资源描述
Artificial Intelligence in the Healthcare Industry in India REPORT BY Yesha Paul, Elonnai Hickok, Amber Sinha, Udbhav Tiwari ECOSYSTEM MAPPING BY Shweta Mohandas, Sidharth Ray, Elonnai Hickok RESEARCH ASSISTANCE FROM Pranav M Bidare The Centre for Internet and Society, India Designed by Saumyaa Naidu Shared under Creative Commons Attribution 4.0 International licenseContents Executive Summary 1 Methodology 1 Introduction 2 Use of AI in Healthcare 3 Descriptive 3 Predictive 4 Prescriptive 4 State of AI in the Indian Healthcare Industry 5 AI and Healthcare Segments in India 6 Government Initiatives 10 Stakeholders in the AI and Healthcare Ecosystem 13 Ethical, Legal, and Cultural Considerations 15 Policy and Regulatory Landscape in India 19 Challenges to AI in India 25 Recommendations and Way Forward 28 Conclusion 30 Annex 1 AI in Healthcare in India Ecosystem Mapping 321 Executive Summary The use of AI in healthcare in India is increasing with new startups and large ICT companies offering AI solutions for healthcare challenges in the country. Such challenges and solutions include addressing the uneven ratio of skilled doctors to patients and making doctors more efficient at their jobs; the delivery of personalized healthcare and high quality healthcare to rural areas; and training doctors and nurses in complex procedures. 1 Companies are offering a range of solutions including automation of medical diagnosis, automated analysis of medical tests, detection and screening of diseases, wearable sensor based medical devices and monitoring equipment, patient management systems, predictive healthcare diagnosis and disease prevention. 2 In developing these solutions, a commonly cited challenge has been the lack of comprehensive, representative, interoperable, and clean data something that is intended to be addressed through the Electronic Health Records Standards developed by the Ministry of Health and Family Welfare in 2016. 3 Other challenges include access to open medical data sets and adoption by practitioners. 4 This report seeks to map the present state of AI in the healthcare sector in India. In doing so, it explores: Use: What is the present use of AI in healthcare? What is the narrative and discourse around AI and healthcare in India? Actors: Who are the key stakeholders involved in the de v el opment, impl ementation and r egulation of AI in the healthcar e industr y? Impact: What is the potential and existing impact of AI in healthcare? Regulation: What are the challenges faced in policy making around AI in the healthcare industry? Methodology This report explores the state of AI in the healthcare industry in India. From CIS literature review undertaken in December 2017, we learned that there is no single definition of AI. 5 For the purposes of this report, we have drawn upon the definitions outlined in the literature review and reached a broad understanding of AI as a dynamic learning system that can be used in decision making and actioning. To do this, the AI ecosystem in the industry was mapped by identifying AI solutions, practitioners, researchers, funders, government, and conferences/exhibitions. For the mapping, the research draws upon news items, company websites, academic articles, industry reports, interviews, and roundtable inputs to identify different AI solutions being used in each sub-segment of the healthcare industry in India. Search terms used when searching for AI solutions in health include artificial intelligence, machine learning, neural networks, multi-agent systems, innovation in healthcare and fuzzy logic. Search terms used for health include health, healthcare, diagnostics, hospitals, telemedicine, pharmaceuticals, medical equipment and supplies, and health insurance, clinical data. The findings in the report and the mind map represent what we have been able to find untill the date of publishing and is not necessarily comprehensive. 1 How innovations in AI, virtual reality are advancing healthcare in India to new frontiers, J Vignes. Retrieved January 5, 2018, from 2 Artificial Intelligence Based Healthcare Startups in India, Tiash Saha. Retrieved January 5, 2018, from 3 Notification on Electronic Health Standards, 2016. Retrieved January 5, 2018, from mohfw.nic.in/ sites/default/files/17739294021483341357_1.pdf 4 Future of Artificial Intelligence in Healthcare in India, Economic Times Healthworld. Retrieved January 5, 2018, from intelligence-in-healthcare-in-india/56174804 5 Artificial Intelligence: Literature Review (2017, December 16). Retrieved January 5, 2018, from cis-india/internet-governance/blog/artificial-intelligence-literature-review2 To broadly understand the function of AI in healthcare, the research categorized the healthcare industry into segments based on the categorization found in the 2017 IBEF report on healthcare market in India. 6 The research adopts three broad functions derived from an interview, that AI is being used for in the healthcare industry: prescriptive, descriptive, and predictive. 7 To give clarity on the state of AI in the industry, each identified solution in each segment was categorized into one of these three functions. The policy and legal components of the report highlights existing policy, case law, and standards that have implications for the use of AI. The research brings in examples from two different contexts, US and UK, to understand how the regulatory framework around AI in health is evolving and how the Indian regime compares. The challenges to the use of AI in healthcare were identified predominantly through a review of literature, interviews and roundtable inputs. Introduction The emerging use cases of Artificial Intelligence (AI) in the healthcare sector can be seen as a collection of technologies enabling machines to sense, comprehend, act and learn so they can perform administrative and clinical healthcare functions, as well as be used in research and for training purposes. 8 Unlike legacy technologies that only complemented human skills, health AI today can significantly expand the scope of human activity. These technologies 9 include, among others, natural language processing, intelligent agents, computer vision, machine learning, expert systems, chatbots and voice recognition. 10 These technologies can also potentially be used to compensate for a physicians cognitive biases (such as “recency bias,” where one is more likely to allow the last case one treated to inform the course of treatment for the next patient.) 11 This use and adoption of AI can be seen at varying levels across the healthcare ecosystem. Machine learning can be used to address the issue of reporting in siloed Electronic Health Records (EHRs) and instead redirect these reports toward analysis and predictive modelling. 12 This technology can also be applied to preventative health programs. Machine learning can be used to merge an individuals omic (genome, proteome, metabolome, microbiome) data with other data sources such EHRs to predict the likelihood of developing a disease, which can then be addressed through timely interventions such as preventative therapy. 13 6 Advantage India Healthcare Report, IBEF. Retrieved January 5, 2018, from ibef/ download/Healthcare-November-2017.pdf 7 Dr. Agrawal, A. (2017, November 24). Personal Interview 8 Ma Si (2017, April 20), New partnership to leverage AI technology in medical fields, Retrieved January 5, 2018, from 9 Artificial Intelligence: Literature Review (2017, December 16). Retrieved January 5, 2018, from cis-india/internet-governance/blog/artificial-intelligence-literature-review 10 A. Ericson, (2017, October 31), Health AI Mythbusters: Separating Fact from Fiction, Retrieved January 5, 2018, from fiction. 11 K. Safavi, (2016, December 15), The AI Will See You Now, Retrieved January 5, 2018, from 12 burnout 13 R. Eubanks, (2017, October 3), Artificial Intelligence and the Healthcare Ecosystem Part One, Retrieved January 5, 2018, from healthcare-ecosystem-part-one3 AI addresses the issue of information overload often faced in healthcare by employing machine learning to make sense of otherwise overwhelming volumes of healthcare data, which can otherwise threaten the adoption of evidence-based practice. This phenomenon is known as “filter failure”, where the main problem is not too much information, but how such information is analysed. This is exemplified by inadequate information retrieval systems for point-of-care settings, difficulty identifying all relevant evidence in an exceedingly diverse landscape of information resources, and lack of basic health information literacy. 14 Programs such as IBMs Watson for Oncology 15 extensively evaluate medical literature to prescribe the best course of treatment. Researchers have used smart algorithms to extract information from radiology reports contained in a repository spanning multiple institutions. They report that their approach “provides an effective automatic method to annotate and extract clinically significant information from a large collection of free-text radiology reports”; that it could be used to help clinicians better understand these radiology reports and prioritize their review process; that it could link radiology reports to information from other data sources such as electronic health records and the patients genome, and that “extracted information also can facilitate disease surveillance, real-time clinical decision support for the radiologist, and content-based image retrieval.” 16 AI can also prevent recidivism/relapse by helping follow up on cases and making further recommendations. EHRs combined with AI can be used to predict how a patients genetic makeup may affect illness or react to a certain medication. Theoretically, AI can use a persons genome to recommend the most effective treatment option with the least side effects. 17 Use of AI in Healthcare The use of AI in the healthcare industry is diverse across sub-sectors. To understand the type of AI that different solutions are being developed around, the uses of AI in healthcare can be categorized into the following broad categories as 18 : Descriptive Descriptive AI is the most widely used in healthcare technology today, and holds the most promise in terms of short-term potential 19 . It quantifies events that have already occurred and uses this data to gain further insights, such as detecting trends and minor changes that may otherwise escape detection by medical professionals. For instance, such technology can be used to identify patterns in fracture detections and skin lesions. Additionally, these technologies have been shown to outperform humans in detecting subtle wrist fractures. 20 14 Klerings I, Weinhandl AS, Thaler KJ (2015, July 27), Information overload in healthcare: too much of a good thing?, Retrieved January 5, 2018, from ncbi.nlm.nih.gov/pubmed/26354128. 15 IBM Watson for Oncology. Retrieved January 05, 2018, from marketplace/ibm-watson-for-oncology. 16 Hassanpour, S., and private hospitals, which include nursing homes and mid-tier and top-tier private hospitals. From a review of solutions adopted it appears that hospitals in India are employing descriptive and predictive AI. For instance, the Manipal Group of Hospitals has tied up with IBMs Watson for Oncology to aid doctors in the diagnosis and treatment of 7 types of cancer. Watson for Oncology is used across its facilities, where more than 2.00,000 patients receive cancer care each year 41 . Here, AI is used to analyse data and research evidence and improve the quality of the report, in turn increasing patient trust. Importantly, patients are fully aware of the process and provide their express consent. Due care is also taken to preserve patient anonymity. However, at the global level, Watson has recently come under fire from physicians across the world for allegedly posing as “a mechanical turk - a human-driven engine masquerading as an artificial intelligence.” It was reported that instead of using AI, it actually works by convening a small panel of cancer experts, who formulate recommendations for specific patient 36 TCS (2017). Getting Smarter by the Sector: How 13 Global Industries Use Artificial Intelligence. Retrieved January 5, 2018, from 37 Artificial Intelligence: Healthcares New Nervous System. Retrieved January 5, 2018, from 38 J. Vignesh (2017, June 9), How innovations in AI, virtual reality are advancing healthcare in India to new frontiers, Retrieved January 5, 2018, from startups/how-innovations-in-ai-virtual-reality-are-advancing-healthcare-in-india-to-new-frontiers/ articleshow/59060040.cms. 39 N. Bareja (2016, August 18), Impact of Novel Technology on Low Cost Healthcare in India, Retrieved January 5, 2018, from 40 IBEF Healthcare (2017, November), Retrieved January 5, 2018, from ibef/download/ Healthcare-November-2017.pdf 41 Manipal Hospitals, Watson for Oncology Report (2016, September 7), Retrieved January 5, 2018, from 7 profiles - “These recommendations represent the best guesses of these experts, supported by medical literature and personal experience. IBM has never allowed an independent study of Watson for Oncology. No follow up is done to evaluate whether its recommendations help patients.” 42 Foreign physicians have also complained that the population that Watson is trained on does not accurately reflect the diversity of cancer patients across the world, and as a result, it is heavily biased towards American patients and standards of care. 43 Aravind Eye Care Systems is presently working with Google Brain, after previously helping Google develop its retinal screening system by contributing images to train its image parsing algorithms. After successful clinical trials to detect signs of diabetes-related eye disease, it is now attempting to put it to routine use with patients. 44 Products such as Microsoft Azure, Machine Learning, Data Analytics, CRM online and Office 365 are being used by private healthcare providers such as Fortis Healthcare, Apollo Hospitals, L V Prasad Eye Institute (LVPEI), Narayana Health and Max Healthcare to improve patient care. 45 2. Pharmaceuticals: These include manufacturing, extraction, processing, purification and packaging of chemical materials for use as medications for humans or animals. From a review of solutions adopted it appears that pharmaceuticals in India are employing descriptive and predictive AI with prototypes for prescriptive AI being developed and tested. The most common use of AI in pharmaceuticals is in drug discovery, where AI is mobilised to scan through all available literature on a particular molecule for a drug (eg. targeted molecule discovery), which would otherwise be impossible for even a group of people to manually carry out. 46 In addition to streamlining the pro
展开阅读全文