refined using the virtual patient and the data available which, in turn, will enable train the insulin pump for correct dose administration on patients.
• Research to build a computer model that simulates many variations in the heart beats and blood pressure anomalies of the patients that will help to design the Pacemakers. By creating the virtual patient’s heart and using the data on the electrical pulses of a large number of patients’ heart, the algorithms can be designed that deal with a variety of situations of heart functioning to design and program the pacemaker device suiting the particular patient.
• Development of an Image analysis algorithm to analyze anomalies in the images collected from patients’ small intestines in correlation with the available data on ulcer, cancer or other inflammations, and training the algorithm to accurately pin point the area for treatment.
• Use of such an algorithm can be extended for the analysis of other images, such as MRI or CT Scan of other parts of patient body, in correlation with the data on anomalies found in patients’ internal body parts, notably lungs and stomach, thereby savings on time for analysis by the Radiologists and be more accurate in findings.
• To aid organ transplant, it is important to administer correct amount of dose, as there is a possibility of dose errors leading to complications. Research to develop math formula with help of AI to determine correct dose to administer or reduce dosage errors taking individual patients data into account are underway.
• AI assisted Robotic surgery helps to analyze data from pre-operative records to physically guide surgeon’s instrument on Orthopedic surgeries. A lot of research is taking place to refine AI algorithms for accurate instrument guidance.
• Many developments have taken place using healthcare informatics applications such as AI powered virtual nursing assistant to help improve efficiency of their workflows to render speedy medical attention, managing expensive back office problems which consume half of nurses’ time and workload, making available voice- to- text transcriptions, using computerized reviews of medical claims and AI based neural networks to search medical claims for patterns associated with any frauds, monitor and detect abnormality in interaction with proprietary data using AI to ensure data security etc.

Acceptance of AI in healthcare and outlook:

Recognizing the immense potential AI has to transform healthcare delivery system in terms of better quality and cost savings to both the providers and patients, major investment of financial and intellectual resources are being made worldwide. Developments that have gone in this sector are ample evidence to suggest and justify more researches also taking place in the area.

While AI has been accepted in the healthcare industry for aiding and augmenting  improved patient care and diagnostics, its acceptance in clinical practice is still an issue. The examples of applying AI for clinical decisions and clinical outcomes are therefore rare. There is also an ethics and social issue of using AI that is mostly arising out of use of patient data, and in AI being used to influence clinical decisions or even diagnostics.

Challenge for the researchers and the industry is to ensure AI applications are  developed and deployed in a way that is transparent and compatible with public interest and acceptance.

It can be stated that AI’s most powerful use is to enhance human capabilities, not replace them, in dealing with healthcare delivery.