Elementary, my dear Watson

One of the enduring mysteries in Arthur Conan Doyle’s delightful Holme’s series is that whether Watson was needed at all. However, it transpires in the text that Watson is more like a second person to whom Holmes is speaking with (much like a sounding board).

Purists although would point out the truth hiding in plain sight.

In real life, the modern day Watson is far from perfect, and its cheerleaders sound even dumber.

In a scathing write up in Wall Street Journal, the reporters have laid bare the hype behind IBM Watson that was supposed to have revolutionised cancer treatment. Doctors from a reputed hospital joined the project to “provide direction” followed by a series of photo-ops. The word out was that we were, at last, on the brink of revolutionising the healthcare as we knew it.

To cut a long story short, it didn’t happen. It was being used for marketing by prominent hospital chains with “opinions” sought from the “latest literature”. There is NO scientific evidence that it made an iota of difference.

I have no idea what is the problem that they were trying to solve in the first place. Everything is available off the shelf, and it is incredibly easy to start off a chemotherapy regime. The promised efficiencies of the scale never materialised. Watson was supposed to tap into existing electronic medical records so that it can pre-fill and auto-populate the fields. However, these are limited to restricted geographies and hence do not have worldwide relevance. They do not have a mechanism to quantify the outcomes and measure responses. How would they know what works and what doesn’t? If the results have to be determined by the clinician, what is the point of having this expensive system in the first place? Another issue with it is its poor adaption to clinical ontology. I have seen and heard about the way it struggles with building up the “clinical dictionary” that makes it difficult for it to adapt to hand-written notes. It is likely to lose out all the context of patient’s clinical history. How accurate would its recommendation be if there are significant gaps in the patient’s verbal history? Clinicians often make educated guesses and implied understanding of patient’s issues.

I remain sceptical about the broad-based “research” that multinational corporations push out. There seems to be no truth in the advertising. The original biomedical research has plateaued concerning outcomes. If we have run out of targets, we dig deeper to find new ones! How much do we need to “understand” about pathogenesis? Let us find something worthwhile from what we have discovered. We require an audit of existing “research” to prevent overlaps.

Instead, we need similar and parallel approaches in patient support and ancillary services. The use of mobile technology with adequate privacy safeguards to assist affected patients is promising.

Even from the perspective of Radiation Oncology, we need more to understand this fascinating science. Radiation Biology hardly ever scores funding; for example, we need a scoring system to quantify plans and qualitative analysis for clinical outcomes. We also lack standards to match it with reported clinical results.

The undue focus on the biomedical research will rob us away from our real targets- the patients. In the quest for mutations, we forget its the real people who need good palliative care and support system.

Imagine, if we could channelise the millions of dollars for something constructive! I strongly feel that it is a wasted effort. AI in healthcare can wait.