Preoperative glioma project

Challenge

One of the market leading healthcare companies needed a model to segment and analyze preoperative brain gliomas based on brain MRI scans. According to the project brief, the model was to segment identified brain areas associated with glioma occurrence and provide quantitative data. The client provided unlabeled brain MRI studies for the project.

What we did

  • We developed and trained the model from scratch.
  • We implemented and performed the process of labeling brain
    MRI studies provided by the client.
  • The model was created using advanced machine learning techniques, deep learning in particular.
  • We developed an algorithm that extracts three regions in a brain MRI image: tumor, tumor enhancing area, and tumor enhancing area in contrast. The algorithm also provides extremely valuable quantitative data on the volume of each area.
  • The entire process takes place in an automated manner.

Results

As a result of our project, the client received an algorithm that allows for a precise and fully automatic analysis of MRI brain images showing gliomas. It is a tool that automatically and accurately segments three brain regions associated with the presence of glioma. In addition, the algorithm provides objective data on the shape of each region and its volume.