I built a private system to examine ten years of medical records. The task was to bring the history together, compare information and prepare hypotheses to discuss with doctors. Previous diagnostic conclusions differed, and the overall picture remained unclear.
From documents to analysis
We uploaded the accumulated records. AI and parsers processed the documents: recognising text, extracting dates and measurements, putting the data into a consistent structure and saving it in a database.
I linked structured data with the original text context. This made it possible to compare measurements over time and return to information in the documents when examining a particular question.
Exploring hypotheses
Using the assembled history, we worked with AI to formulate hypotheses and explore possible scenarios for the course of the illness. The aim was to find explanations worth discussing with doctors and bring the relevant material together for that discussion.
My contribution and the outcome
I designed and implemented document processing, storage, search and the bot interface, then used the system to analyse the accumulated history.
In our view, the analysis helped focus the doctors’ diagnostic investigation. Symptoms were also eased. The outcome remains uncertain: the diagnosis has not been definitively confirmed.