Artificial Intelligence In Healthcare - Should We Let The Genie Out Of The Bottle?

By Prof. (Dr.) Mahipal S. Sachdev, Chairman, Medical Director & Senior Consultant - Ophthalmology, CENTRE FOR SIGHT Having completed his graduate studies in medicine from AIIMS Delhi, and Post Graduation in Ophthalmology from the Dr. R.P. Centre for Ophthalmic Sciences, Dr. Mahipal is one of the most well known names in the India eye care scenario, and has also been felicitated with the prestigious Padma Shri award from the President of India.

Airplanes are becoming far too complex to fly. Instead of pilots, the planes require computer scientists from MIT. I don't know about you, but I don't want Albert Einstein to be my pilot. I want great flying professionals that are allowed to easily and quickly take control of a plane.I see it all the time in many products’. This famous statement by an even famous personality was precipitated by a couple of Air Disasters in the recent past where the sensors over rode the manual maneuvers of the pilots to plunge the aircraft into destruction and death.

One cannot but help thinking of the Jewish legends, where Golems were inanimate objects made of matter like mud and clay, but magically brought to life, making them prophesiers of AI machines. The interest in AI started to catch the attention of the industry when the IBM computer, ‘Deep Blue’defeated Russian grandmaster Garry Kasparov in 1997. This popularity is becoming more and more relevant with each passing day, as it becomes fashionable to flaunt the AI aspect of technology and related terms like machine learning, deep learning, neural networks, random forests, and unsupervised learning, which are becoming part of the everyday lingo in every field.

We stand today on the threshold of true Autonomous Intelligence in all walks of life. And the Healthcare industry is no exception, with Cognitive hospitals now becoming a reality. We already have deep learning algorithms that can read CT scans, software and programmes that can detect earliest changes of Diabetic Retinopathy in a Retinal Map. According to a 2016 study by Frost & Sullivan, the market for AI in healthcare is projected to reach $6.6 billion by 2021.Oxford University and Yale University in a study in 2017 predicted that by 2053, surgical jobs could be the exclusive purview of AI Tools. In the same year, a robot from a Chinese developer passed the Chinese medical Licensing Exam with distinction. Indeed, heightened patient expectations, burgeoning population and lifestyle and genetic factors affecting our health have made the use of technology and AI a potent and ubiquitous tool in healthcare.

The entry of AI into medicine is accompanied by insecurity amongst the medical fraternity that these AI processes will be better, stronger, faster and smarter, leading to loss of jobs and opportunities for those made of flesh and blood. This fear of
the unknown is coupled by a much more genuine and reasonable fear of the drawbacks and an irrevocable change in the doctor patient relationship. Surprisingly, the other and most important stakeholder in this equation, the patient, is much more amenable to AI entering as the third pillar in healthcare. A recent survey by Accenture saw 29 percent of the respondents saying they would not prefer to use AI, whereas 71 percent said they would prefer AI aided treatment, expecting it to be faster, available 24x7, less prone to errors and biases, and having data of all the past visits on its ‘fingertips’.

In dynamic environments such as medicare where decision making is based on many variable parameters, full automation is a minefield, with high stakes at every step

These indeed are cited as the advantages AI can offer to the healthcare industry. Rapid analysis of humungous data, machine learning aided billing, predictive monetary estimates, functioning as Virtual Health Assistant for scheduling appointments, aiding in rapid and early diagnosis and lately, functioning as Medical Bots and Chatbots ensuring round the clock availability for the patient are the tangible benefits AI can offer.This will ultimately result in decreasing the workload of the medical professionals, circumventing the shortage of trained medical working force and preventing stress and burnout in the medicos. But (and it is a big ‘But’), in dynamic environments such as medicare where decision making is based on many variable parameters, full automation is a minefield, with high stakes at every step.

The main argument against AI rests on three premises, those of Privacy, Ethics and Errors. Any AI application worth its salt needs to have the decision making based on an algorithm developed after analysis of a large repository of data. Such a large database is possible only by meticulous and standardized data capture, data collation and data collaboration. In medicine, legal and regulatory issues preclude free sharing of patient data, so getting a large database to make the AI aided decisions accurate will be a herculean task. Even if achieved, such a large dump of data stored at one single place will be a goldmine for hackers and unscrupulous elements, compromising the privacy of the patients. The second issue is of Ethics and Trust. A small tinkering of an algorithm by vested interest for commercial gains, can turn a benign skin mole into a cancerous growth, aninnocuous spot on the retina into a sinister melanoma of the retina. With such manipulation possible, who will be the Legal watchdog? Till now, FDA and other regulatory authorities have been dealing with drugs, devices and procedures. Will they have the know-how to validate AI algorithms? As the machines keep learning and improving, will every software update entail a fresh regulatory approval? Add to this the ethical issue of decreasing facetime and eye contact with the patient as more and more technology enters our clinics. Thirdly, and most importantly, who will take the blame for life threatening mistakes of omission or commission made in the process of AI aided diagnosis and therapy. Will the AI systems be indemnified for malpractice?

As we find lasting solutions to these vexing problems of changing the mindset of the doctors to be more receptive of AI, getting in place water tight regulations and put safeguards to prevent Automation complacency, my personal roadmap for introduction of AI in healthcare is , what I call, the TRT model of AI Induction.TRT is Training, Research and Therapy. In my opinion, initially AI should make its worth felt in imparting training of Ophthalmology. From simulated case scenarios, to analyzing the individual deficiencies in a student’s surgical skill sets, the applications of AI in training are numerous. Then, as a next step, AI should find use in research, where its inherent capability of data capture and instant memory will make it an invaluable tool. Then, and only then, should we allow AI to enter the hallowed portals of our clinics and hospitals, to help in the domain of actual patient care.

A tandem of Human experience and digital augmentation will ensure optimal use of AI in healthcare. To ensure this, we need to make sure that in our enthusiasm and euphoria, we do not hand over the joystick to the machine. The safety valve of AI has to be ‘man overriding the machine’ in a scenario of conflict. Then and only then will AI be able to live up to the gold standards of Trust and Transparency.