The art of building a "continuously learning" diagnostic tool

separator

Do medical devices change? Yes, of course. Better and better new versions come out. But is the piece of equipment that you purchased two years ago any smarter today than it was then? Can it do more than what it did two years ago? That’s impossible, right?

Wrong!

Welcome to SigTuple, the maker of continuously improving screening and diagnostic devices.

So, how do we do it?

Let me explain. Lets take the case of Shonit, our flagship product for peripheral blood smear analysis. The hardware that you need to use it is standard – a microscope, a cell phone, and the adapter which attaches the cell phone camera to the microscope eyepiece. Images of the blood smear slide are captured through the same setup over time. But, with every sample that you analyze with Shonit, our backend systems learn a bit more about blood – maybe a new type of cell, or a different form of pathogen. Every element in those blood images that our current systems are unable to classify, goes to our in-house panel of doctors – pathologists in this case. They help us recognize it. The learning goes back into the artificial intelligence platform. It becomes a bit smarter. Next time we see something similar, we know exactly what it is! So, today we may be able to identify one strain of malaria and flag it. Tomorrow, we identify three other varieties. And all this without the user needing to change anything. A small lab without a dedicated pathologist can now perform non-routine blood tests, without having to send the sample to a bigger lab. It is the patient who benefits in the end.

Pretty neat, huh?

It is time the medical industry meets the power of the cloud. The regulatory landscape also needs to change to accommodate such self learning, continuously improving systems.

Image source: LinkedIn Pulse