AI is a sector that was on an exponential rise even before the COVID outbreak, however since the beginning of the year its growth has further accelerated.

The present pandemic has been unprecedented in history in terms of the number of people affected by it, directly and indirectly. The communities have gradually started adapting to this new normal even as the second, and at some places, the third wave, of COVID-19 infections are being reported. It is clear that, as we attempt to find a balance between health and economic priorities, life will not be business as usual, at least not yet. Healthcare has suddenly come back into round table discussions for policy and practice, more to ascertain that a situation of this magnitude that paralyzed entire global societies and economies overnight, can be averted in future.

To have a sharper outlook on existing and emerging infections and diseases, it becomes imperative that data-driven intelligence complements human intelligence such that outbreaks can be better predicted, and their spread can be better managed and controlled. Additionally, diseases need to be monitored, managed, and treated more efficiently remotely to make the best use of limited resources while protecting caregivers from risk of infections, wherever possible.

Healthcare is a global priority, with a wide spectrum of diseases, acute and chronic, that needs to be monitored for hundreds of parameters for millions of people at any given point of time. To make informed choices, decipher emerging trends, plan public health interventions, and ascertain optimal treatment regimens from this humongous quantum of data will require real-time analytics and interpretation that can be used to make critical time-sensitive decisions. This is where artificial intelligence (AI)comes in to complement and simulate human intelligence.

AI is a sector that was on an exponential rise even before the COVID outbreak, however since the beginning of the year its growth has further accelerated. Although the application of AI in healthcare has so far been dominated by fitness trackers for wellness monitoring, health notifications and personalized recommendations, there are avenues in which AI has started impacting healthcare which arelesser known. AI techniques such as machine learning (ML), optimization and natural language processing (NLP) have been useful in providing critical insights and predictions on the spread of the coronavirus and the effectiveness and impact of countermeasures. Additionally, as many other healthcare services were disrupted or severely compromised due to the pandemic, the need for technological advancements, like AI, to sustain healthcare delivery in an already challenging ecosystem becomes imperative.

AI has already been deployed to assess chest CT scans, MRI scans and X-Rays to screen for potential SARS-Cov-2 infection in patients, especially in early stages of the disease where the scans might appear normal on visual inspection.

AI for surveillance

Bringing together AI/ML and geographic information systems (GIS) has created GeoAI which has an emerging role in health and healthcare at population and individual levels. Factors such as the environment, environmental exposures, and social determinants allow us a better understanding of the risk for specific diseases and help in identification of appropriatestrategies for ramping up prevention and containment efforts. To achieve this,spatial big data such as social media, electronic health records, satellite remote sensing, and personal sensors, are being used to advance the science of public health.

BlueDot, an AI company, used data to predict the emerging risk of a potential outbreak of COVID-19 in the Hubei province of China on 31st December 2019. They also went on to correctly predict the first 8 out of the 10 cities where the virus would spread based on various transport, environmental and other factors. This highlights how critical an information of this nature could have been to break the chain of transmission early on and prevent the spread of the disease.

Globally many efforts are presently ongoing for real-time syndromic surveillance systemto detect disease outbreaks earlier by classifying health-related geotagged tweets that allowed for the geo-visualisation of health symptoms. Going forward these systems aim to improve the surveillance system by incorporating disease-specific information (e.g., mode of transmission) to enhance disease forecasting accuracy. With major changes in determinants of environmental quality in recent times, the effect of climate change is also becoming important in disease resurgences and management. Facebook has developed highly detailed population maps to assist health organizations, researchers and universities understand as well as tackle disease outbreaks and plan public health interventions. As per the research the American Red Cross is one of the users of such maps.

Aside from population-level surveillance, efforts are also underway to monitor pathogens and their transmission. Under the Global Virome Project developed by scientists at EcoHealth Alliance, unknown viral threats are being analysed. Combining factors such as deforestation, land use, level of wildlife diversity, population density, human-wildlife interaction, etc. a map was created. They claim that with this data they have identified potential hotspots that could kickstart future pandemics if timely counter measures are not taken.

AI for screening and diagnosis

The COVID pandemic has left millions infected across the globe and a larger proportion that have been exposed and at potential risk of disease outbreak. When millions in scattered communities across the globe are at risk, screening them to ascertain their risk to plan interventional strategies becomes a herculean task, not to mention a very resource-intensive one in a fragmented and under-resourced sector.

AI has already been deployed to assess chest CT scans, MRI scans and X-Rays to screen for potential SARS-Cov-2 infection in patients, especially in early stages of the disease where the scans might appear normal on visual inspection. Studies are also underway to identify infection and disease progression from a patient’s cough pattern. AI enabled interventions, such as these, can ramp up the screening efforts and connect the patients to appropriate care for their treatment and for prevention of infection from spreading further.

AI-based screening solution have already made inroads in screening for chronic disease and NCDs. In early 2018, IDx-DR, an AI-based software algorithm for analysis of images of the eye, received US FDA approval, marking a historic moment in healthcare. IDx-DR achieved 87.4% accuracy rate while detecting ‘more than mild’ diabetic retinopathy. Closer home, Google partnered with leading eye care chains Narayana Nethralaya, Aravind Eye Hospital and Sankara Nethralaya to train its AI system on the detection of diabetes retinopathy, especially using low-quality images.

In India considerable focus is on AI in cancer due to a high incidence to mortality rate. According to an EY report, 55% of the breast cancer cases in India were detected at late stage compared to 11% in the UK. Similarly, for cervical cancer, 85% of the cases were detected at the third or fourth stage, against 25% in the UK. A large number of resulting mortality are avoidable with awareness and early screening. Having low cost, minimally invasive screening solutions can democratize cancer care across strata. Niramai Health Analytix, and OncoStem Diagnostics are two startups offering AI-based solutions in this space.

Ten3T’sAI based palm-sized cardiac care monitor, Cicer, tracks ECG, respiration, pulse, temperature easier and on real-time basis. The data is streamed to the doctor or at the clinic for monitoring the patient. It is adapted for various clinical, nursing and in-home applications.

To overcome the lack of infrastructure and expertise for reading pathology samples, Sigtuple leverages AI and machine learning to analyse blood samples thereby helping hospitals and healthcare centres improve the speed and accuracy of blood reports and save crucial time, while at the same time decreasing the cases of mis-diagnosis. Pune based Optrascan does a similar task with their digital pathology solution that can replace microscopes in laboratories. As policy frameworks are set in place, the screening and diagnosis by AI is positioned to aide a technician or clinician requiring their consent and sign-off for further action.

AI for precision medicine

Precision medicine is being made possible by person-specific data (eg clinical, nutritional, behavioural) and information on personal genomes. This will find application in disease management in which assessing the best treatment regime to follow based on individual drug susceptibility profiles can be crucial to ensure adherence, treatment success, and minimizing resurgence of disease. TheApplication of deep learning, a variation of machine learning, to genomic datasets is an exciting area that is rapidly developing and is primed to revolutionize genome analysis. It extracts novel features from 3 billion base pairs of human DNA faster thereby providing invaluable insights which would have taken months, or years, to identify.

Orbuculum has brought together AI with genomics to predict cancer, diabetes, neurological disorders, cardiovascular diseases in a fast and cost-effective way. It extracts meaningful information from the enormous amount of genomic data generated globally and utilizes it to understand the genetic basis of many life-threatening diseases.

In drug development, insights into the patient’s genomic data can ascertain the drug, or the drug combination, that they are likely to respond to without much side effects. This is important from the patient’s standpoint of immediate relief from suffering but also for drug manufacturers who are developing and testing newer drugs for a particular population. The focused genomic information will enable them to select the drug which is most likely to succeed in the target population. The clinical trials on the drugs can also be more efficiently planned and executed thereby saving them millions, if not billions, of dollars.

Deep learning is also making substantial contributions to the understanding of gene regulation, genome organization, and mutation effects which will be useful in disease assessment and management.

AI in Drug Design, Discovery & Development

Another area that would benefit from AI in healthcare is drug discovery. The task of developing a new drug conventionally takes years to fructify and many billions of dollars in investment. This sector is expected to benefit from the quick identification of drug targets and prediction of structures of potential drug molecules due to AI engagement throughout this pipeline. Atomwise’s algorithm scans through a database of molecular structures through its AI platform for drugs which could potentially be safe and efficient against Ebola. Their success in identification of two potential drugs would conventionally take months or years to accomplish.

In case no known molecule is identified, the AI will be able to make suggestions on what the structure of the drug should be and what kind of physio-chemical characteristics it should have to enhance the change of its success. Additionally, AI-based simulations can be used to assess whether prospective drugs will be effective before going to a full-on clinical trial also.

DeepMind, an AI arm of Google’s parent Alphabet Inc., outperformed biologists at predicting the shapes of proteins; this information is vital for efficiently speeding up the designing, discovery and development of new drugs. However, DeepMind simulation doesn’t yet produce the atomic-level resolution that is important for drug discovery. This has been a challenge for biologists and AI could be used to scan millions of high-resolution cellular images more than humans could ever process on their own, and also to decipher other appropriate interventional molecules[1]. In the context of the present pandemic, AI is being deployed to identify and test from millions of potential drug molecules to select the ones with a higher chance to succeed.[1]AI could give Big Pharma a run for its money.

AI Mental Health

The social distancing norms and lockdowns during the ongoing pandemic has had various psychiatric, psychological and psycho-social behavioural effects on people as communities were forced into self-isolation and quarantine amidst fear, stress and an unfamiliar lifestyle.

Before the onset of the pandemic it was estimated that about 15.5% of the global population is affected by mental illnesses, and those numbers are rising steeply during the present situation, the numbers for which currently unavailable. A large fraction of these require treatment but more than half of these go untreated.Diagnosis of mental health disorders are based on an age-old method that can be subjective and unreliable. In addition, the lack of adequately skilled practioners can lead to many of the patients being missed. Now machine learning technology is able to precisely detect day-to-day changes in speech that hint at mental health decline is being developed. Picking up behavioural and cognitive changes can be crucial for timely intervention. For example, sentences without a logical pattern can be a critical symptom in schizophrenia. The perspectives pertaining to the shift in the tone or pace can hint for mania or depression, and memory loss is also a sign of both cognitive and mental health problems.It is predicted that in the next decade will solve ‘specific’ problems with high accuracy. AI’s speech & image recognition is expected to be 100% accurate, and in 5-10 yrs, AI speech recognition will be better than human.

Touchkin is a predictive healthcare app which records parameters as sleep, activity, and patterns of communication through sensors and smartphone to identify changes in behavioural pattern. Whereas, World Well-Being Project (WWBP) analyzed social media with an AI algorithm to identify language markers of depression. These markers could predict depression up to three months before the person receives a formal diagnosis. Furthermore, facial expression, use of certain words, tone or language could indicate suicidal progressions. With remote consultations and therapies becoming common place, there is a good chance for patients struggling with mental health issues to access care in a timely manner.

Since 2012 the medical data has been digitized in India in public health systems and an equivalent patient data also exists in the private sector. Sadly, this valuable information that can help build and train many AI/ML algorithm for advancements inpatient care is non-uniform, siloed and inaccessible. Guidelines of usage, safety nets, and frameworks need to be put in place so these can be openly accessed and utilized.

The inevitable change in the healthcare ecosystem is set to redefine, reshape and reimagine this space to keep pace with the emerging challenges and evolving needs in the domain.

IT service firms are extending their AI and ML capabilities to healthcare by partnering with hospital chains for domain expertise, and more importantly, curated data. For example, NTT DATA Services, a Japanese technology firm, in partnership with Pune’s Deenanath Mangeshkar Hospital last year, achieved a 170% higher detection rate in its proof-of-concept AI based solution to diagnose emphysema, a chronic condition of the lungs, as compared to traditional systems. People are participating by volunteering their curated data to develop AI solutions in exchange for early access to algorithms for making their data actionable efficiently. Microsoft partnered with SRL Diagnostics to source 1 million biopsy samples from confirmed patients to train its AI to detect cancer. Theyhave also partnered with Apollo Hospitals to build an AI system that can detect heart irregularities in patients, to give them a health score.

Although collaborative projects and alliances such as these will ensure that these AI solutions are sharpened for diseases analysis and they survive in a competitive market, however data assess will need to be assured and facilitated for the development of neglected diseases like tuberculosis, malaria etc. which are not driven by market economics but remain a major healthcare problem in developing countries, like India. For such diseases open access to requisite data maybe the only way to attract AI solution developers.

To conclude, we’re at a turning point in healthcare delivery with increasing AI based interventions. The inevitable change in the healthcare ecosystem is set to redefine, reshape and reimagine this space to keep pace with the emerging challenges and evolving needs in the domain.

Dr Swati Subodh is a scientist and healthcare professional in the field of Infectious Diseases for nearly 20 years. Her work spans from basic research to identification of high potential innovations for better public health outcomes. She has published widely and has led her research team in various government and industry supported projects. She has written forvarious media platforms; delivered talks, including at TEDx, UN Headquarters, and a keynote address. Additionally, as an entrepreneur, a guest faculty and member in national institutes she evangelizes #bridgingsilos to accelerate ideas to action.

InnoHEALTH magazine digital team

Author InnoHEALTH magazine digital team

More posts by InnoHEALTH magazine digital team

Leave a Reply

Connect with us now

%d bloggers like this: