Artificial Intelligence in Healthcare
- As the Artificial Intelligence revolution permeates through societies at a global level, its role in shaping India with its sixth of the world’s population, could be substantial. India’s ratio 0.8 doctors per one thousand head of population (UK: 2.8, Australia: 5, China: ~ 4), the inability to meet World Health Organisation (WHO) guidelines for ratio of skilled healthcare workers (Weber, 2019), and resulting average patient-to-doctor face-to-face contact of just two minutes; illustrates the challenges of extremely heavy workloads on Indian doctors and opportunities for AIbased solutions to make a difference.
- Healthcare systems in many developing countries are bursting at the seams with as much as 77% of a doctor’s time spent on preventive services that could be safely delegated to non-clinicians. With the ubiquitous reach of mobile technology within rural areas, opportunities exist for AI to help in the achievement of good health and well-being within remote communities where access to healthcare and skilled medical professionals are in short supply.
Opportunities and Applications:
The modern era of global connectivity and high levels of mobile usage in India presents significant opportunities for access to AI technology focused healthcare within the following areas:
AI in Assistance to Physicians:
- One of the ways in which AI can enhance healthcare delivery in India is to relieve highly-skilled medical professionals from routine activities, freeing up doctors to concentrate on the higher-value cognitive application of medical practice, truly connect with patients and positively impact cases of medical errors and misdiagnosis.
- Given the resource constraints and stress on the healthcare system, a significant part of a doctor’s workload could be safely offloaded to carefully-designed AI systems, reserving the serious cases for more detailed physician’s attention.
- AI-based technologies can offer improvements with speedy diagnosis and therapy selection, reducing medical errors, improving productivity, assessing and modeling risk and stratifying disease.
- Researchers have highlighted the success for AI in healthcare using Machine Learning (ML) image-interpretation methods within radiology, pathology, dermatology, using AI in ophthalmology, diagnosis of atrial fibrillation in cardiology, identifying the best available treatment in oncology and interpreting subtle cues from online communications within mental health with greater efficiency over human medical practitioners.
AI in Diagnostics:
- One of the key healthcare challenges in India is acute shortage of radiologists. AI based diagnosis can be especially helpful for radiology, pathology, skin diseases, and ophthalmology. For example, Aravind Eye Care Systems and Sankara Nethralaya have developed and validated an AI-based algorithm for diabetic retinopathy, which assists the ophthalmologists in screening for diabetic retinopathy on the basis of images of retina set to the doctor from peripheral centres. While CT scan, MRI and X-ray facilities have proliferated in India, there are only about 10,000 radiologists available. This is where AI can be of great assistance. The Tamil Nadu e-Governance Agency is helping the health department with the shortage of radiologists by developing an AI based system to read CT brain scans and grade them for further interventions.
AI for Optimising Treatment Plans:
- AI can also be used for assisting doctors and patients to choose an optimal treatment protocol. ML can be used to mine not only doctor’s notes and patient’s lab reports, but also link to the extant medical literature to provide optimal treatment options. (Wahl et al., 2018). Such technology is in use in India, China and Thailand to provide appropriate recommendation plans for cancer treatment using patient’s details linked to medical literature.
AI for Monitoring/Ensuring Compliance:
- The potential for AI application in remote monitoring has enhanced manifolds via the use of wearables. These can be used for monitoring various aspects such as movements, physiological parameters, temperature and alerts that can be communicated to healthcare professionals. Devices can be used for helping people exercise and adopt healthy eating. While these aspects have largely been used for chronic disease management (diabetes, stroke, epilepsy) and for elderly people, specific aspects can also be designed for monitoring during epidemics.
AI in the COVID-19 Epidemic:
- The COVID-19 epidemic highlights the need for an AI based epidemic monitoring system that can model and predict outbreaks and help optimise scarce resources. Researchers from Imperial College, London have identified scenarios of up to 40 million deaths in 2020 from COVID-19 if measures are not taken to address the pandemic but highlight that over 38 million lives could be saved if countries across the globe implement high levels of testing, enforced isolation and wider social distancing.
- AI can help fight the virus via Machine Learning-based applications including population screening, notifications of when to seek medical help and tracking how infection spreads across swathes of the population. A Chinese tech firm uses AI systems to flag anyone who has a temperature above 37.3 degrees within Beijing’s Qinghe Railway Station using cameras equipped with computer vision and infrared sensors to predict people’s temperatures. The system can screen up to 200 people per minute and detect their temperature within a range of 0.5 degrees Celsius. AI was also used for tracking individuals in China and contacts2 by combining face recognition technology, GPS tracking and a network of cameras covering the public places.
Challenges and Controversies:
- The major challenges for India to deliver the benefits to its citizens from the adoption of AI technology within healthcare are significant. Leveraging AI in a meaningful way to enhance healthcare in India; needs emphasis across the healthcare industry to address technological, socio cultural, regulatory, legal and ethical issues.
Healthcare Industry Issues:
- Due to the nature of the industry as well as people dynamics, the healthcare industry has been slow to adopt technological innovations. The challenges of migrating to an AI-technology-based healthcare infrastructure are numerous as medical professionals attempt to transition to new ways of working and adopt new systems and processes. Traditional healthcare personnel may resist new innovations, doctors may not trust AI systems, patients may question AI-based decision-making and medical staff could view the changes as disenfranchising them from their key roles and decision-making powers.
- The changes required to realise the benefits of AI systems must be centred around clinicians and the problems they face, to enhance, not replace the need for highly-skilled medical practitioners. The required transformation to an AI-centric healthcare system requires trust from medical professionals, but also from patients unaccustomed to new ways of diagnosis and decision-making. The key challenge for policy makers is the engendering of confidence in the outcomes and trust that a human medical practitioner has an active role within the AI system.
- The future role of doctors and other medical professionals is likely to change within an era where AI is integrated within diagnosis and disease forecasting. The challenge for the training of doctors is to address the transformational nature of AI-based healthcare, whilst not elongating the period for learning and qualification to integrate these new systems alongside everyday working practices.
- AI systems and the underlying algorithms are reliant on the quality of data to enable the ML elements to perform the necessary processing and decision-making. The challenge within India is the disparate nature of healthcare related data. Each state has its own system and working process. Initiatives are needed at state and national government levels to ensure shared data standards, data security and exchange processes are incorporated within healthcare systems. This is complicated by the massworker migration between states, but highlights the need for solutions at a national level.
Socio-cultural Issues in Technology Implementation:
- Although India is seeing significant development and positive societal change over the last decade or so, the country has a long road ahead in the context of nationwide technological development and adoption.
- Although policy makers have tended to view successful ideas from other countries and naturally assume these can be transplanted to India, researchers have warned of the inefficiency, even danger of such an approach. Studies have advocated that decisions are made to take account of cultural context and existing social conditions.
- Within India, access to the internet is primarily undertaken via mobile phones. While the penetration of mobile phones would at face value seem to be a positive factor for the adoption of AI, it could inadvertently amplify the gender disadvantage.
- Research highlights that women in South Asia are 38% less likely to own a mobile phone than men and when overlaid with patriarchal and misogynistic social factors, the real access figure could be less. Without positive action from policy makers the resulting outcomes for AI adoption are likely to become segmented along gender lines.
Regulatory and Ethical issues:
- There are several ethical and regulatory challenges in implementation of AI in healthcare in India. Data security and privacy is especially important with the increasing use of wearables which can potentially cause identity theft through hacking of devices and data.
- AI is set to alter the traditional relationship between the doctor and the patient as technology plays the role of a third substantial actor. Under these circumstances, the regulators need to provide clear and concise user agreement and privacy policies to enhance widespread and safe adoption of these devices.
- To enhance the adoption of technology by healthcare providers, AI and its applications should be incorporated within the curriculum for medical and paramedical training.
- Technology should be recognised as socio-culturally embedded; hence the technology design and implementation should take into account cultural practices and address the gender divide in India.
- Ethical guidelines regarding security and privacy of data should be protected, especially as more and more the data is available through wearables and IOT. The data should be strictly used for clinical purposes only. AI systems when used for healthcare would have to be tested against all 7 DEEP-MAX parameters.
- The AI system must be explainable and auditable. All decisions made in the context of diagnosis or recommendations can impact on human lives. As such the underlying algorithms must be transparent and explainable to ensure ease of audit rather than acting as a black-box based system.
- AI systems should not exhibit bias. The algorithms developed for the AI system must not exhibit any racial, gender or Pincode-based decision-making that disenfranchise or favour any population groups.
- AI healthcare systems must conform to human values and ethics. Regulatory bodies must ensure that human ethical values are an integral element of AI algorithms and resulting decision-making.
- Adoption of AI based healthcare must be benefits-driven. The migration toward greater levels of technology use may not be universally accepted or trusted by the medical staff within healthcare institutions. The impact and change in working practices must not be underestimated by policy makers, who need to ensure that changes are geared to the benefits to patients and the overall healthcare of the Indian people.
- Pilot initiatives should be developed within key states to trial the impact that AI systems could have on existing healthcare systems and infrastructure. Lessons should be learned from these initiatives before, wider rollout at a national level
Go Back to Main Page