Neuropathology remains a periphery in the field of digital medicine, especially deep learning.
Automated assistance systems in neuropathological diagnostics are still pretty poorly developed or not even existent. Neuropathologists still look at tissue through microscopes. However, our research group enters the field of digital neuropathology from the side of physicians bringing pathogenetic, pathophysiologic and histologic insight to the table. We try to solve problems encountered in routine neuropathologic work-up with AI to improve healthcare.
By digitizing histopathological slides and processing them with algorithms it is possible to automate and enhance the diagnostic process. A computer can evaluate more tissue in less time with the same accuracy as humans and therefore lowers the error rate and simultaneously decreases the time to diagnosis. Techniques like CNNs that are already widely used in other fields like autonomous driving could have a huge impact on diagnostical procedures.
A major burden are non-standardized procedures. Therefore as a prerequisite whole pipelines and easy to use web-apps are mandatory. We want to illustrate the potential of deep learning aided virtual tissue slide analysis to predict diseases, their molecular genetic background and disease progression. We will emphasize some of our projects in different subsections of AI in neuropathology which try to combine the potential of deep learning with the usefulness of pipelines in online tools.
Samir Jabari (Institute for Neuropathology, University Hospitals Erlangen)