Digitization of histological slides and use of AI

Neuropathological diagnosis is largely based on the examination of tissue sections. Tissue collections from clinical practice are thus often accompanied by histology slides. The staining applied to the tissue during preparation of these slides enables microscopic observation and identification of characteristic lesions of the underlying diseases.

Traditionally, histology slides were examined manually by pathologists, but the introduction of digital pathology techniques, now largely AI-powered, makes it possible to rapidly process a large volume of prepared slides. This approach is well suited to large collections like the Geneva Brain Bank, which includes over 200,000 tissue slides.

The purpose of this module is to digitise the histology slides and to train AI tools to classify them by disease and locate the lesions. We will establish proof of concept based on cases of neurosyphilis. The final product of this work will be confirmation of the original diagnosis, potential reclassification of certain cases, and localisation of the regions of interest on the tissues for identification of molecular biomarkers and characterisation of the pathogen (Treponema pallidum).