Automated Classification of Diseases from Medical Reports
Our Projects

Automated Classification of Diseases from Medical Reports

The University Hospital in Berne wanted to relieve a team of highly trained medical pro-fessionals – Zühlke established an AI-based solution.

Executive Summary

  • The University Hospital of Bern wanted to relieve a team of highly qualified professionals in the evaluation of patient files. 
  • Zühlke developed a solution based on machine learning that automatically recognised 74% of the main diagnoses. 
  • After delivery of the solution, Zühlke supported the University Hospital in setting up a data science team including infrastructure.

Algorithms to Support Trained Medical Professionals

Medical patient reports need to be classified with disease codes. At the University Hos-pital in Berne, this process was manually performed by a coding team consisting of high-ly trained medical professionals. The hospital wanted to develop and evaluate machine learning algorithms to automatically classify patient reports in order to support the team in their daily work.

First prototype after only three weeks

To increase the efficiency of the coding team, Zühlke developed a machine learning based solution allowing for an automated classification of medical patient reports that can be employed in an expert in-the-loop setting. In just three weeks, a team of data scien-tists and consultants developed a first prototypical machine learning solution and demonstrated its feasibility.

2 diagonalestriche lightgray

Establishment of data science team

Based on the first prototype, key decision of the hospital makers decided to invest in fur-ther development. Zühlke developed a state-of-the-art solution based on deep learning in another four months and supported the University Hospital Berne in the establishment of a data science team and the corresponding infrastructure.