Aignostics’ foundation model for histopathology takes the lead on key benchmarks

We used the holidays to benchmark our foundation model RudolfV and are humbled to share that it achieved higher accuracy compared to published alternatives on key public datasets.

The strong performance of our foundation model is particularly noteworthy as it was not trained on the highest number of slides in the comparison, while still covering H&E, IHC, and other stains, across almost 60 tissue types and 15+ labs in Europe and the US. We attribute this to our “pathologist-led” development approach and careful data curation as described in our preprint (linked below) and are excited to see how this approach will scale over the next months as we ramp-up the dataset and model size.

We are also excited to announce that we will roll-out our foundation model as the “base model” across all of our histopathology-related client work over the coming months, which should not only further improve accuracy of our solutions and time-to-model for new developments, but also open up completely new opportunities for applications.

Lastly: Rudolf Virchow was not only the founder of modern pathology, he was also the very founder of the institute our labs and most of the data for this work originate from. We therefore name our model "RudolfV" and join in honoring his groundbreaking work and innovative spirit.