E0BA74E8-2B56-4F4A-9045-81AB513E5CE6

Update on Vet AI

Remember seeing cats in a blog post a few months ago?  It’s time for an update!  In this post I introduced Mirzakhani the cat and talked about my hopes of using machine learning for early prediction of kidney disease in cats. 

I started thinking about how to obtain enough data for the task, and I had conversations with my cat community about what data would be relevant.  Cats are diagnosed with kidney disease based on key blood markers which become raised, along with urine properties and behaviour changes.  Weight can also be an important indicator of the onset of kidney disease.  But who has this data?

Philipe, Silverpond's AI cat in residence.

 

Vets.  The clinics that the cats go to, the labs that do the blood work, and surely research organisations such as universities? I went down 2 of these tracks:
 
The vet clinic track

As a foster carer with Tom and Queenie Cat Rescue, I was able to use my network to understand what happens to a cat’s data when they visit the vet.  Clinics often use an externally managed database system, where a history record is created for each visit.  If the cat has tests performed (such as blood or urine), these tests are often sent off to an external lab.  The test results come back in a variety of formats, so far I have seen: PDFs, word documents, .txt files, HTML files.  These files get attached directly to the history record. Fun…! 

I was lucky to get access to a clinic’s dataset, which I scraped from the various formats and collated. I ended up with around 2000 cats’ records over a 5 year period!  Analysis will begin soon.
 
The research track
While searching the internet for a magical database full of all the cat histories my heart could desire I came across something called VetCompass Australia.  VetCompass is an initiative that began in the UK to collate data from vet clinics in order to better study health in companion animals.  Data from individual clinics is merged into a giant anonymised database.  This is exactly what I need!  I got in touch with veterinary academics at Melbourne University, a VetCompass Australia partner.  They were very supportive.  We put in a proposal to access the database for my kidney disease project, and it was approved! Coming soon is 20 years worth of cat data, containing somewhere between 4 – 5 million cats. So. Exciting. Right?
 

I’ve learned a lot along the way so far and we are still only at the beginning of this story.  On Saturday 27th October I spoke about the project thus far at the Melbourne Android GDG DevFest 2018.  Video of my talk coming soon! 

Stay tuned and keep all those paws crossed for a successful outcome.
 
Meow.
 

Share this post

Leave a Reply

Your email address will not be published. Required fields are marked *