Authors: Avik Sarkar and Poorva Singh

Artificial Intelligence (AI) aims to create machines or computer programs to perform tasks that typically require human intelligence. Along with the advancement in computer algorithms and computational capabilities, the availability of large volumes of data in digital format is essential for success in AI. These technological advancements in the last decade have led to AI creating applications globally across sectors like banking, insurance, education, entertainment, healthcare, etc. Access to quality healthcare is a challenge in India, with less than recommended doctor-to-population ratio and a significant shortage of modern medical partitioners. The situation aggravates further in rural areas of India, where seventy percent of the population resides but has less than thirty percent of trained healthcare professionals, which includes physicians and nurses. The ongoing research at the ISB Institute of Data Science (IIDS) is looking into the current progress of healthcare AI in India and factors that would enhance higher adoption.

Let us look at the key stakeholders, actors, and processes in healthcare where one may apply AI. Further, we have patients consult a physician in anticipation of a disease or upon detection of the disease. Diagnostics tests help in the detection and diagnosis of diseases. The doctor would treat the patients through medication, procedures, or surgery in acute cases. As people grow older, multiple health issues arise, or patients might be detected with terminal illnesses which cannot be cured entirely, and here the role of the healthcare professional is to provide the best healthcare support to the patient. The hospitals or clinics manage many processes that help deliver healthcare services to patients, like managing appointments, scheduling critical resources like operation theatres, scheduling shifts of healthcare staff, ensuring replenishment of essential medicines, critical asset management, etc. There are many diseases for which effective medicines do not exist, and there is always room for improvement in the efficacy and safety of the existing drugs – drug discovery is the process of identifying new compounds or molecules that can be developed into effective drugs to treat diseases.  A public health system in place would periodically check for disease trends and symptoms in the population across different ages and social groups to improve the overall health of citizens and the associated healthcare delivery system. Our research identifies these six broad areas which require human intelligence and are amenable to the application of AI in healthcare.

  • Diagnosis and detection of disease
  • Effective treatment of diseases
  • Geriatric (old age) and palliative (terminal illness) care
  • Drug discovery, pharmaceutical research & development
  • Hospital and healthcare ecosystem management
  • Public Health and Pandemic

Early and correct disease diagnosis helps institute effective treatment leading to better and faster recovery and mitigating disease transmission. Often a disease diagnosis requires years of experience, and young doctors entering the system may need to be trained more to diagnose complex health conditions. Clinical time is limited due to the increasing workload of the doctors and changing disease dynamics over time, leading to challenges in the diagnosis process. We see a substantial advancement in disease diagnosis through AI in a few areas like diabetic retinopathy, cancer, and cardiology. Diabetic retinopathy (DR), a morbid microvascular complication of diabetes, affects one in three people with diabetes and is one of the leading causes of preventable blindness. There are multiple US Food and Drug Administration (FDA) approved systems for DR detection by non-ophthalmic healthcare professionals like IDx-DR and EyeArt. India is home to over 72 million people with diabetes with an estimated DR prevalence of 18%, making routine retinal examination essential to early detection of DR. Automated DR applications need thorough testing and prior approval from the medical regulatory body in the country before being administered to the patients – several AI-based interventions in this area have shown promising results. A team of doctors at L V Prasad Eye Institute (LVPEI), Hyderabad, has developed a low-cost method to capture good-quality retinal images for diagnosing diabetic retinopathy, enabling the images from the rural areas to be sent to the urban areas for review by ophthalmologists. Smartphone-based AI algorithm (Medios AI) captures retinal images of patients visiting dispensaries to show good sensitivity and specificity results for a small sample case.

Diagnosis of cancer requires screening tests carried out on patients with no symptoms or diagnostics tests based on biopsy or imaging for people with symptoms and monitoring tests on people diagnosed with cancer to gauge their response to the ongoing treatment. AI methods have been used to detect lung, brain, and breast cancer, based on the analysis of patient reports of imaging studies such as X-ray, ultrasound, MRI, and CT scan, using a diverse range of machine learning techniques providing good accuracy. In India, researchers at Advanced Study in Science and Technology (IASST), Guwahati, have developed an AI-based method for prognostication of breast cancer based on the evaluation of hormone status. India-based Niramai Health Analytix received US FDA clearance for their ‘SMILE-100’ system, which uses thermal imaging for breast cancer screening. There is a vast prevalence of cancer in India. Still, the lack of large-scale cancer biobank data is one of the leading factors for not having more India-specific AI research for cancer detection.

Early detection of heart diseases can prevent many deaths. Some genetic factors can contribute to its development, but the disease is mainly attributable to poor lifestyle habits like a high-cholesterol diet, lack of regular exercise, tobacco smoking, alcohol or drug abuse, and high-stress levels. AI methods have shown good accuracy in predicting congestive heart failure based on chest radiographs. ECG (electrocardiogram), the most widely used diagnostic tool for cardiology, was used to train AI algorithms to differentiate between acceptable and unacceptable ECG image quality and provide real-time feedback on the patient’s heart condition, which the cardiologist can later use for final patient assessment and evaluation. EchoMD AutoEF tool uses an AI algorithm that automates the clip selection and ejection fraction calculation from cardiovascular imaging, thus assisting cardiologists by reducing the time taken for their decision-making process received FDA approval in 2018. In 2021, Apollo Hospital, based in Hyderabad, launched an AI-based tool to predict the risk of cardiovascular disease and classify patients as high, moderate, or minimal risk. Doctors can use this risk assessment tool for proactive and preventive care to individuals with significant cardiac risk, thus reducing the future burden on the healthcare system and improving the quality of lives of citizens. 

The standard and robust approach to medical treatment is evidence-based medicine, which is based on leveraging thorough clinical research-backed evidence for treating patients. A step forward is precision or personalized medicine, which entails using cellular and/or genetic biomarkers to decide the optimum treatment plan for patients and has already been extensively applied in the field of oncology, being the treatment of choice for many cancer subtypes. AI is changing the surgery practice with technological advancements in imaging, navigation, and robotic intervention. Methods for diagnosis and treatment of cancer are evolving every day. These findings are shared in different scientific journals. It is impossible for practicing doctors in their busy schedules to review these research materials to understand the latest trends thoroughly. IBM Watson for Oncology is an AI-based “Expert Advisor” that assists doctors and oncologists by increasing their treatment and diagnosis capacity by going through the latest healthcare research and reviewing multiple patient records. Bangalore-based health-tech start-up Oncostem Diagnostics uses AI on genomics-based data for personalized breast cancer therapy and chances of cancer recurrence based on the patient’s condition.

With the life expectancy in India steadily rising and reaching 70 years in 2020, there is a growing need to manage chronic illness and make the last years of patients with dementia, heart failure, and osteoporosis more comfortable. Artificial intelligence in healthcare promotes and transforms traditional elderly care services through deep integration with elderly care delivery. A range of AI solutions are helping in the care of elderly people. The simplest ones are AI chatbots that help patients keep on top of care plans, reminding elderly people about their doctor’s appointments, when to take medication, or when to eat. With the decline in physical function, nursing care has always been the most critical aspect of elderly care services and an area that requires immense manpower input. Intelligent nursing robots have humanoid features, move around the house, collect data through sensors and cameras, and relay the information to the cloud. These nursing robots can take emergency action or call a person when required for help.  AI-based robots also help the elderly with household chores like housekeeping and cleaning.

Hospital and healthcare ecosystem management is one area where we see multiple interventions in the Indian context. Given this plethora of choices patients often find it challenging to choose a reputed doctor or diagnosis center in their vicinity who would charge an appropriate amount. The Indian healthcare start-up Practo AI technology for finding the appropriate doctor in the patient’s vicinity based on various criteria like doctor’s specialty, years of experience, fees charged, etc., allows the patient to book an appointment with the doctor and further introduces the teleconsultation facility over time. India lacks a common system for sharing healthcare records in digital format, which patients and doctors can easily access in an emergency. The Indian healthcare start-up DocTalk provides a subscription-based service for patients to store their healthcare records and prescriptions in a digital format, thus allowing them to chat with the doctors based on the stored health records. Based on the success in e-commerce, several healthcare start-ups like 1mg, Netmeds, Pharmeasy, Medlife, TABLT, etc. have identified this opportunity and launched mobile applications through which citizens can order the medicine by uploading their prescriptions and getting them delivered to their doorsteps.

In drug discovery, the recent revolution in human genomics, proteomics and metabolomics has generated an immense amount of biological data, which can become the substrate for AI-based algorithms to give insights on prospective targets for the ever-increasing number of diseases. AI drug discovery platforms have accelerated drug discovery & development from a multi-year timeline to a matter of months and design novel drugs that can have the desired clinical effect & safety profile. A paradigm shift is happening in treating illnesses from managing diseases to curing them altogether. India has not witnessed much growth of AI in drug discovery research. Indian tech-based companies have collaborated with pharmaceutical companies abroad to provide AI and analytical capabilities to them. Still, there have not been examples where Indian pharmaceutical companies have leveraged AI in drug development.

Public health is the application of medicine to improve people's overall wellness and prevent disease onset. AI can aid significantly to this field by identifying risk factors and high-risk groups susceptible to an illness and helping them by dedicating limited public resources to the right population cohorts. In the Indian context, AI has not been widely used in the public health context. Still, there lies an enormous opportunity to apply AI for disease surveillance, health education, monitoring & evaluation, workforce development, public health research, development of public health policy, and enforcing public health laws and regulations.

Our ongoing research found several areas where AI can be applied in healthcare, but only a few have gained prominence in India. Based on our study, we identified some of the underlying reasons for the same. There needs to be more IT infrastructure in the healthcare domain, so the interaction between the various stakeholders can be captured in digital format. India needs a standardized format for electronic health records (EHR), which helps collate or porting of health records between various healthcare institutions. There is a lack of large-scale healthcare datasets of Indian patients available for AI research. There needs to be clear regulations relating to the privacy and security of healthcare records. AI applications often have inaccuracies at the start, which improves over time. A range of ethical aspects arise related to applying AI-based recommendations to the patients, or the recommendations must always be routed via a trained physician. Like the FDA in the USA, there need to be explicit safety norms on what sort of applications can be administered to patients. Our ongoing research focuses on some aspects like healthcare AI adoption in India, challenges in healthcare AI adoption in India, privacy issues related to patient data and regulations enabling the higher adoption of healthcare AI in India.