Project Title:

Federated Learning as a step forward in digital pathology: a multi-centric pipeline and a quality
assurance (QA) platform to support diagnosis of prostate cancer, in WSI of core needle
biopsies, to evaluate artificial intelligence systems (CAD and QC tools).

Histopathological diagnosis, the gold standard for detecting prostate cancer in biopsies, is time consuming and subjected to
the pathologist’s expertise and interpretation criteria. Digital pathology (DP), with the use of whole slide imaging (WSI) may
allow the support of computer-aided diagnosis systems (CAD) which could increase the level of diagnostic accuracy and
turn-around time. However the available approved CE-IVD CADs are not completely validated with heterogeneous routine
series of digital slides (DS) obtained by different institutions.

Our project aims to estimate the feasibility of a digital pathology workflow in real life, focusing on the diagnosis of prostate biopsies, by evaluating the inter/intra-reader concordance with/without CAD support, and to create a Federated-Learning (FL) [Rieke 2020, doi: 10.1038/s41746-020-00323-1] Quality Assurance platform to exchange the inter-centre knowledge.

Burndown Studio, part of NVIDIA Inception program and Italian startup in AI and data sciences, with a significant experience in FL tools, will support the centres in the QA platform implementation.