About the talk
00:35 Problem of working with EHRs
00:57 Specific Challenges
01:25 Data Harmonization
02:50 Validation tool
I am a PhD student at Cambridge working on machine learning problems for healthcare. I am part of the AI group in the computer science department, but I am also a medical student (my studies are paused between the 4th and 5th clinical years to allow me to complete this PhD). I have previously completed parts IA and IB of the pre-clinical medical and veterinary sciences tripos, part IIA of the engineering tripos and one year of clinical medicine.View the profile
Dr Ari Ercole became an anaesthetist after completing a PhD in physics at the University of Cambridge. He divides his time between intensive care, anaesthesia and research. His particular focus on data-driven research: Developing novel analytical techniques including machine learning and feature discovery for intensive-, acute- and perioperative care.View the profile
Hello. I'm interested. Welcome to my presentation on. I see Unity. This is a project that was initially developed at the Milan. Chris cool. Cat dates pain in January and contributions to the community and its Naples and increasing amounts of waxing machine learning. Many of us will be all too familiar with the challenges of working. Single mom, that's very difficult. This is something that we'd really like to change. So the central question is, Tommy ultimate, some of the process is not
showing cross States, that's I'm progressing. Dimensions and specific examples when matching willfully a child that says we can have differences and formats labeling. The same variable language barriers units in a similarly named different measurements, Federal types granularity, and finally variation, in the ditch distribution States population differences. Say we've developed the general framework for the integration of data amuse. The I'm some UMC
as defined by various ontology to that corresponding variables. In each stage set a minimum number of single cat to edit. So you have to make to get a match. So we can see that in some cases, the good specificity, the many variables failed to sleep PC and the Prothrombin time. So We can possibly Medicated by also considering the distributions. So we tested some simple tests and interquartile range of a lot. To show some examples of your system in action. This is an example of a correct Maps based on their string and distribution and sometimes after I
provided a useful, sanity check. So this is a genuine for pH but some reason that the distribution is quite strange in minute, And finally, we developed an interactive tool that can be used by conditions to buy a 5.4 matches. So they can quickly accept or reject as much as to generate. Harmonized a squirrel with its from birthday too. Thank you for listening to my talk and I'd love to take questions Lisa.
Buy this talk
Buy this video
Our other topics
With ConferenceCast.tv, you get access to our library of the world's best conference talks.