We`re developing software for different types of optical microscopy (fluorescent and non-fluorescent).
Our algorithms mimic what the human eye does when a researcher analyzes an image.
This approach does not require large annotated databases on which to train a model. These types of databases typically do not exist for discovery research.
Our validation process is straightforward and fast – we compare our algorithm with human analysis on a fraction of the data.
Our previous software achieved more than 95% agreement with human scoring. This allows a short development time to a validated, accurate solution.