Notes 2021, episode 8
Updates from a relatively quiet month
A short update this month: August was a quiet month during which I visited some of my family, and the procedures of travel — all those swabs! — took their toll. So here’s a brief overview of what happened.
- The new Data/Technology Lead in my team, Amadeus, started on 1/8 and he’s got quickly up to speed with the team and the new sector he finds himself in. I now have a full team (as far as staff is concerned), so recruitment is off for a bit, although we might be getting pools of contractors to help on specific projects
- We completed what’s ended up as one of our most interesting projects to date, the forecasting of (needless) hospital long stays that was pitched by Gloucestershire Hospitals. What’s brilliant about this project is that it really showed how AI can contribute to addressing a problem through what — I believe — AI does best: providing clinical teams with useful information to support their decision process. I wrote a blog post and we also published an in-depth case study, alongside the source code.
- The other project that is approaching completion is the creation of a tool for the semi-automatic alignment of CT scans, which we’re helping develop for George Eliot Hospital in Nuneaton. Manoj Srivastava, the radiologist that came to us with the idea, was also recently interviewed by Digital Health as part on a broader report on AI in medical imaging, and he mentions the work we’re doing together.
- Some delays in data access put on hold our project about using AI in shortlisting to control for bias, which we are working on with the London Talent Team of NHS England and Improvement. One thing I’m definitely developing is a thick skin for these drawbacks — Information Governance is hard, especially when there are multiple parties that own the data or are meant to work on it — but there are usually good reasons why data sharing is hard, so we’re here to navigate those processes and make sure that data access is always secure, ethical, and properly governed. The project will likely resume in October.
- We kicked off the Bed Management project that Kettering General Hospital pitched a few months back. This is another interesting issue of optimisation of resources in what is an issue that affects both the patient, in terms of quality of care received, and the hospital, from a logistics and financial point of view. We hope that AI could provide a way to better manage bed occupancy.
- We’ve been also working to better define three problems that have come to us through various routes: a project to see if AI could improve the current accuracy of Acute Kidney Injury predictors at Leicester’s Hospitals is in the pipeline already; we’re working with Mor Rubinstein’s team at Parkinsons UK on unleashing AI on their brain data bank; and with Queen’s University Belfast on building a classifier for phenotyping in Acute Respiratory Distress Syndrome.
- About to kick-off: our first two-hospital projects, about using NLP to improve automation in Clinical Coding, which we’ll run in parallel at Kettering General and Royal Free Hospitals by providing two local teams with extra data science and project delivery resources. Very excited about this one.
- My team of data scientists and technologists have also been working hard with a variety of partners, mostly NHS Trusts, with roles ranging from the advisory to the actual hands on set up of data pipelines to enable the launch of projects. More on this soon. We’ve also offered the AI Award team a way to better support their applicants with data science expertise, so that we can help increase the overall understanding of what AI can and cannot do.
- I finally had a good chat with Ahmad Barclay after so much tweeting and totally admiring his census dataviz.
Oh, and the tweet below really pleased me 🙂 It’s going to be 10 years since I started the newsletter next year, and it’s good to see people still appreciating it.
via Web of Weeknotes https://weeknot.es?source=rss—-424e247e31cc—4
September 1, 2021 at 11:56AM