Artificial Intelligence and the future of healthcare

This article was first published in my Times of India Blog Staying Wired.

Anyone reading this post would know how predominant has been the terms ‘Artificial Intelligence’ and ‘Machine Learning’ in this millennium. Though these fields have been under the research radar of scientists from a much earlier period, it is only in the last few decades that AI and ML have matured out of the test-tubes to face real-world challenges. And healthcare is one field which stands to gain from that.

It would not be too far-fetched to predict that the advances in medical science like vaccination and genetics which helped to further human longevity, Artificial Intelligence would bring about the next quantum leap. And as we discuss multiple instances of applications of AI and ML have already started coming to the forefront.

In the detection of cancer, metastasis pathologists are faced with the time-consuming task of reviewing slides of high-resolution images covering every tissue under investigation. The images are in the magnitude of 10 gigapixels and a pathologist must go through each pixel to ensure that he isn’t missing anything important.

AI algorithms reading the visual patterns are expected to be between 5-10% more accurate than pathologists with unlimited time studying the same images. And since being machines they are expected to be more consistent as well. This increased accuracy and consistency can strongly fight off misdiagnoses.

In another brilliant example, Madurai based Arvind Eye Hospital has started testing a Google AI based interface which can provide early against diabetic retinopathy, which if untreated can lead to permanent blindness. However, with just 11 ophthalmologists for 10 lakh people, not everyone is expected to undergo regular eye checkups in India. Such AI interfaces, providing preliminary examination results fast and at low cost, can go a long way to bridge the gap w.r.t availability and reach of doctors.

AI systems have found extensive usage in diagnoses of common ailments along with predicting how long a patient would stay in hospital or chances of readmission and even death.

In China studies conducted on a wide base of patients have shown amazing results. AI systems diagnosed asthma with 90% accuracy as against physicians whose accuracy ranged from 80% to 94%. For gastrointestinal diseases, AI systems were 87% accurate while physicians were in the 82-90% bracket. Thus AI systems can be overwhelmingly reliable in terms of generating a consistently accurate outcome. And hopefully, with further proliferation, it will be cheap and maybe even open source sometime in future.

But that doesn’t mean there are no pitfalls. Anyone with access can fudge the medical records to fool AI systems. And many scientists have already proved so my performing tricks with AI systems which made these machines fall flat. So it can be well said that the introduction of AI will no way reduce the importance of the work a physician or a pathologist does. It only shifts the nature of work.

Pathologists and doctors can let AI systems perform the preliminary investigation while they get look at exceptions and outliers. For the patients, especially in countries like India where competent doctors are scarce to be found, AI systems could be the first line of the interface at primary healthcare facilities where preliminary diagnoses can take place.

We should be aiming at an augmented Healthcare system where physicians and AI can work hand in hand providing the best possible outcome for the patient. With the growing population and ever new complications in diseases arising, the field where humanity needs to embrace the combined might of Machine Learning and Artificial Intelligence most are not self-driving cars or productivity enhancing robots but healthcare.