'AI can help doctors fight cancer, but it’s not meant to replace them'

From mining medical records and designing treatment plans to developing precision medicines, artificial intelligence (AI) is playing a bigger role in healthcare. Shobita Dhar spoke to Rob Merkel, general manager of oncology and genomics, IBM Watson Health, about how the cognitive computing system is working with clinicians, researchers, businesses and governments worldwide, and in India, to solve healthcare challenges that range from cancer to diabetes.

How can AI help the doctor?
There is an overwhelming amount of information available to physicians today. In oncology alone, there are more than 160,000 studies published each year. Yet there are few tools to help doctors draw meaning from this knowledge. With training, AI technologies like Watson are assisting physicians to summarise patient records and synthesise the vast body of medical data — from clinical records to medical literature to genomics and even data from wearable devices like the Apple Watch — so that they can deliver personalised, evidence-based treatment options to their patients.

Can over-dependence on machines make doctors complacent and lazy?

The responsibility for medical decisions ultimately lies with the doctor. Just as physicians have relied on textbooks in the past, a technology like Watson can help doctors make more informed, data-driven decisions, but it is not intended to supersede their expert judgments.

Manipal Hospitals in Bengaluru has been using Watson and a recent study showed that the supercomputer suggested the same treatment options as the hospital's oncologists in 96.4% of 362 cancer cases analysed. This is quite a high accuracy. How was this achieved?
After winning the TV quiz show Jeopardy! in 2011, Watson began training at Memorial Sloan Kettering (MSK) Cancer Center in New York, one of the world's leading cancer institutes. The resulting platform, Watson for Oncology, went through rigorous training and testing with MSK before it was released for any given cancer type.

(It is currently available to support breast, lung, colorectal, gastric, ovarian, cervical and prostate cancers.) So it's not surprising that these MSK-level opinions frequently matched Manipal Hospital's multi-disciplinary tumour board.
How well are oncologists responding and adapting to AI?

Watson for Oncology is already in use in 13 countries. Oncologists are sceptical of almost any new technology, but when they experience Watson directly, they are impressed. A University of North Carolina oncologist who used Watson for Genomics has said it can analyse genomic data and surface an accurate list of targeted treatment options in mere seconds — a task that generally requires hours of time from a molecular tumour board. An oncologist from the Mayo clinic shared that the amount of time to screen a patient for clinical trials was reduced from 30 minutes to 8 minutes.

What are the limits of AI in medical care and what can go wrong?

Doctors should always be at the centre of patient care decisions. AI can go wrong when an underlying assumption is made that it can replace human judgment. Watson is a valuable tool for clinicians — not a replacement for clinicians. We provide insights to oncologists in a very transparent manner so that they can directly access relevant patient information or medical literature which supports these insights.

Will technologies like Watson mean a need for fewer doctors and nurses thereby reducing healthcare costs?

It is physically impossible for doctors and nurses to read the amount of information to stay current in their profession — there are not enough hours in the day. Watson provides an opportunity to partner with doctors and nurses to assist in analysing this information so that they can spend more time with their patients. It is also our hope that Watson can help doctors and nurses deliver efficient and effective care to their patients during the increasingly short period of time available for patient visits. This is especially important in markets like India where there is only one oncologist for every 1,600 patients.



Popular posts from this blog

EMA Recommends Extension of Indications for Ramucirumab

Dr. Hootie Warren takes helm of Fred Hutch Global Oncology

Bristol-Myers Squibb and Clovis Oncology Announce a Broad Clinical Collaboration to Evaluate Combination of Opdivo (Nivolumab) and Rubraca® (Rucaparib) in Phase 2 and Pivotal Phase 3 Clinical Trials in Multiple Tumor Types