It is not surprising that India has a booming healthcare sector, considering that a population of 1.35 billion in 2018 is likely to exhibit myriad morbidities (World Bank, 2018). With a three-fold increase in the healthcare market and governmental projections pushing a 372 billion USD mark in 2022, the sector is ready for significant technological interventions. (IBEF, 2019). On the downside however, the sector is beleaguered by concerns that range from access barriers and poor doctor-patient ratio to affordability and poor healthcare infrastructure. Artificial intelligence (AI) comes with a promise of not only overcoming a majority of these barriers, but also eliminating predispositions, such as the recency bias, in medical sciences. Riding the investment wave AI—robotics and Internet of Things (IoTs) can revolutionise healthcare.
Healthcare concerns can be broadly classified into those which are predictive in nature—pre-empting a problem and providing a solution to abrogate the issue; and, prescriptive, where a treatment is offered based on an informed decision. When AI is deployed to prepare algorithms that help map patterns by collecting and analysing gathered data—both spatial and temporal, it is found that it can provide astounding results in preparing a response ahead of time and influencing outcomes. From early detection of diseases based on analysis of past data, to decentralised diagnostic testing, AI can singularly alleviate healthcare problems in rural and remote areas. AI algorithms are able, with a certain degree of precision, to screen diseases, which can help triaging high priority cases, enhancing the productivity of healthcare professionals.
In India, companies such as Artelus, a Bangalore based AI enabled healthcare unit, works to provide an image-based early detection facility for diabetic retinopathy . The Deep Learning and AI based setup in Artelus can help identify at a primary screening, lesions or abnormalities present in fundus images and report it to the doctor, making it impactful for areas that lack this facility. These healthcare capacities extend beyond the boundaries of hospitals and specialised clinics, reduce cost and improve health outcomes. There are many such applications for various anatomical disorders world over, such as IBM Watson for oncology and many private hospitals in India, such as the Manipal group of hospitals, make use of such interfaces.
Interestingly, AI can also screen mental disorders in its early stages, like depression. Wysa, a bot developed by Bangalore based start-up Touchkin, is delving into the domain of emotional wellness. Supported by human coaches, the bot helps cure depressive thoughts (wysa.io).The app records and analyses various physiological factors like sleep patterns, blood sugar levels and other behavioural insights and predicts the user’s mental health. In the event that the bot identifies an individual who needs intensive care, it refers the case to professionals, who can then intervene.
Niramai, a Bengaluru based healthcare start-up has developed a non-invasive, radiation free breast cancer detection software that uses a high resolution thermal sensing device and a cloud hosted analytics solution for screening thermal images that accurately leads to the early detection of tumours. Orbuculum, another start-up in Bengaluru analyses genomic data to predict a gamut of diseases.
Telangana in fact, has adopted a cloud-based analytics tool developed by Microsoft, for the state’s Rashtriya Bal Swasthya Karyakram, to reduce avoidable blindness among children by screening them for the ailment. Thus predictive AI usage in healthcare can provide actionable insights based on available data and thus improve healthcare penetration.
Prescriptive analytics on the other hand, make use of machine learning to determine the best solution or outcome among various courses of action. For instance, an AI algorithm in IBM Watson for oncology will use information from relevant literature to assess the information from a patient’s medical record and throw up potential treatment options ranked by level of confidence. The oncologist can then use the results along with the supporting evidence to arrive at the appropriate treatment option. Such AI interfaces aid human decision making for doctors and health administrators to use critical data to support clinical, financial and operational decisions. It can lower the cost of healthcare, improve patient efficiency and mitigate operational risks.
This technology has found its way into hospitals in India as well. Manipal uses IBM Watson for Oncology. Max Healthcare,India has deployed the GE Healthcare’s web-based radiology information system—the Centricity RIS-IC. Integrated with the GE -picture archive and communication system (PACS), the programme addresses a healthcare unit’s evolving radiology workflow to enable seamless access to images like X-Ray, MRI and more for patients across locations. It can therefore be used to create an integrated customer record of patients. Fortis Hospital, Bengaluru has partnered with Phable, a healthcare start-up in India, to provide an App to the patients that allows for constant monitoring by the doctor in the event any new symptoms emerge and can also help patients manage medication, tests, diet, exercise etc.
The medical equipment industry are also using AI and machine learning to develop smart wearables and insertables that gather individual data and detect anomalies. The US drug major Abott has launched an Insertable Cardiac Monitor (ICM) that can alert users about irregular heartbeats (arrhythmias) on their smartphone screen (cardiovascular.abbott). Ten3T, another Indian healthcare start-up has launched a wearable device named ‘Smart Patch’ that inter alia measures the patients’ temperature, pulse and blood pressure and provides real time monitoring facility (ten3thealth.com). AI also delves into solving problems related to the pharma supply chain with tools streamlining the entire process from drug generation to delivery. Pharmarack, a Pune based start-up has developed a tool to automate the sales and operational processes of pharma companies (pharmarack.com). It offers management solutions from the origin of order to its completion, in a seamless platform to process, track, and settle all orders, creating complete visibility of business operations in real time.
The AI and machine learning has begun to partner with the medical insurance sector as well. Embedded into existing insurance frameworks, the platform helps insurers to automate and expedite the process, minimising delays and frauds. ICICI Lombard and HDFC bank with their AI and Natural Language Processing (NLP) based chat bots named MyRA and SPOK, respectively, are using AI to categorise, prioritise and respond to customer emails and mine appropriate information for an improved operational efficiency (myralabs.com;
Dhawan 2018) .
Ethical Concerns in AI Healthcare
In 2008, Google Flu Trends (GFT), began aggregating and analysing big data from a range of countries based on Google search queries. GFT then went on to predict or ‘nowcast’ the onset on flu outbreaks days before they were reported by the global Centres for Disease Control and Prevention (Lazer and Kennedy, 2015). However, GFT failed to accurately predict the 2009 global swine flu pandemic, as its algorithm over relied on the Google search patterns rather than the traditional reporting of the disease. In 2013, the GFT failed again, missing predictions by 140 per cent at the peak of the flu season. The project was thereafter closed. It is true that big data is competent to model disease spread and identify emergencies, way faster than traditional methods, but the method and the data used becomes critical in identifying a trend—which is why the GFT lost out. AI and machine learning therefore, needs to be understood in the perspective of its own set of challenges. Apart from the data accuracy concerns, a faulty algorithm can distort the results. In a scenario where humans begin to depend on AI for its decisions, such errors can lead to critical drawbacks in healthcare.
AI, if handled improperly can result in data leaks, which would lead to privacy violations. In India the consent forms for data sharing are not mandatorily filled by the healthcare units—and patients too are barely aware of its need. In practice, doctor-patient confidentiality is in the realm of ethics. Therefore, there is always a chance that the profile of patients can be exploited by companies and consequent data breach can lead to an erosion of trust among the general public. For instance, in 2016, a Mumbai based diagnostic laboratory Health Solutions had to remove over 35,000 medical records of patients which included HIV reports, when its data leaked (Indian Express, 2019). Such breaches are a constant threat that AI needs to combat in order to maintain patient confidentiality.
Then of course there is the single language predisposition that AI holds, where English predominates, making it difficult for the technology to penetrate rural areas. With a huge internet connectivity of over 566 million people in 2018 (The Economics Time, 2019), AI can make greater headway if vernacular usage is encouraged.
Importantly, an AI enabled system thrives on data—generation of poor quality data coupled with poor digital infrastructure and storage can skew results, rendering programmes ineffective. In India lack of trained professionals for data handling also impedes the penetration of AI in this sector. Also, the focus of AI in healthcare in India is nascent and fairly narrow, with disease specific solutions.
AI has significant scope in developing solutions in bettering the lives of humans, and healthcare is a priority area of research. Despite the many challenges AI and machine learning exhibit at present, ground is being made for larger and more accurate predictive and prescriptive programmes. An ambient policy framework by the Indian government that makes for favourable investments in the AI and machine learning sector can help augment the paucity of quality healthcare professionals in many locations of the nation