AI has demonstrated the capability to be a valuable companion to pathologists, with respect to meeting accuracy levels of pathologists, removing bias and increasing diagnostic throughput. However, its adoption to clinical practice is still disputed with sources of resistance stemming from several facets of clinical practice, one of them being the transparency and domain understanding of the recommendation. In our approach we represent the content of histology images through a hierarchy of biological entities. This representation captures information from cell nuclei and the surrounding tissues, as well as how they are connected to each other. We combine this graph representation with a specific neural network architecture and domain concepts to create interpretable and domain-recognizable predictions. In the presentation we will demonstrate the power of the approach in breast cancer subtyping and heterogeneity quantification, as well as in prostate cancer simultaneous segmentation and scoring.
Maria Gabrani is at IBM Research Zurich since 2001. Her research has focused on AI technologies for Healthcare and Lifesciences applications. She currently has 3 major roles. She is an IBM Research strategist in Accelerated Discovery for Healthcare and Lifesciences, specifically in the areas of oncology and biomarkers discovery. She serves also as member of the Exploratory Lifesciences Council of IBM Research. Her local, IBM Research Zurich, role is managing the Cognitive Healthcare and Lifesciences group that that has 3 subgroups, namely, computational pathology, computational single-cell research and computational cancer immunology.
Before joining IBM, from 1999 until 2001, Maria worked for Philips Research, in Eindhoven, The Netherlands. She holds a Ph.D and MS degree in Electrical and Computer Engineering from Drexel University, Philadelphia, PA, USA, and a Diploma in Electrical Engineering from Aristotle University, Thessaloniki Greece. She has received numerous awards including several Outstanding technical Achievement awards, from IBM Corporation, and Technical Achievement Award, from IBM Research Division and numerous best paper awards, from NASA/Goddard Space Flight Center, Image Registration Workshop (1997), to best paper awards in MICCAI (2020, 2021, 2022) and more than 60 patents.