Radiologists and pathologists can have a profound influence during a patient’s therapy. The analysis of medical images or pathology slides is a very complex and time consuming task, and it requires years of specialized training beforehand. Every system that can speed up this process means treatments get cheaper and ultimately more lives can be saved.

Computer vision with deep learning proved to be a good solution to this problem. We developed a model that learns to annotate medical images, based on historical pathologist annotations. A localization score (FROC) of 89% was reached using this technique.

This tool now does a first analysis and lets the pathologist or radiologist focus its attention (and time) on relevant regions. The perfect example of a case where AI goes hand in hand with humans!