20th Meeting "The Imaging Data Commons as a platform for reproducible reasearch in digital pathology"

Please join us Friday, 09/15/2023 11AM ET for a talk titled “The Imaging Data Commons as a platform for reproducible reasearch in digital pathology”. When developing machine learning (ML)-based solutions in computational pathology one of the often underestimated but major challenges is to ensure their reproducibilty. The National Cancer Institute (NCI) Imaging Data Commons (IDC) is a new cloud-based repository within the US national Cancer Research Data Commons (CRDC), which provides more than 120 cancer image collections according to the FAIR principles and is designed to be used with cloud ML services. In our work, we investigated how the IDC and cloud ML services can be used in combination to facilitate reproducibility in computational pathology research. For this purpose we developed a use case, in which we trained and applied a tissue classification model to distinguish between healthy tissue and two different tumor subtypes. We used different datasets from the IDC and implemented our experiments using Google Colaboratory and Google Vertex AI Notebooks. This talk will be about our experiences with setting up a reproducible computational pathology pipeline and summarize some of our pitfalls and learnings.

Presenters: Daniela Schacherer, M.Sc.: Daniela Schacherer has a background in Molecular Biotechnology and Computer Science and is working as an Image Analysis researcher with Fraunhofer MEVIS in Bremen (Germany) since May 2020. She is part of the IDC team since July 2020 and in the project focusses on the work with digital pathology images. André Homeyer, PhD: André Homeyer is a senior scientist in computational pathology at Fraunhofer MEVIS in Bremen, Germany. He has been working in this field for more than 10 years and is the IDC project lead on Fraunhofer MEVIS side since the start of the project in 2019.