AI-based tools for mining electronic health record data

Analyzing electronic health record (EHR) data has already provided us with new insights in medical research and holds great promise for the future.

Unfortunately, many of the EHR systems were not designed with retrospective scientific study in mind and much of the interesting information lies in digital, but unstructured form. The largest source for this unstructured information in the hospital are different kinds of dictations and written notes that describe the patients’ status.

In the video, I quickly go over some of the approaches that we have taken at the Auria Biobank in analyzing such information and show some examples about the promising results we have obtained. Our tools vary from the trivial but very effective regular expressions all the way up to neural net driven vector space representations of words.

The author of this post is Antti Karlsson, PhD, Development Manager at Auria Biobank (picture by Mikko Tukiainen). 

Read more:

Possibilities of machine learning and artificial intelligence in RWE studies