Role of AI in Health Tech

At the second edition of DQDeepTech Virtual Event themed Unlocking the Next Tech Frontier innovators in the healthcare sector gave their insights on the role of artificial intelligence in every stage of healthcare from clinical trials to diagnosis to treatment and also the impact of technology in treating mental health.

Niranjan Subbarao, Co-founder, Cyclops Medtech, said AI is evolving in developing drugs and diagnosis. We are using eye tracking, computer vision, and AI/ML to build diagnostic and therapeutic products in the areas of neurotology, and ENT and audiology. The core product captures and analyzes the eye movements. It helps the neurologists and ENT specialists find out the underlying root cause behind dizziness. A stroke or tumor in the brain can also cause dizziness. It is important that a thorough, differential diagnosis is done, and the root cause is put in two different buckets.

We extensively use computer vision to analyze the pupil's image. We also have a consumer version of the product. We are using deep learning to make a model. We can map the eye movement and arrive at diagnostics. When it comes to actual interpretation of the eye movement, we are building ML models to help clinicians connect the dots.

In the healthcare scenario, it is important to analyze the patient's journey. There are four stages -- detection, intervention, prevention, and management. There is a definite requirement of AI-based apps. AI usage is higher when we get to prevention and management. As we get to detection and intervention, it starts reducing.

Ravi Chivukula, CEO, Hearthealth, said that the use of AI needs repeatability and accuracy in diagnostics. We are using it in imaging analysis. We go for second opinion to ensure the diagnostics taken by one doctor is correct. Also, there are overloaded health systems today. How many doctors are there today? In India, there is GDP spend of under 2%, and so, we have a huge shortage. AI can provide some assistance as support in remote locations for telehealth environments. Going into the drug discovery phase, it modifies the metabolism.

Today, we work in imaging studies for impact of drug on organ, and see progression or regression. These use cases are relevant, apart from hospital side. We also need support from the operational side. The apps are vast.

Talking about mental health, Ms. Laina Emmaneul, CEO, BrainSight AI, said we are pioneering new tools for understanding the human brain. Our goal is to help patients shorten the path to recovery and good health by taking the leap from informed estimation to data-assured diagnosis and data-predicted prognosis. We are creating a Google Map of the brain. We overlay the traffic map using functional magnetic resonance imaging (fMRI), which actually helps building the traffic map.

In mental health issues, there is potential for using the traffic map of the brain to understand the issues. There may be patterns, such as schizophrenia. You can use imaging to help typify patients, and predict the path they may take in future. A lot of these disorders are extremely debilitating for the patient and caregiver. We have to do planning, especially in the early stages. There is work being done on rehabilitation of patients such as deep brain stimulation, non-invasive stimulation, etc. When we create traffic patterns, we can also say these are the areas of the brain that need stimulation. Rehabilitation also requires pattern matching.

Mental health is both neurological and biological. There can be lot of data, and also from the brain. Then, we can find even better, nuanced patterns to make drugs and provide rherapy. Lot of work is done in data mining and data pattern matching to help the patient.

Infrastructure will take a huge haul. Brain is an unexplored territory largely. To even extract a brain signal is very difficult. We need to democratize this. You need the infrastucture that can allow this. You need to do infrastructure optimization. In the coming years, infrastructure will play a key role in how this can be democratized.

​Ravi Chivukula, Hearthealth, said with so much data flowing across, there is learning at the edge and at the core, and they need to be connected with data pipes. The digital infrastructure should be able to handle this data -- compute and network. The latencies need to be really small.

Remote patient monitoring and radiology are two examples. You probably have lot of time in radiology. For remote patient monitoring, there is risk of spreading the infection. That kind of infrastructure has to evolve so that we can handle remote patient monitoring. India is among the largest data-hungry nations. Compute and network infrastructure have to be significantly scaled up as the pickup of AI happens. That can have real impact on health outcomes.

Talking about security for patient privacy, Niranjan Subbarao, Cyclops Medtech, said when we talk of AI, it is all blackbox architecture. When it comes to healthcare, this needs to be tackled. We can move away to a glassbox or whitebox. Physicians will clearly need to know that the result will be accurate, and continue to be so. When we have blockchain with AI, a lot of security aspects will be catered to. We have the Ayushman Bharat initiative with sandbox architecture. It is built in secure manner. The same can be extended to AI-based apps. All healthcare apps in India have some regulatory guideline.

Ms. Laina Emmaneul, BrainSight AI, added that explainable AI is there. We can have a 3D twin of any organ, which is like a glassbox. We can show that to the doctor. That will probably help people understand how AI is being used.

Ravi Chivukula, Hearthealth, noted the machine can intervene at rapid pace. There is a transition to decision support. AI does not take any decision, and only serves as a recommendation system. It will naturally impact the regulatory process. From the privacy preserving nature, federated learning is coming out with models. Only the relevant information is transferred. Security of patient data is preserved, with privacy within the sandbox. The AI part of this is done elsewhere.

From skillset perspective, AI needs lot of inter-disciplinary skills. There should be strong connect between healthcare system and technology system. IIT Kanpur has set up a hospital inside the campus. IISc Bangalore is also doing it. There are lot of inter-disciplinary capabilities that one has to develop. We can also bring all of this safely. Bringing it in graded fashion will shape the future.

Ms. Laina Emmaneul, BrainSight AI, added that their work is very inter-disciplinary. That is one policy thing that we would like to see, going forward. Cyber security and HIPAA compliance are also there. That would be another set of recomendations.

Niranjan Subbarao, Cyclops Medtech, said we have large data sets. We need to have a more disciplined and structured approach to ensure data integrity is good. The quality of data is extremely high. The AI model can produce results that are as good as the data being fed in. There needs to be underlying discipline regarding gathering of data.

In future, as we move forward, AI will become more relevant in disease detection and intervention stage. We need to ensure that doctors are part of the equation. We need to be cautious that AI remains as an aid to diagnosis, and not diagnosis in itself. AI is going to make inroads into disease detection and prevention.

Ravi Chivukula, Hearthealth, felt AI can help in healthcare. AI needs to be at small interventions at the right time that can give added improvement in outcome. From clinical and operations sides, we will see it take on more.

Ms. Laina Emmaneul, BrainSight AI, said there are conversations around ethical AI and explainable AI. There will be more focus on including patients in decisions that support them. We will be able to find more targeted uses of AI. It will be ubiquitous.